AI Video Tools

Explore the best AI Video Tools — independent reviews, comparisons, pricing and step-by-step how-to guides, curated by Aizhi.

• 

  Outline of deep learning

  The following outline is provided as an overview of, and topical guide to, deep learning: Deep learning is a subfield of machine learning and artificial intelligence based on artificial neural networks with multiple processing layers. It emphasizes representation learning and is widely used in areas such as computer vision, natural language processing, speech recognition, recommender systems, robotics, and generative artificial intelligence. == Ways to categorize deep learning == A field of study A branch of artificial intelligence A subfield of machine learning A subfield of computer science A form of representation learning A class of methods based on artificial neural networks An approach used in computational statistics == History == === Precursors === Cybernetics Perceptron Connectionism Neocognitron Backpropagation === Milestones === LeNet Long short-term memory Deep belief network AlexNet Sequence to sequence learning Generative adversarial network Residual neural network Transformer BERT Generative pre-trained transformer Diffusion model === Related histories === History of artificial intelligence History of machine learning Timeline of machine learning == Core concepts == == Learning settings == Supervised learning Unsupervised learning Self-supervised learning Semi-supervised learning Reinforcement learning Transfer learning Multitask learning Multimodal learning Online machine learning Continual learning == Common tasks == Image classification Object detection Image segmentation Automatic speech recognition Neural machine translation Question answering Automatic summarization Text-to-image model Protein structure prediction == Architectures == === Feedforward and convolutional architectures === Feedforward neural network Multilayer perceptron Convolutional neural network Radial basis function network Residual neural network U-Net === Recurrent and sequence architectures === Recurrent neural network Long short-term memory Gated recurrent unit Sequence to sequence learning Recursive neural network === Representation-learning architectures === Autoencoder Denoising autoencoder Sparse autoencoder Variational autoencoder Restricted Boltzmann machine Deep belief network === Attention and transformer architectures === Attention (machine learning) Transformer BERT Generative pre-trained transformer Vision transformer === Generative and probabilistic architectures === Autoregressive model Diffusion model Energy-based model Generative adversarial network Mixture of experts === Graph and memory architectures === Graph neural network Graph convolutional network Siamese network Neural Turing machine Memory network Echo state network Capsule neural network == Neural network components and techniques == Artificial neuron Activation function Rectified linear unit Sigmoid function Softmax function Embedding Convolution Pooling layer Attention Batch normalization Layer normalization Residual connections == Training and optimization == Backpropagation Gradient descent Stochastic gradient descent Adam optimization Learning rate Loss function Cross-entropy Mean squared error Regularization Dropout Early stopping Batch normalization Data augmentation Transfer learning Knowledge distillation Ensemble learning Curriculum learning == Datasets and benchmarks == CIFAR-10 ImageNet MNIST database Common Objects in Context (COCO) General Language Understanding Evaluation (GLUE) benchmark LibriSpeech SQuAD == Applications == === Computer vision === Computer vision Facial recognition system Image classification Image segmentation Medical imaging Object detection Optical character recognition === Natural language processing === Automatic summarization Chatbot Information retrieval Large language model Natural language processing Neural machine translation Question answering Sentiment analysis === Speech and audio === Automatic speech recognition Music information retrieval Speaker recognition Speech synthesis === Science and medicine === Bioinformatics Computational biology Drug discovery Medical diagnosis Protein structure prediction === Robotics and control === Autonomous car Computer game bot Control theory Robotics === Recommendation, search, and forecasting === Anomaly detection Forecasting Fraud detection Recommender system Search engine === Generative artificial intelligence === Deepfake Generative artificial intelligence Large language model Speech synthesis Text-to-image model === Computer graphics and video games === Deep Learning Anti-Aliasing (DLAA) Deep Learning Super Sampling (DLSS) == Hardware == AMD Instinct AMD XDNA Application-specific integrated circuit Deep learning processor, Neural processing unit (NPU), or Neural Engine Field-programmable gate array General-purpose computing on graphics processing units (GPGPU) Graphics processing unit NVIDIA Deep Learning Accelerator (NVDLA) Tensor processing unit Vision processing unit Wafer-scale integration === Supporting software platforms === CUDA Metal ROCm == Software == === Open-source frameworks and libraries === === Neural network software === EDLUT Emergent Encog JOONE Neuroph NeuroSolutions OpenNN Peltarion Synapse SNNS === Platforms, tools, and deployment === Amazon SageMaker Google Colab Hugging Face Kaggle Kubeflow MLflow ONNX OpenVINO TensorFlow Hub == Algorithms for deep learning and neural networks == Backpropagation Conjugate gradient method Generalized Hebbian algorithm Gradient descent Levenberg–Marquardt algorithm Perceptron Quasi-Newton method Wake-sleep algorithm == Methods and related topics == === Representation and metric learning === Contrastive learning Embedding Feature learning Manifold learning Metric learning === Generative modeling === Autoregressive model Diffusion model Generative adversarial network Generative model Variational inference === Efficient and scalable deep learning === Knowledge distillation Low-rank approximation Mixture of experts Quantization Sparsity === Reliability, safety, and interpretability === Adversarial machine learning AI alignment Algorithmic bias Catastrophic forgetting Differential privacy Explainable artificial intelligence Federated learning Hallucination (artificial intelligence) == Conferences and workshops == Annual Meeting of the Association for Computational Linguistics Conference on Computer Vision and Pattern Recognition Conference on Neural Information Processing Systems International Conference on Computer Vision International Conference on Learning Representations International Conference on Machine Learning == Organizations == === Research laboratories and institutions === Allen Institute for AI Alberta Machine Intelligence Institute European Laboratory for Learning and Intelligent Systems Google DeepMind Meta AI Mila Microsoft Research Vector Institute === Companies === Anthropic Cerebras Cohere DeepSeek Mistral AI OpenAI Stability AI xAI == Publications == === Books === Deep Learning – Ian Goodfellow and Yoshua Bengio Neural Networks and Deep Learning – Michael Nielsen Perceptrons – Marvin Minsky and Seymour Papert === Journals === IEEE Transactions on Neural Networks and Learning Systems Neural Networks Neural Computation == Influential persons ==

  Read more →
• 

  Web performance

  Web performance refers to the speed in which web pages are downloaded and displayed on the user's web browser. Web performance optimization (WPO), or website optimization is the field of knowledge about increasing web performance. Faster website download speeds have been shown to increase visitor retention and loyalty and user satisfaction, especially for users with slow internet connections and those on mobile devices. Web performance also leads to less data travelling across the web, which in turn lowers a website's power consumption and environmental impact. Some aspects which can affect the speed of page load include browser/server cache, image optimization, and encryption (for example SSL), which can affect the time it takes for pages to render. The performance of the web page can be improved through techniques such as multi-layered cache, light weight design of presentation layer components and asynchronous communication with server side components. == History == In the first decade or so of the web's existence, web performance improvement was focused mainly on optimizing website code and pushing hardware limitations. According to the 2002 book Web Performance Tuning by Patrick Killelea, some of the early techniques used were to use simple servlets or CGI, increase server memory, and look for packet loss and retransmission. Although these principles now comprise much of the optimized foundation of internet applications, they differ from current optimization theory in that there was much less of an attempt to improve the browser display speed. Steve Souders coined the term "web performance optimization" in 2004. At that time Souders made several predictions regarding the impact that WPO as an "emerging industry" would bring to the web, such as websites being fast by default, consolidation, web standards for performance, environmental impacts of optimization, and speed as a differentiator. One major point that Souders made in 2007 is that at least 80% of the time that it takes to download and view a website is controlled by the front-end structure. This lag time can be decreased through awareness of typical browser behavior, as well as of how HTTP works. == Optimization techniques == Web performance optimization improves user experience (UX) when visiting a website and therefore is highly desired by web designers and web developers. They employ several techniques that streamline web optimization tasks to decrease web page load times. This process is known as front end optimization (FEO) or content optimization. FEO concentrates on reducing file sizes and "minimizing the number of requests needed for a given page to load." In addition to the techniques listed below, the use of a content delivery network—a group of proxy servers spread across various locations around the globe—is an efficient delivery system that chooses a server for a specific user based on network proximity. Typically the server with the quickest response time is selected. The following techniques are commonly used web optimization tasks and are widely used by web developers: Web browsers open separate Transmission Control Protocol (TCP) connections for each Hypertext Transfer Protocol (HTTP) request submitted when downloading a web page. These requests total the number of page elements required for download. However, a browser is limited to opening only a certain number of simultaneous connections to a single host. To prevent bottlenecks, the number of individual page elements are reduced using resource consolidation whereby smaller files (such as images) are bundled together into one file. This reduces HTTP requests and the number of "round trips" required to load a web page. Web pages are constructed from code files such JavaScript and Hypertext Markup Language (HTML). As web pages grow in complexity, so do their code files and subsequently their load times. File compression can reduce code files by about 40 percent, thereby improving site responsiveness. Web Caching Optimization reduces server load, bandwidth usage and latency. CDNs use dedicated web caching software to store copies of documents passing through their system. Many website platforms, such as SiteGround, IONOS, Wix, and Hostinger, rely on global CDNs and caching technologies to deliver faster page loads across different geographical regions. Subsequent requests from the cache may be fulfilled should certain conditions apply. Web caches are located on either the client side (forward position) or web-server side (reverse position) of a CDN. Web browsers are also able to store content for re-use through the HTTP cache or web cache. Requests web browsers make are typically routed to the HTTP cache to validate if a cached response may be used to fulfill a request. If such a match is made, the response is fulfilled from the cache. This can be helpful for reducing network latency and costs associated with data-transfer. The HTTP cache is configured using request and response headers. Code minification distinguishes discrepancies between codes written by web developers and how network elements interpret code. Minification removes comments and extra spaces as well as crunch variable names in order to minimize code, decreasing files sizes by as much as 60%. In addition to caching and compression, lossy compression techniques (similar to those used with audio files) remove non-essential header information and lower original image quality on many high resolution images. These changes, such as pixel complexity or color gradations, are transparent to the end-user and do not noticeably affect perception of the image. Another technique is the replacement of raster graphics with resolution-independent vector graphics. Vector substitution is best suited for simple geometric images. Lazy loading of images and video reduces initial page load time, initial page weight, and system resource usage, all of which have positive impacts on website performance. It is used to defer initialization of an object right until the point at which it is needed. The browser loads the images in a page or post when they are needed such as when the user scrolls down the page and not all images at once, which is the default behavior, and naturally, takes more time. == HTTP/1.x and HTTP/2 == Since web browsers use multiple TCP connections for parallel user requests, congestion and browser monopolization of network resources may occur. Because HTTP/1 requests come with associated overhead, web performance is impacted by limited bandwidth and increased usage. Compared to HTTP/1, HTTP/2 is binary instead of textual is fully multiplexed instead of ordered and blocked can therefore use one connection for parallelism uses header compression to reduce overhead allows servers to "push" responses proactively into client caches Instead of a website's hosting server, CDNs are used in tandem with HTTP/2 in order to better serve the end-user with web resources such as images, JavaScript files and Cascading Style Sheet (CSS) files since a CDN's location is usually in closer proximity to the end-user. == Metrics == In recent years, several metrics have been introduced that help developers measure various aspects of the performance of their websites. In 2019, Google introduced metrics such as Time to First Byte (TTFB), First Contentful Paint (FCP), First Paint (FP), First Input Delay (FID), Cumulative Layout Shift (CLS) and Largest Contentful Paint (LCP) allow for website owner to gain insights into issues that might hurt the performance of their websites making it seem sluggish or slow to the user. Other metrics including Request Count (number of requests required to load a page), DOMContentLoaded (time when HTML document is completely loaded and parsed excluding CSS style sheets, images, etc.), Above The Fold Time (content that is visible without scrolling), Round Trip Time, number of Render Blocking Resources (such as scripts, stylesheets), Onload Time, Connection Time, Total Page Size help provide an accurate picture of latencies and slowdowns occurring at the networking level which might slow down a site. Modules to measure metrics such as TTFB, FCP, LCP, FP etc are provided with major frontend JavaScript libraries such as React, NuxtJS and Vue. Google publishes a library, the core-web-vitals library that allows for easy measurement of these metrics in frontend applications. In addition to this, Google also provides the Lighthouse, a Chrome dev-tools component and PageSpeed Insight a site that allows developers to measure and compare the performance of their website with Google's recommended minimums and maximums. In addition to this, tools such as the Network Monitor by Mozilla Firefox help provide insight into network-level slowdowns that might occur during transmission of data.

  Read more →
• 

  Mixvoip

  Mixvoip S.A. is a Luxembourg-based telecommunications service provider founded in 2008. The company offers IP telephony, high-speed Internet connectivity, and IT solutions to businesses and individuals. == Company history == In November 2017, Mixvoip expanded its operations to Belgium and Germany. At the beginning of 2019, the company acquired the telecommunications provider Voipgate. In December 2019, Mixvoip was named Telecom Company of the Year at the Luxembourg ICT Awards 2019 organized by Farvest and IT One. A 2024 article in Duke described the company's transition during the 2010s from traditional telephony services to cloud-based communication platforms. In the end of 2024, the ILR published the statistics about electronic communications in Luxembourg, including Mixvoip in the fix telephony section. In July 2025, Mixvoip acquired Crossing Telecom. In 2026, Mixvoip acquired Nomado's portfolio.

  Read more →
• 

  Supercomputer operating system

  A supercomputer operating system is an operating system intended for supercomputers. Since the end of the 20th century, supercomputer operating systems have undergone major transformations, as fundamental changes have occurred in supercomputer architecture. While early operating systems were custom tailored to each supercomputer to gain speed, the trend has been moving away from in-house operating systems and toward some form of Linux, with it running all the supercomputers on the TOP500 list in November 2017. In 2021, top 10 computers run for instance Red Hat Enterprise Linux (RHEL), or some variant of it or other Linux distribution e.g. Ubuntu. Given that modern massively parallel supercomputers typically separate computations from other services by using multiple types of nodes, they usually run different operating systems on different nodes, e.g., using a small and efficient lightweight kernel such as Compute Node Kernel (CNK) or Compute Node Linux (CNL) on compute nodes, but a larger system such as a Linux distribution on server and input/output (I/O) nodes. While in a traditional multi-user computer system job scheduling is in effect a tasking problem for processing and peripheral resources, in a massively parallel system, the job management system needs to manage the allocation of both computational and communication resources, as well as gracefully dealing with inevitable hardware failures when tens of thousands of processors are present. Although most modern supercomputers use the Linux operating system, each manufacturer has made its own specific changes to the Linux distribution they use, and no industry standard exists, partly because the differences in hardware architectures require changes to optimize the operating system to each hardware design. == Context and overview == In the early days of supercomputing, the basic architectural concepts were evolving rapidly, and system software had to follow hardware innovations that usually took rapid turns. In the early systems, operating systems were custom tailored to each supercomputer to gain speed, yet in the rush to develop them, serious software quality challenges surfaced and in many cases the cost and complexity of system software development became as much an issue as that of hardware. In the 1980s the cost for software development at Cray came to equal what they spent on hardware and that trend was partly responsible for a move away from the in-house operating systems to the adaptation of generic software. The first wave in operating system changes came in the mid-1980s, as vendor specific operating systems were abandoned in favor of Unix. Despite early skepticism, this transition proved successful. By the early 1990s, major changes were occurring in supercomputing system software. By this time, the growing use of Unix had begun to change the way system software was viewed. The use of a high level language (C) to implement the operating system, and the reliance on standardized interfaces was in contrast to the assembly language oriented approaches of the past. As hardware vendors adapted Unix to their systems, new and useful features were added to Unix, e.g., fast file systems and tunable process schedulers. However, all the companies that adapted Unix made unique changes to it, rather than collaborating on an industry standard to create "Unix for supercomputers". This was partly because differences in their architectures required these changes to optimize Unix to each architecture. As general purpose operating systems became stable, supercomputers began to borrow and adapt critical system code from them, and relied on the rich set of secondary functions that came with them. However, at the same time the size of the code for general purpose operating systems was growing rapidly. By the time Unix-based code had reached 500,000 lines long, its maintenance and use was a challenge. This resulted in the move to use microkernels which used a minimal set of the operating system functions. Systems such as Mach at Carnegie Mellon University and ChorusOS at INRIA were examples of early microkernels. The separation of the operating system into separate components became necessary as supercomputers developed different types of nodes, e.g., compute nodes versus I/O nodes. Thus modern supercomputers usually run different operating systems on different nodes, e.g., using a small and efficient lightweight kernel such as CNK or CNL on compute nodes, but a larger system such as a Linux-derivative on server and I/O nodes. == Early systems == The CDC 6600, generally considered the first supercomputer in the world, ran the Chippewa Operating System, which was then deployed on various other CDC 6000 series computers. The Chippewa was a rather simple job control oriented system derived from the earlier CDC 3000, but it influenced the later KRONOS and SCOPE systems. The first Cray-1 was delivered to the Los Alamos Lab with no operating system, or any other software. Los Alamos developed the application software for it, and the operating system. The main timesharing system for the Cray 1, the Cray Time Sharing System (CTSS), was then developed at the Livermore Labs as a direct descendant of the Livermore Time Sharing System (LTSS) for the CDC 6600 operating system from twenty years earlier. In developing supercomputers, rising software costs soon became dominant, as evidenced by the 1980s cost for software development at Cray growing to equal their cost for hardware. That trend was partly responsible for a move away from the in-house Cray Operating System to UNICOS system based on Unix. In 1985, the Cray-2 was the first system to ship with the UNICOS operating system. Around the same time, the EOS operating system was developed by ETA Systems for use in their ETA10 supercomputers. Written in Cybil, a Pascal-like language from Control Data Corporation, EOS highlighted the stability problems in developing stable operating systems for supercomputers and eventually a Unix-like system was offered on the same machine. The lessons learned from developing ETA system software included the high level of risk associated with developing a new supercomputer operating system, and the advantages of using Unix with its large extant base of system software libraries. By the middle 1990s, despite the extant investment in older operating systems, the trend was toward the use of Unix-based systems, which also facilitated the use of interactive graphical user interfaces (GUIs) for scientific computing across multiple platforms. The move toward a commodity OS had opponents, who cited the fast pace and focus of Linux development as a major obstacle against adoption. As one author wrote "Linux will likely catch up, but we have large-scale systems now". Nevertheless, that trend continued to gain momentum and by 2005, virtually all supercomputers used some Unix-like OS. These variants of Unix included IBM AIX, the open source Linux system, and other adaptations such as UNICOS from Cray. By the end of the 20th century, Linux was estimated to command the highest share of the supercomputing pie. == Modern approaches == The IBM Blue Gene supercomputer uses the CNK operating system on the compute nodes, but uses a modified Linux-based kernel called I/O Node Kernel (INK) on the I/O nodes. CNK is a lightweight kernel that runs on each node and supports a single application running for a single user on that node. For the sake of efficient operation, the design of CNK was kept simple and minimal, with physical memory being statically mapped and the CNK neither needing nor providing scheduling or context switching. CNK does not even implement file I/O on the compute node, but delegates that to dedicated I/O nodes. However, given that on the Blue Gene multiple compute nodes share a single I/O node, the I/O node operating system does require multi-tasking, hence the selection of the Linux-based operating system. While in traditional multi-user computer systems and early supercomputers, job scheduling was in effect a task scheduling problem for processing and peripheral resources, in a massively parallel system, the job management system needs to manage the allocation of both computational and communication resources. It is essential to tune task scheduling, and the operating system, in different configurations of a supercomputer. A typical parallel job scheduler has a master scheduler which instructs some number of slave schedulers to launch, monitor, and control parallel jobs, and periodically receives reports from them about the status of job progress. Some, but not all supercomputer schedulers attempt to maintain locality of job execution. The PBS Pro scheduler used on the Cray XT3 and Cray XT4 systems does not attempt to optimize locality on its three-dimensional torus interconnect, but simply uses the first available processor. On the other hand, IBM's scheduler on the Blue Gene supercomputers aims to exploit locality a

  Read more →
• 

  Actor-critic algorithm

  The actor-critic algorithm (AC) is a family of reinforcement learning (RL) algorithms that combine policy-based RL algorithms such as policy gradient methods, and value-based RL algorithms such as value iteration, Q-learning, SARSA, and TD learning. An AC algorithm consists of two main components: an "actor" that determines which actions to take according to a policy function, and a "critic" that evaluates those actions according to a value function. Some AC algorithms are on-policy, some are off-policy. Some apply to either continuous or discrete action spaces. Some work in both cases. == Overview == The actor-critic methods can be understood as an improvement over pure policy gradient methods like REINFORCE via introducing a baseline. === Actor === The actor uses a policy function π ( a | s ) {\displaystyle \pi (a|s)} , while the critic estimates either the value function V ( s ) {\displaystyle V(s)} , the action-value Q-function Q ( s , a ) , {\displaystyle Q(s,a),} the advantage function A ( s , a ) {\displaystyle A(s,a)} , or any combination thereof. The actor is a parameterized function π θ {\displaystyle \pi _{\theta }} , where θ {\displaystyle \theta } are the parameters of the actor. The actor takes as argument the state of the environment s {\displaystyle s} and produces a probability distribution π θ ( ⋅ | s ) {\displaystyle \pi _{\theta }(\cdot |s)} . If the action space is discrete, then ∑ a π θ ( a | s ) = 1 {\displaystyle \sum _{a}\pi _{\theta }(a|s)=1} . If the action space is continuous, then ∫ a π θ ( a | s ) d a = 1 {\displaystyle \int _{a}\pi _{\theta }(a|s)da=1} . The goal of policy optimization is to improve the actor. That is, to find some θ {\displaystyle \theta } that maximizes the expected episodic reward J ( θ ) {\displaystyle J(\theta )} : J ( θ ) = E π θ [ ∑ t = 0 T γ t r t ] {\displaystyle J(\theta )=\mathbb {E} _{\pi _{\theta }}\left[\sum _{t=0}^{T}\gamma ^{t}r_{t}\right]} where γ {\displaystyle \gamma } is the discount factor, r t {\displaystyle r_{t}} is the reward at step t {\displaystyle t} , and T {\displaystyle T} is the time-horizon (which can be infinite). The goal of policy gradient method is to optimize J ( θ ) {\displaystyle J(\theta )} by gradient ascent on the policy gradient ∇ J ( θ ) {\displaystyle \nabla J(\theta )} . As detailed on the policy gradient method page, there are many unbiased estimators of the policy gradient: ∇ θ J ( θ ) = E π θ [ ∑ 0 ≤ j ≤ T ∇ θ ln ⁡ π θ ( A j | S j ) ⋅ Ψ j | S 0 = s 0 ] {\displaystyle \nabla _{\theta }J(\theta )=\mathbb {E} _{\pi _{\theta }}\left[\sum _{0\leq j\leq T}\nabla _{\theta }\ln \pi _{\theta }(A_{j}|S_{j})\cdot \Psi _{j}{\Big |}S_{0}=s_{0}\right]} where Ψ j {\textstyle \Psi _{j}} is a linear sum of the following: ∑ 0 ≤ i ≤ T ( γ i R i ) {\textstyle \sum _{0\leq i\leq T}(\gamma ^{i}R_{i})} . γ j ∑ j ≤ i ≤ T ( γ i − j R i ) {\textstyle \gamma ^{j}\sum _{j\leq i\leq T}(\gamma ^{i-j}R_{i})} : the REINFORCE algorithm. γ j ∑ j ≤ i ≤ T ( γ i − j R i ) − b ( S j ) {\textstyle \gamma ^{j}\sum _{j\leq i\leq T}(\gamma ^{i-j}R_{i})-b(S_{j})} : the REINFORCE with baseline algorithm. Here b {\displaystyle b} is an arbitrary function. γ j ( R j + γ V π θ ( S j + 1 ) − V π θ ( S j ) ) {\textstyle \gamma ^{j}\left(R_{j}+\gamma V^{\pi _{\theta }}(S_{j+1})-V^{\pi _{\theta }}(S_{j})\right)} : TD(1) learning. γ j Q π θ ( S j , A j ) {\textstyle \gamma ^{j}Q^{\pi _{\theta }}(S_{j},A_{j})} . γ j A π θ ( S j , A j ) {\textstyle \gamma ^{j}A^{\pi _{\theta }}(S_{j},A_{j})} : Advantage Actor-Critic (A2C). γ j ( R j + γ R j + 1 + γ 2 V π θ ( S j + 2 ) − V π θ ( S j ) ) {\textstyle \gamma ^{j}\left(R_{j}+\gamma R_{j+1}+\gamma ^{2}V^{\pi _{\theta }}(S_{j+2})-V^{\pi _{\theta }}(S_{j})\right)} : TD(2) learning. γ j ( ∑ k = 0 n − 1 γ k R j + k + γ n V π θ ( S j + n ) − V π θ ( S j ) ) {\textstyle \gamma ^{j}\left(\sum _{k=0}^{n-1}\gamma ^{k}R_{j+k}+\gamma ^{n}V^{\pi _{\theta }}(S_{j+n})-V^{\pi _{\theta }}(S_{j})\right)} : TD(n) learning. γ j ∑ n = 1 ∞ λ n − 1 1 − λ ⋅ ( ∑ k = 0 n − 1 γ k R j + k + γ n V π θ ( S j + n ) − V π θ ( S j ) ) {\textstyle \gamma ^{j}\sum _{n=1}^{\infty }{\frac {\lambda ^{n-1}}{1-\lambda }}\cdot \left(\sum _{k=0}^{n-1}\gamma ^{k}R_{j+k}+\gamma ^{n}V^{\pi _{\theta }}(S_{j+n})-V^{\pi _{\theta }}(S_{j})\right)} : TD(λ) learning, also known as GAE (generalized advantage estimate). This is obtained by an exponentially decaying sum of the TD(n) learning terms. === Critic === In the unbiased estimators given above, certain functions such as V π θ , Q π θ , A π θ {\displaystyle V^{\pi _{\theta }},Q^{\pi _{\theta }},A^{\pi _{\theta }}} appear. These are approximated by the critic. Since these functions all depend on the actor, the critic must learn alongside the actor. The critic is learned by value-based RL algorithms. For example, if the critic is estimating the state-value function V π θ ( s ) {\displaystyle V^{\pi _{\theta }}(s)} , then it can be learned by any value function approximation method. Let the critic be a function approximator V ϕ ( s ) {\displaystyle V_{\phi }(s)} with parameters ϕ {\displaystyle \phi } . The simplest example is TD(1) learning, which trains the critic to minimize the TD(1) error: δ i = R i + γ V ϕ ( S i + 1 ) − V ϕ ( S i ) {\displaystyle \delta _{i}=R_{i}+\gamma V_{\phi }(S_{i+1})-V_{\phi }(S_{i})} The critic parameters are updated by gradient descent on the squared TD error: ϕ ← ϕ − α ∇ ϕ ( δ i ) 2 = ϕ + α δ i ∇ ϕ V ϕ ( S i ) {\displaystyle \phi \leftarrow \phi -\alpha \nabla _{\phi }(\delta _{i})^{2}=\phi +\alpha \delta _{i}\nabla _{\phi }V_{\phi }(S_{i})} where α {\displaystyle \alpha } is the learning rate. Note that the gradient is taken with respect to the ϕ {\displaystyle \phi } in V ϕ ( S i ) {\displaystyle V_{\phi }(S_{i})} only, since the ϕ {\displaystyle \phi } in γ V ϕ ( S i + 1 ) {\displaystyle \gamma V_{\phi }(S_{i+1})} constitutes a moving target, and the gradient is not taken with respect to that. This is a common source of error in implementations that use automatic differentiation, and requires "stopping the gradient" at that point. Similarly, if the critic is estimating the action-value function Q π θ {\displaystyle Q^{\pi _{\theta }}} , then it can be learned by Q-learning or SARSA. In SARSA, the critic maintains an estimate of the Q-function, parameterized by ϕ {\displaystyle \phi } , denoted as Q ϕ ( s , a ) {\displaystyle Q_{\phi }(s,a)} . The temporal difference error is then calculated as δ i = R i + γ Q θ ( S i + 1 , A i + 1 ) − Q θ ( S i , A i ) {\displaystyle \delta _{i}=R_{i}+\gamma Q_{\theta }(S_{i+1},A_{i+1})-Q_{\theta }(S_{i},A_{i})} . The critic is then updated by θ ← θ + α δ i ∇ θ Q θ ( S i , A i ) {\displaystyle \theta \leftarrow \theta +\alpha \delta _{i}\nabla _{\theta }Q_{\theta }(S_{i},A_{i})} The advantage critic can be trained by training both a Q-function Q ϕ ( s , a ) {\displaystyle Q_{\phi }(s,a)} and a state-value function V ϕ ( s ) {\displaystyle V_{\phi }(s)} , then let A ϕ ( s , a ) = Q ϕ ( s , a ) − V ϕ ( s ) {\displaystyle A_{\phi }(s,a)=Q_{\phi }(s,a)-V_{\phi }(s)} . Although, it is more common to train just a state-value function V ϕ ( s ) {\displaystyle V_{\phi }(s)} , then estimate the advantage by A ϕ ( S i , A i ) ≈ ∑ j ∈ 0 : n − 1 γ j R i + j + γ n V ϕ ( S i + n ) − V ϕ ( S i ) {\displaystyle A_{\phi }(S_{i},A_{i})\approx \sum _{j\in 0:n-1}\gamma ^{j}R_{i+j}+\gamma ^{n}V_{\phi }(S_{i+n})-V_{\phi }(S_{i})} Here, n {\displaystyle n} is a positive integer. The higher n {\displaystyle n} is, the more lower is the bias in the advantage estimation, but at the price of higher variance. The Generalized Advantage Estimation (GAE) introduces a hyperparameter λ {\displaystyle \lambda } that smoothly interpolates between Monte Carlo returns ( λ = 1 {\displaystyle \lambda =1} , high variance, no bias) and 1-step TD learning ( λ = 0 {\displaystyle \lambda =0} , low variance, high bias). This hyperparameter can be adjusted to pick the optimal bias-variance trade-off in advantage estimation. It uses an exponentially decaying average of n-step returns with λ {\displaystyle \lambda } being the decay strength. == Variants == Asynchronous Advantage Actor-Critic (A3C): Parallel and asynchronous version of A2C. Soft Actor-Critic (SAC): Incorporates entropy maximization for improved exploration. Deep Deterministic Policy Gradient (DDPG): Specialized for continuous action spaces.

  Read more →
• 

  HTTP cookie

  An HTTP cookie (also called web cookie, Internet cookie, browser cookie, or simply cookie) is a small block of data created by a web server while a user is browsing a website and placed on the user's computer or other device by the user's web browser. Cookies are placed on the device used to access a website, and more than one cookie may be placed on a user's device during a session. Cookies serve useful and sometimes essential functions on the web. They enable web servers to store stateful information (such as items added in the shopping cart in an online store) on the user's device or to track the user's browsing activity (including clicking particular buttons, logging in, or recording which pages were visited in the past). They can also be used to save information that the user previously entered into form fields, such as names, addresses, passwords, and payment card numbers for subsequent use. Authentication cookies are commonly used by web servers to authenticate that a user is logged in, and with which account they are logged in. Without the cookie, users would need to authenticate themselves by logging in on each page containing sensitive information that they wish to access. The security of an authentication cookie generally depends on the security of the issuing website and the user's web browser, and on whether the cookie data is encrypted. Security vulnerabilities may allow a cookie's data to be read by an attacker, used to gain access to user data, or used to gain access (with the user's credentials) to the website to which the cookie belongs (see cross-site scripting and cross-site request forgery for examples). Tracking cookies, and especially third-party tracking cookies, are commonly used as ways to compile long-term records of individuals' browsing histories — a potential privacy concern that prompted European and U.S. lawmakers to take action in 2011. European law requires that all websites targeting European Union member states gain "informed consent" from users before storing non-essential cookies on their device. == Background == === Origin of the name === The term cookie was coined by web-browser programmer Lou Montulli. It was derived from the term magic cookie, which is a packet of data a program receives and sends back unchanged, used by Unix programmers. === History === Magic cookies were already used in computing when computer programmer Lou Montulli had the idea of using them in web communications in June 1994. At the time, he was an employee of Netscape Communications, which was developing an e-commerce application for MCI. Vint Cerf and John Klensin represented MCI in technical discussions with Netscape Communications. MCI did not want its servers to have to retain partial transaction states, which led them to ask Netscape to find a way to store that state in each user's computer instead. Cookies provided a solution to the problem of reliably implementing a virtual shopping cart. Together with John Giannandrea, Montulli wrote the initial Netscape cookie specification the same year. Version 0.9beta of Mosaic Netscape, released on 13 October 1994, supported cookies. The first use of cookies (out of the labs) was checking whether visitors to the Netscape website had already visited the site. Montulli applied for a patent for the cookie technology in 1995, which was granted in 1998. Support for cookies was integrated with Internet Explorer in version 2, released in October 1995. The introduction of cookies was not widely known to the public at the time. In particular, cookies were accepted by default, and users were not notified of their presence. The public learned about cookies after the Financial Times published an article about them on 12 February 1996. In the same year, cookies received a lot of media attention, especially because of potential privacy implications. Cookies were discussed in two U.S. Federal Trade Commission hearings in 1996 and 1997. The development of the formal cookie specifications was already ongoing. In particular, the first discussions about a formal specification started in April 1995 on the www-talk mailing list. A special working group within the Internet Engineering Task Force (IETF) was formed. Two alternative proposals for introducing state in HTTP transactions had been proposed by Brian Behlendorf and David Kristol respectively. But the group, headed by Kristol himself and Lou Montulli, soon decided to use the Netscape specification as a starting point. In February 1996, the working group identified third-party cookies as a considerable privacy threat. The specification produced by the group was eventually published as RFC 2109 in February 1997. It specifies that third-party cookies were either not allowed at all, or at least not enabled by default. At this time, advertising companies were already using third-party cookies. The recommendation about third-party cookies of RFC 2109 was not followed by Netscape and Internet Explorer. RFC 2109 was superseded by RFC 2965 in October 2000. RFC 2965 added a Set-Cookie2 header field, which informally came to be called "RFC 2965-style cookies" as opposed to the original Set-Cookie header field which was called "Netscape-style cookies". Set-Cookie2 was seldom used, however, and was deprecated in RFC 6265 in April 2011 which was written as a definitive specification for cookies as used in the real world. No modern browser recognizes the Set-Cookie2 header field. == Terminology == === Session cookie === A session cookie (also known as an in-memory cookie, transient cookie or non-persistent cookie) exists only in temporary memory while the user navigates a website. Session cookies expire or are deleted when the user closes the web browser. Session cookies are identified by the browser by the absence of an expiration date assigned to them. === Persistent cookie === A persistent cookie expires at a specific date or after a specific length of time. For the persistent cookie's lifespan set by its creator, its information will be transmitted to the server every time the user visits the website that it belongs to, or every time the user views a resource belonging to that website from another website (such as an advertisement). For this reason, persistent cookies are sometimes referred to as tracking cookies because they can be used by advertisers to record information about a user's web browsing habits over an extended period of time. Persistent cookies are also used for reasons such as keeping users logged into their accounts on websites, to avoid re-entering login credentials at every visit. (See § Uses, below.) === Secure cookie === A secure cookie can only be transmitted over an encrypted connection (i.e. HTTPS). They cannot be transmitted over unencrypted connections (i.e. HTTP). This makes the cookie less likely to be exposed to cookie theft via eavesdropping. A cookie is made secure by adding the Secure flag to the cookie. === Http-only cookie === An http-only cookie cannot be accessed by client-side APIs, such as JavaScript. This restriction eliminates the threat of cookie theft via cross-site scripting (XSS). However, the cookie remains vulnerable to cross-site tracing (XST) and cross-site request forgery (CSRF) attacks. A cookie is given this characteristic by adding the HttpOnly flag to the cookie. === Same-site cookie === In 2016 Google Chrome version 51 introduced a new kind of cookie with attribute SameSite with possible values of Strict, Lax or None. With attribute SameSite=Strict, the browsers would only send cookies to a target domain that is the same as the origin domain. This would effectively mitigate cross-site request forgery (CSRF) attacks. With SameSite=Lax, browsers would send cookies with requests to a target domain even it is different from the origin domain, but only for safe requests such as GET (POST is unsafe) and not third-party cookies (inside iframe). Attribute SameSite=None would allow third-party (cross-site) cookies, however, most browsers require secure attribute on SameSite=None cookies. The Same-site cookie is incorporated into a new RFC draft for "Cookies: HTTP State Management Mechanism" to update RFC 6265 (if approved). Chrome, Firefox, and Edge started to support Same-site cookies. The key of rollout is the treatment of existing cookies without the SameSite attribute defined, Chrome has been treating those existing cookies as if SameSite=None, this would let all website/applications run as before. Google intended to change that default to SameSite=Lax in Chrome 80 planned to be released in February 2020, but due to potential for breakage of those applications/websites that rely on third-party/cross-site cookies and COVID-19 circumstances, Google postponed this change to Chrome 84. === Supercookie === A supercookie is a cookie with an origin of a top-level domain (such as .com) or a public suffix (such as .co.uk). Ordinary cookies, by contrast, have an origin of a specific domain name, such as ex

  Read more →
• 

  Acousto-electronics

  Acousto-electronics (also spelled 'Acoustoelectronics') is a branch of physics, acoustics and electronics that studies interactions of ultrasonic and hypersonic waves in solids with electrons and with electro-magnetic fields. Typical phenomena studied in acousto-electronics are acousto-electric effect and also amplification of acoustic waves by flows of electrons in piezoelectric semiconductors, when the drift velocity of the electrons exceeds the velocity of sound. The term 'acousto-electronics' is often understood in a wider sense to include numerous practical applications of the interactions of electro-magnetic fields with acoustic waves in solids. In particular, these are signal processing devices using surface acoustic waves (SAW), different sensors of temperature, pressure, humidity, acceleration, etc.

  Read more →
• 

  CSS box model

  In web development, the CSS box model refers to how HTML elements are modeled in browser engines and how the dimensions of those HTML elements are derived from CSS properties. It is a fundamental concept for the composition of HTML webpages. The guidelines of the box model are described by web standards World Wide Web Consortium (W3C) specifically the CSS Working Group. For much of the late-1990s and early 2000s there had been non-standard compliant implementations of the box model in mainstream browsers. With the advent of CSS2 in 1998, which introduced the box-sizing property, the problem had mostly been resolved. == Specifics == The Cascading Style Sheets (CSS) specification describes how elements of web pages are displayed by graphical browsers. Section 4 of the CSS1 specification defines a "formatting model" that gives block-level elements—such as p and blockquote—a width and height, and three levels of boxes surrounding it: padding, borders, and margins. While the specification never uses the term "box model" explicitly, the term has become widely used by web developers and web browser vendors. All HTML elements can be considered "boxes", this includes div tag, p tag, or a tag. Each of those boxes has five modifiable dimensions: the height and width describe dimensions of the actual content of the box (text, images, ...) the padding describes the space between this content and the border of the box the border is any kind of line (solid, dotted, dashed...) surrounding the box, if present the margin is the space around the border According to the CSS1 specification, released by W3C in 1996 and revised in 1999, when a width or height is explicitly specified for any block-level element, it should determine only the width or height of the visible element, with the padding, borders, and margins applied afterward. Before CSS3, this box model was known as W3C box model, in CSS3, it is known as the content-box. The total width of a box is therefore margin-left + border-left + padding-left + width + padding-right + border-right + margin-right. Similarly, the total height of a box equals margin-top + border-top + padding-top + height + padding-bottom + border-bottom + margin-bottom. For example, the following CSS code would specify the box dimensions of each block belonging to 'my-class'. Moreover, each such box will have total height 140px and width 240px. CSS3 introduced the Internet Explorer box model to the standard, known referred to as border-box. == History == Before HTML 4 and CSS, very few HTML elements supported both border and padding, so the definition of the width and height of an element was not very contentious. However, it varied depending on the element. The HTML width attribute of a table defined the width of the table including its border. On the other hand, the HTML width attribute of an image defined the width of the image itself (inside any border). The only element to support padding in those early days was the table cell. Width for the cell was defined as "the suggested width for a cell content in pixels excluding the cell padding." In 1996, CSS introduced margin, border and padding for many more elements. It adopted a definition width in relation to content, border, margin and padding similar to that for a table cell. This has since become known as the W3C box model. At the time, very few browser vendors implemented the W3C box model to the letter. The two major browsers at the time, Netscape 4.0 and Internet Explorer 4.0 both defined width and height as the distance from border to border. This has been referred to as the traditional or the Internet Explorer box model. Internet Explorer in "quirks mode" includes the content, padding and borders within a specified width or height; this results in a narrower or shorter rendering of a box than would result following the standard behavior. The Internet Explorer box model behavior was often considered a bug, because of the way in which earlier versions of Internet Explorer handle the box model or sizing of elements in a web page, which differs from the standard way recommended by the W3C for the Cascading Style Sheets language. As of Internet Explorer 6, the browser supports an alternative rendering mode (called the "standards-compliant mode") which solves this discrepancy. However, for backward compatibility reasons, all versions still behave in the usual, non-standard way by default (see quirks mode). Internet Explorer for Mac is not affected by this non-standard behavior. === Workarounds === Internet Explorer versions 6 and onward are not affected by the bug if the page contains certain HTML document type declarations. These versions maintain the buggy behavior when in quirks mode for reasons of backward compatibility. For example, quirks mode is triggered: When the document type declaration is absent or incomplete; When an HTML 3 or earlier document is encountered; When an HTML 4.0 Transitional or Frameset document type declaration is used and a system identifier (URI) is not present; When an SGML comment or other unrecognized content appears before the document type declaration Internet Explorer 6 also uses quirks mode if there is an XML declaration prior to the document type declaration. Various workarounds have been devised to force Internet Explorer versions 5 and earlier to display Web pages using the W3C box model. These workarounds generally exploit unrelated bugs in Internet Explorer's CSS selector processing in order to hide certain rules from the browser. The best known of these workarounds is the "box model hack" developed by Tantek Çelik, a former Microsoft employee who developed this idea while working on Internet Explorer for the Macintosh. It involves specifying a width declaration for Internet Explorer for Windows, and then overriding it with another width declaration for CSS-compliant browsers. This second declaration is hidden from Internet Explorer for Windows by exploiting other bugs in the way that it parses CSS rules. The implementation of these CSS “hacks” has been further complicated by the public release of Internet Explorer 7, which has had some issues fixed, but not others, causing undesired results in pages using these hacks. Box model hacks have proven unreliable because they rely on bugs in browsers' CSS support that may be fixed in later versions. For this reason, some Web developers have instead recommended either avoiding specifying both width and padding for the same element or using conditional comment and/or CSS filters to work around the box model bug in older versions of Internet Explorer. == Support for Internet Explorer's box model == Web designer Doug Bowman has said that the original Internet Explorer box model represents a better, more logical approach. Peter-Paul Koch gives the example of a physical box, whose dimensions always refer to the box itself, including potential padding, but never its content. He says that this box model is more useful for graphic designers, who create designs based on the visible width of boxes rather than the width of their content. Bernie Zimmermann says that the Internet Explorer box model is closer to the definition of cell dimensions and padding used in the HTML table model. The W3C has included a "box-sizing" property in CSS3. When box-sizing: border-box; is specified for an element, any padding or border of the element is drawn inside the specified width and height, "as commonly implemented by legacy HTML user agents". Internet Explorer 8, WebKit browsers such as Apple Safari 5.1+ and Google Chrome, Gecko-based browsers such as Mozilla Firefox 29.0 and later, Opera 7.0 and later, and Konqueror 3.3.2 and later support the CSS3 box-sizing property. Gecko browsers previous than 29.0 support the same functionality using the browser-specific -moz-box-sizing property. border-box is the default box model used in Bootstrap framework.

  Read more →
• 

  Altibase

  Altibase is a hybrid database, relational database management system manufactured by the Altibase Corporation. The software's hybrid architecture allows it to access both memory-resident and disk-resident tables using single interface. It supports both synchronous and asynchronous replication and offers real-time ACID compliance. Support is also offered for a variety of SQL standards and programming languages. Other important capabilities include data import and export, data encryption for security, multiple data access command sets, materialized view and temporary tables, and others. == History == From 1991 through 1997 the Mr. RT project was an in-memory database research project, conducted by the Electronics and Telecommunications Research Institute a government-funded research organization in South Korea. Altibase was incorporated in 1999. Altibase acquired an in-memory database engine from the Electronics and Telecommunications Research Institute in February 2000, and commercialized the database in October of the same year. In 2001, Altibase changed the name of the in-memory database product from "Spiner" to "Altibase" in 2001. In 2004, Altibase integrated the in-memory database with a disk-resident database to create a hybrid DBMS, released version 4.0 and renamed it as ALTIBASE HDB. Altibase released version 5.5.1 and 6.1.1 in 2012, version 6.3.1 in November 2013, and 6.5.1 in May 2015. Altibase claims that this is the world's first hybrid DBMS. Altibase released its open source edition version 7.1, however, closed the source in 2023. In August 2023, Altibase released its cloud-optimized version 7.3. === Awards === In 2006, Received the Presidential Award at the Korea Software Awards In 2007, Selected as World-Class Product by the Ministry of Commerce, Industry and Energy In 2009, Awarded the Outstanding Product Award in China's Telecommunications Industry In 2009, Received Outstanding Product Award at the China Billing China 2009 Telecommunication Industry Awards In 2010, Commendation from the Minister of Knowledge Economy for Technological Practicalization In 2011, Received the Grand Prize at the 10th Software Enterprise Competitiveness Award In 2011, Selected as Top 10 Emerging Technologies and received Special Award at the Korea Technology Grand Prize In 2012, Awarded for Contributions to Military Manpower Administration In 2014~2016, Included in Gartner Magic Quadrant for Operational DBMS In 2015, Selected as Outstanding BSS by China Fujian Mobile. In 2023, Awarded as the Excellent Research and Development Institution by the Korean Ministry Science and ICT In 2023, Won the Global Premium Commercial Software Presidential Award at the 9th Global Commercial Software Grand Exhibition in Korea === Release === The first version, called Spiner, was released in 2000 for commercial use. It took half of the in-memory DBMS market share in South Korea. In 2002 the second version was released renamed to Altibase v2.0. By 2003, Altibase v3.0 was released and it entered the Chinese market. Released version 4.0 with hybrid architecture, combining RAM and disk databases, was released in 2004. In 2005 Altibase began working with Chinese telecommunications providers for billing systems, and some financial companies in Taiwan, China, for home trading systems. The software was certified by the Telecommunications Technology Association. The Ministry of Government Administration and Home Affairs gave it an award in 2006. Offices in China and United States opened in 2009. In 2011, version 5.5.1 was renamed it to HDB (for "hybrid database"). The Altibase Data Stream product for complex event processing was renamed DSM. The product received a Korean technology award. Altibase introduced certification services. In 2012, HDB Zeta and Extreme were announced, and DSM renamed to CEP. In 2013, yet another variant called XDB was announced, and the company received ISO/IEC 20000 certification. In 2018, Altibase went open source. Altibase went open source in February, 2018. Altibase Corp has made the decision to discontinue the Altibase 7.1 open source edition, effective March 17, 2023. As a result, the open-source edition of Altibase 7.1 will no longer be available for download or use. Altibase released version 7.3 in September, 2023, its notable feature is the world’s first hybrid partition, allowing data to be stored in both memory and on disk at the partition level. Version 7.3 also added parallel processing capabilities for high-speed performance in both partitioned and non-partitioned scenarios. Improving potential bottlenecks associated with Commit and logging that impact transaction performance, version 7.3 has achieved an approximately 490% enhancement in performance compared to previous versions. === Release history === == Clients == According to marketing research, Altibase have over 700 customers and more than 8,000 of installations and deployments, including 22 Fortune Global 500 Companies. Altibase's clients in the telecommunications, financial services, manufacturing, and utilities sectors include Bloomberg, AT&T, LG, Intel, LGU+, ETRADE, HP, UAT Inc., POSCO, SK Telecom, KT Corporation, Samsung Electronics, Shinhan Bank, Woori Bank, Canon(Toshiba), Hanhwa, The South Korean Ministry of Defense, G-Market, CJ, and Chung-Ang University. === Global clients === Japan FX Prime, a foreign exchange services company Retela Crea Securities United States AT&T Implemented Altibase for its PS-LTE Safety network, where the Presence service plays a vital role. This service handles the reception and storage of user information, conducting real-time checks for online presence and location as needed. Canada Telus One of the major telecommunication companies. Utilizes Altibase for its operations involving real-time user management, processing high volumes of dedicated terminal data, and managing real-time location information (GIS) for terminals. Altibase contributes to the company's in-house solution for maintaining uninterrupted services during national disasters or similar situations, ensuring efficiency and reliability. China China Mobile, China Unicom, China Telecom The three major telecommunications companies. Utilize ALTIBASE HDB in 29 of 31 Chinese provinces. Turkish Ziraat Bank, Halk Bank, Deniz Bank, Garanti BBVA, TEB, Oyak Bank, QNB, Burgan Bank, and others. In 2018, Altibase entered the market through a partnership with ATP-Tradesoft, a subsidiary of Ata Holdings. Collaborating with ATP-Tradesoft. Altibase integrated into the Online Trading System XFront. This integration was well-received by major financial institutions and securities firms in Turkey. Altibase is currently implemented in the XFront Online Trading System, used by 13 significant financial institutions and banks in the Turkey. Thailand Bualuang Securities Altibase has been supplied its DBMS to support the construction of the online stock trading platform. Mongolia MobiCom The Mongolian telecommunication giant, has adopted Altibase’s 7.0 version for its mobile platform for storing the infrequently used data. Azerbaijan M1 highway Altibase has been supplied as the Database Management System (DBMS) for the electronic toll collection system. One of the most crucial transportation networks in the country. India State-owned Karur Vysya Bank In 2013, Altibase provided its hybrid database solution and was deployed for the online banking system === Industries === Telecommunications LGU+ SK Telecom KT Corporation AT&T Telus Financial services Shinhan Bank Woori Bank KakaoPay Securities Implemented Altibase in its stock trading system Leveraging Altibase's replication feature, along with offline replication through shared disk and adapter functionality, the system ensures a high level of availability and consistency, with a reliability rate of 99.999% even in the event of system failures. COREDAX Cryptocurrency market Altibase has entered into a strategic partnership by signing a database management system (DBMS) supply contract with the cryptocurrency exchange Bloomberg ETRADE Manufacturing Samsung Electronics LG POSCO Hanhwa Canon(Toshiba) Intel HP Utilities South Korean Ministry of Defense G-Market CJ UAT Inc. Chung-Ang University == Features == Altibase is a so-called "hybrid DBMS", meaning that it simultaneously supports access to both memory-resident and disk-resident tables via a single interface. It is compatible with Solaris, HP-UX, AIX, Linux, and Windows. It supports the complete SQL standard, features Multiversion concurrency control (MVCC), implements Fuzzy and Ping-Pong Checkpointing for periodically backing up memory-resident data, and ships with Replication and Database Link functionality. High performance, large -capacity service Fast real-time data processing and large amounts of data stable Provide parallel processing architecture for large data management Developed and provided Hybrid Partitioned Table function for efficiency according to data personality High stability

  Read more →
• 

  Mean opinion score

  Mean opinion score (MOS) is a measure used in the domain of Quality of Experience and telecommunications engineering, representing overall quality of a stimulus or system. It is the arithmetic mean over all individual "values on a predefined scale that a subject assigns to his opinion of the performance of a system quality". Such ratings are usually gathered in a subjective quality evaluation test, but they can also be algorithmically estimated. MOS is a commonly used measure for video, audio, and audiovisual quality evaluation, but not restricted to those modalities. ITU-T has defined several ways of referring to a MOS in Recommendation ITU-T P.800.1, depending on whether the score was obtained from audiovisual, conversational, listening, talking, or video quality tests. == Rating scales and mathematical definition == The MOS is expressed as a single rational number, typically in the range 1–5, where 1 is lowest perceived quality, and 5 is the highest perceived quality. Other MOS ranges are also possible, depending on the rating scale that has been used in the underlying test. The Absolute Category Rating scale is very commonly used, which maps ratings between Bad and Excellent to numbers between 1 and 5, as seen in below table. Other standardized quality rating scales exist in ITU-T Recommendations (such as ITU-T P.800 or ITU-T P.910). For example, one could use a continuous scale ranging between 1–100. Which scale is used depends on the purpose of the test. In certain contexts there are no statistically significant differences between ratings for the same stimuli when they are obtained using different scales. The MOS is calculated as the arithmetic mean over single ratings performed by human subjects for a given stimulus in a subjective quality evaluation test. Thus: M O S = ∑ n = 1 N R n N {\displaystyle MOS={\frac {\sum _{n=1}^{N}{R_{n}}}{N}}} Where R {\displaystyle R} are the individual ratings for a given stimulus by N {\displaystyle N} subjects. == Properties of the MOS == The MOS is subject to certain mathematical properties and biases. In general, there is an ongoing debate on the usefulness of the MOS to quantify Quality of Experience in a single scalar value. When the MOS is acquired using a categorical rating scales, it is based on – similar to Likert scales – an ordinal scale. In this case, the ranking of the scale items is known, but their interval is not. Therefore, it is mathematically incorrect to calculate a mean over individual ratings in order to obtain the central tendency; the median should be used instead. However, in practice and in the definition of MOS, it is considered acceptable to calculate the arithmetic mean. It has been shown that for categorical rating scales (such as ACR), the individual items are not perceived equidistant by subjects. For example, there may be a larger "gap" between Good and Fair than there is between Good and Excellent. The perceived distance may also depend on the language into which the scale is translated. However, there exist studies that could not prove a significant impact of scale translation on the obtained results. Several other biases are present in the way MOS ratings are typically acquired. In addition to the above-mentioned issues with scales that are perceived non-linearly, there is a so-called "range-equalization bias": subjects, over the course of a subjective experiment, tend to give scores that span the entire rating scale. This makes it impossible to compare two different subjective tests if the range of presented quality differs. In other words, the MOS is never an absolute measure of quality, but only relative to the test in which it has been acquired. For the above reasons – and due to several other contextual factors influencing the perceived quality in a subjective test – a MOS value should only be reported if the context in which the values have been collected in is known and reported as well. MOS values gathered from different contexts and test designs therefore should not be directly compared. Recommendation ITU-T P.800.2 prescribes how MOS values should be reported. Specifically, P.800.2 says:it is not meaningful to directly compare MOS values produced from separate experiments, unless those experiments were explicitly designed to be compared, and even then the data should be statistically analysed to ensure that such a comparison is valid. == MOS for speech and audio quality estimation == MOS historically originates from subjective measurements where listeners would sit in a "quiet room" and score a telephone call quality as they perceived it. This kind of test methodology had been in use in the telephony industry for decades and was standardized in Recommendation ITU-T P.800. It specifies that "the talker should be seated in a quiet room with volume between 30 and 120 m³ and a reverberation time less than 500 ms (preferably in the range 200–300 ms). The room noise level must be below 30 dBA with no dominant peaks in the spectrum." Requirements for other modalities were similarly specified in later ITU-T Recommendations. == MOS estimation using quality models == Obtaining MOS ratings may be time-consuming and expensive as it requires the recruitment of human assessors. For various use cases such as codec development or service quality monitoring purposes – where quality should be estimated repeatedly and automatically – MOS scores can also be predicted by objective quality models, which typically have been developed and trained using human MOS ratings. A question that arises from using such models is whether the MOS differences produced are noticeable to the users. For example, when rating images on a five point MOS scale, an image with a MOS equal to 5 is expected to be noticeably better in quality than one with a MOS equal to 1. Contrary to that, it is not evident whether an image with a MOS equal to 3.8 is noticeably better in quality than one with a MOS equal to 3.6. Research conducted on determining the smallest MOS difference that is perceptible to users for digital photographs showed that a MOS difference of approximately 0.46 is required in order for 75% of the users to be able to detect the higher quality image. Nevertheless, image quality expectation, and hence MOS, changes over time with the change of user expectations. As a result, minimum noticeable MOS differences determined using analytical methods such as in may change over time.

  Read more →
• 

  Mini-STX

  Mini-STX (mSTX, Mini Socket Technology EXtended, originally "Intel 5x5") is a computer motherboard form factor that was released by Intel in 2015 (as "Intel 5x5"). These motherboards measure 147mm by 140mm (5.8" x 5.5"), making them larger than "4x4" NUC (102x102mm / 4.01" x 4.01" inches) and Nano-ITX (120x120mm / 4.7" x 4.7") boards, but notably smaller than the more common Mini-ITX (170x170mm / 6.7" x 6.7") boards. Unlike these standards, which use a square shape, the Mini-STX form factor is 7mm longer from front-to-rear, making it slightly rectangular. == Mini-STX design elements == The Mini-STX design suggests (but does not require) support for: Socketed processors (e.g. LGA or PGA CPUs) Onboard power regulation circuitry, enabling direct DC power input IO ports embedded on the front and rear of the motherboard (akin to NUC, but unlike typical motherboards which often use headers instead to connect built-in ports on enclosures) == Adoption by manufacturers == This motherboard form factor is still not in particularly common use with consumer-PC manufacturers, although there are a few offerings: ASRock offers both DeskMini kits (that use mini-STX boards) and standalone motherboards, Asus offer VivoMini kits (that use mini-STX boards) and standalone motherboards, Gigabyte offers a few motherboards, and industrial PC suppliers (e.g. Kontron, Iesy, ASRock Industrial) also provide some options for mini-STX equipment. == Derivatives == ASRock developed a derivative of mini-STX, dubbed micro-STX, for their 'DeskMini GTX/RX' small form-factor PCs and industrial motherboards. Micro-STX adds an MXM slot which allows the use of special PCI Express expansion cards, including graphics or machine learning accelerators, but increases the width of the board to be extended two inches, resulting in measurements of 147 x 188 mm (5.8" x 7.4")

  Read more →
• 

  Packingham v. North Carolina

  Packingham v. North Carolina, 582 U.S. 98 (2017), is a case in which the Supreme Court of the United States held that a North Carolina statute that prohibited registered sex offenders from using social media websites was unconstitutional because it violated the First Amendment to the U.S. Constitution, which protects freedom of speech. In 2010, Lester Gerard Packingham, a registered sex offender, posted on Facebook under a pseudonym to comment favorably on a recent traffic court experience. Police then identified Packingham and charged him with violating North Carolina's law. Packingham moved to dismiss the charges, arguing that the state's law violated the First Amendment. The trial court dismissed this motion and ultimately convicted Packingham. A state appellate court initially reversed the trial court, holding that the law did violate the First Amendment, but the North Carolina Supreme Court, the state's highest court, disagreed and reinstated the conviction. In June 2017, the U.S. Supreme Court unanimously reversed the North Carolina Supreme Court's judgment. In the majority opinion authored by Justice Anthony Kennedy, the Court held that social media—defined broadly to include Facebook, Amazon.com, The Washington Post, and WebMD, among many others—is a "protected space" under the First Amendment for lawful speech. The Court offered that North Carolina could protect children through less restrictive means, such as prohibiting "conduct that often presages a sexual crime, like contacting a minor or using a website to gather information about a minor". == Background == === North Carolina statute === In 2008, the state of North Carolina passed a law that made it a felony for a registered sex offender "to access a commercial social networking Web site where the sex offender knows that the site permits minor children to become members or to create or maintain personal Web pages". The law defined a "commercial social networking Web site" using four criteria. Specifically, the website must: be "operated by a person who derives revenue from membership fees, advertising, or other sources related to the operation of the Web site". facilitate "the social introduction between two or more persons for the purposes of friendship, meeting other persons, or information exchanges". allow "users to create Web pages or personal profiles that contain information such as the name or nickname of the user, photographs placed on the personal Web page by the user, other personal information about the user, and links to other personal Web pages on the commercial social networking Web site of friends or associates of the user that may be accessed by other users or visitors to the Web site". provide "users or visitors... mechanisms to communicate with other users, such as a message board, chat room, electronic mail, or instant messenger". The law exempted websites that "Provid[e] only one of the following discrete services: photo-sharing, electronic mail, instant messenger, or chat room or message board platform", as well as websites that have as their primary purpose "the facilitation of commercial transactions involving goods or services between [their] members or visitors". === Facts of the case === In 2002, Lester Gerard Packingham was convicted of taking "indecent liberties with a child", a felony that required him to register as a sex offender. A North Carolina court sentenced him to 10–12 months in prison with 24 months of supervised release. He was given no other special instructions on his behavior outside of prison other than to "remain away from" the minor. In 2010, after a state court dismissed a traffic ticket against Packingham, he submitted a post on Facebook under the name "J. R. Gerrard", stating: "Man God is Good! How about I got so much favor they dismissed the ticket before court even started? No fine, no court cost, no nothing spent. . . . . .Praise be to GOD, WOW! Thanks JESUS!" The Durham Police Department identified Packingham as the author of the post after cross-checking the time of the post with recently dismissed traffic tickets, and a grand jury indicted him for violating the North Carolina statute. === Lower court proceedings === Initially, Packingham moved to dismiss his indictment, arguing that it violated the First Amendment. A North Carolina Superior Court judge denied this motion, and he was convicted of violating the North Carolina social media law. Packingham appealed his conviction to the North Carolina Court of Appeals, which reversed the trial court's decision in 2013. Applying intermediate scrutiny, the court of appeals determined that North Carolina's law violated the First Amendment because it was too broad, applying to all registered sex offenders regardless of whether the offender had committed a crime involving a minor or whether the offender was a continuing threat to minors. The appeals court also stated that the law had been defined broadly enough to prohibit a registered sex offender from conducting a wide array of Internet activity, such as "conducting a 'Google' search, purchasing items on Amazon.com, or accessing a plethora of Web sites unrelated to online communication with minors". In 2015, the North Carolina Supreme Court, the state's highest court, reversed the court of appeals, holding that the law was "constitutional in all respects". The North Carolina Supreme Court found that the statute was a "limitation on conduct" and did not impede any free speech. The state had a vested interest in “forestalling the illicit lurking and contact of minors” by registered sex offenders and potential future victims, and upheld Packingham's conviction. == Supreme Court ruling == Packingham filed a petition for a writ of certiorari with the Supreme Court of the United States. The federal government also filed a brief recommending that the Supreme Court grant certiorari, arguing that the North Carolina Supreme Court incorrectly decided the case in favor of the state. The U.S. Supreme Court granted certiorari in October 2016. Amicus briefs in support of Packingham were filed by the libertarian Cato Institute and the American Civil Liberties Union. The North Carolina Supreme Court filed a brief supporting its prior decision, urging the importance of protecting minors from being stalked online. === Oral argument === The oral argument took place in February 2017. Packingham’s lawyer, David T. Goldberg, argued that the law banned “vast swaths of First Amendment activity”, went too far in restricting which Internet sites could be accessed, and forbade use of the Internet in general. The law targeted speech on some of the platforms that Americans use most often, Goldberg noted, and that under the law Packingham could not even use Twitter to read the myriad messages discussing his own case. He further noted that the law imposes punishment without regard to whether the offender actually did anything wrong. North Carolina’s senior deputy Attorney General, Robert C. Montgomery, argued for the state, and claimed that communication through social media sites is a “crucial channel”. Justice Sonia Sotomayor asked Montgomery to provide evidence as to the claim that by giving Packingham Internet privileges, he would commit another crime. Justice Stephen Breyer added that “It seems to be well-settled law that the state can’t (bar usage) unless there is a 'clear and present danger'." === Opinion of the Court === In June 2017 the Supreme Court delivered a judgment in favor of Packingham, unanimously voting to reverse the state court's ruling. Justice Anthony Kennedy authored the decision, joined by Justice Ginsburg, Justice Breyer, Justice Sotomayor, and Justice Kagan. Kennedy explained the decision: "A fundamental principle of the First Amendment is that all persons have access to places where they can speak and listen, and then, after reflection, speak and listen once more." He continued that "By prohibiting sex offenders from using those websites, North Carolina with one broad stroke bars access to what for many are the principal sources for knowing current events, checking ads for employment, speaking and listening in the modern public square, and otherwise exploring the vast realms of human thought and knowledge." Citing Ashcroft v. Free Speech Coalition as a precedent, Kennedy also wrote: "It is well established that, as a general rule, the Government 'may not suppress lawful speech as the means to suppress unlawful speech'." === Concurring opinion === Justice Samuel Alito wrote an opinion concurring in the judgment, joined by John Roberts and Clarence Thomas. While Alito agreed that the state statute at issue violated the First Amendment, he noted that there are reasonable scenarios for which legal bans for sex offenders can be placed, such as for sites targeted at teenagers. Justice Gorsuch took no part in the decision of the case. == Impact == Packingham v. North Carolina was one of the first U.S. Supreme Court cases to ana

  Read more →
• 

  Neural style transfer

  Neural style transfer (NST) software algorithms are able to manipulate digital images, or videos, in order to adopt the appearance or visual style of another image. NST algorithms are characterized by their use of deep neural networks for the sake of image transformation. Common uses for NST are the creation of artificial artwork from photographs, for example by transferring the appearance of famous paintings to user-supplied photographs. Several notable mobile apps use NST techniques for this purpose, including DeepArt and Prisma. This method has been used by artists and designers around the globe to develop new artwork based on existent style(s). == History == NST is an example of image stylization, a problem studied for over two decades within the field of non-photorealistic rendering. The first two example-based style transfer algorithms were image analogies and image quilting. Both of these methods were based on patch-based texture synthesis algorithms. Given a training pair of images–a photo and an artwork depicting that photo–a transformation could be learned and then applied to create new artwork from a new photo, by analogy. If no training photo was available, it would need to be produced by processing the input artwork; image quilting did not require this processing step, though it was demonstrated on only one style. NST was first published in the paper "A Neural Algorithm of Artistic Style" by Leon Gatys et al., originally released to ArXiv 2015, and subsequently accepted by the peer-reviewed CVPR conference in 2016. The original paper used a VGG-19 architecture that has been pre-trained to perform object recognition using the ImageNet dataset. In 2017, Google AI introduced a method that allows a single deep convolutional style transfer network to learn multiple styles at the same time. This algorithm permits style interpolation in real-time, even when done on video media. == Mathematics == This section closely follows the original paper. === Overview === The idea of Neural Style Transfer (NST) is to take two images—a content image p → {\displaystyle {\vec {p}}} and a style image a → {\displaystyle {\vec {a}}} —and generate a third image x → {\displaystyle {\vec {x}}} that minimizes a weighted combination of two loss functions: a content loss L content ( p → , x → ) {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}})} and a style loss L style ( a → , x → ) {\displaystyle {\mathcal {L}}_{\text{style }}({\vec {a}},{\vec {x}})} . The total loss is a linear sum of the two: L NST ( p → , a → , x → ) = α L content ( p → , x → ) + β L style ( a → , x → ) {\displaystyle {\mathcal {L}}_{\text{NST}}({\vec {p}},{\vec {a}},{\vec {x}})=\alpha {\mathcal {L}}_{\text{content}}({\vec {p}},{\vec {x}})+\beta {\mathcal {L}}_{\text{style}}({\vec {a}},{\vec {x}})} By jointly minimizing the content and style losses, NST generates an image that blends the content of the content image with the style of the style image. Both the content loss and the style loss measures the similarity of two images. The content similarity is the weighted sum of squared-differences between the neural activations of a single convolutional neural network (CNN) on two images. The style similarity is the weighted sum of Gram matrices within each layer (see below for details). The original paper used a VGG-19 CNN, but the method works for any CNN. === Symbols === Let x → {\textstyle {\vec {x}}} be an image input to a CNN. Let F l ∈ R N l × M l {\textstyle F^{l}\in \mathbb {R} ^{N_{l}\times M_{l}}} be the matrix of filter responses in layer l {\textstyle l} to the image x → {\textstyle {\vec {x}}} , where: N l {\textstyle N_{l}} is the number of filters in layer l {\textstyle l} ; M l {\textstyle M_{l}} is the height times the width (i.e. number of pixels) of each filter in layer l {\textstyle l} ; F i j l ( x → ) {\textstyle F_{ij}^{l}({\vec {x}})} is the activation of the i th {\textstyle i^{\text{th}}} filter at position j {\textstyle j} in layer l {\textstyle l} . A given input image x → {\textstyle {\vec {x}}} is encoded in each layer of the CNN by the filter responses to that image, with higher layers encoding more global features, but losing details on local features. === Content loss === Let p → {\textstyle {\vec {p}}} be an original image. Let x → {\textstyle {\vec {x}}} be an image that is generated to match the content of p → {\textstyle {\vec {p}}} . Let P l {\textstyle P^{l}} be the matrix of filter responses in layer l {\textstyle l} to the image p → {\textstyle {\vec {p}}} . The content loss is defined as the squared-error loss between the feature representations of the generated image and the content image at a chosen layer l {\displaystyle l} of a CNN: L content ( p → , x → , l ) = 1 2 ∑ i , j ( A i j l ( x → ) − A i j l ( p → ) ) 2 {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}},l)={\frac {1}{2}}\sum _{i,j}\left(A_{ij}^{l}({\vec {x}})-A_{ij}^{l}({\vec {p}})\right)^{2}} where A i j l ( x → ) {\displaystyle A_{ij}^{l}({\vec {x}})} and A i j l ( p → ) {\displaystyle A_{ij}^{l}({\vec {p}})} are the activations of the i th {\displaystyle i^{\text{th}}} filter at position j {\displaystyle j} in layer l {\displaystyle l} for the generated and content images, respectively. Minimizing this loss encourages the generated image to have similar content to the content image, as captured by the feature activations in the chosen layer. The total content loss is a linear sum of the content losses of each layer: L content ( p → , x → ) = ∑ l v l L content ( p → , x → , l ) {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}})=\sum _{l}v_{l}{\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}},l)} , where the v l {\displaystyle v_{l}} are positive real numbers chosen as hyperparameters. === Style loss === The style loss is based on the Gram matrices of the generated and style images, which capture the correlations between different filter responses at different layers of the CNN: L style ( a → , x → ) = ∑ l = 0 L w l E l , {\displaystyle {\mathcal {L}}_{\text{style }}({\vec {a}},{\vec {x}})=\sum _{l=0}^{L}w_{l}E_{l},} where E l = 1 4 N l 2 M l 2 ∑ i , j ( G i j l ( x → ) − G i j l ( a → ) ) 2 . {\displaystyle E_{l}={\frac {1}{4N_{l}^{2}M_{l}^{2}}}\sum _{i,j}\left(G_{ij}^{l}({\vec {x}})-G_{ij}^{l}({\vec {a}})\right)^{2}.} Here, G i j l ( x → ) {\displaystyle G_{ij}^{l}({\vec {x}})} and G i j l ( a → ) {\displaystyle G_{ij}^{l}({\vec {a}})} are the entries of the Gram matrices for the generated and style images at layer l {\displaystyle l} . Explicitly, G i j l ( x → ) = ∑ k F i k l ( x → ) F j k l ( x → ) {\displaystyle G_{ij}^{l}({\vec {x}})=\sum _{k}F_{ik}^{l}({\vec {x}})F_{jk}^{l}({\vec {x}})} Minimizing this loss encourages the generated image to have similar style characteristics to the style image, as captured by the correlations between feature responses in each layer. The idea is that activation pattern correlations between filters in a single layer captures the "style" on the order of the receptive fields at that layer. Similarly to the previous case, the w l {\displaystyle w_{l}} are positive real numbers chosen as hyperparameters. === Hyperparameters === In the original paper, they used a particular choice of hyperparameters. The style loss is computed by w l = 0.2 {\displaystyle w_{l}=0.2} for the outputs of layers conv1_1, conv2_1, conv3_1, conv4_1, conv5_1 in the VGG-19 network, and zero otherwise. The content loss is computed by w l = 1 {\displaystyle w_{l}=1} for conv4_2, and zero otherwise. The ratio α / β ∈ [ 5 , 50 ] × 10 − 4 {\displaystyle \alpha /\beta \in [5,50]\times 10^{-4}} . === Training === Image x → {\displaystyle {\vec {x}}} is initially approximated by adding a small amount of white noise to input image p → {\displaystyle {\vec {p}}} and feeding it through the CNN. Then we successively backpropagate this loss through the network with the CNN weights fixed in order to update the pixels of x → {\displaystyle {\vec {x}}} . After several thousand epochs of training, an x → {\displaystyle {\vec {x}}} (hopefully) emerges that matches the style of a → {\displaystyle {\vec {a}}} and the content of p → {\displaystyle {\vec {p}}} . As of 2017, when implemented on a GPU, it takes a few minutes to converge. == Extensions == In some practical implementations, it is noted that the resulting image has too much high-frequency artifact, which can be suppressed by adding the total variation to the total loss. Compared to VGGNet, AlexNet does not work well for neural style transfer. NST has also been extended to videos. Subsequent work improved the speed of NST for images by using special-purpose normalizations. In a paper by Fei-Fei Li et al. adopted a different regularized loss metric and accelerated method for training to produce results in real-time (three orders of magnitude faster than Gatys). Their idea was to use not the pixel-based loss defined above but rather a 'perceptual loss' measuring t

  Read more →
• 

  Creator economy

  The creator economy, also known as influencer economy, is a platform-driven economy in which creators produce content, products, or services and distribute them directly to their audience through social media platforms and emerging technologies. This economic model is based on the ability of creators to build and maintain communities of users, monetizing their creative activity through multiple channels including advertising, sponsorships, product sales, crowdfunding, and subscription-based services. Creators include various professional categories such as social media influencers, YouTubers, bloggers, artists, online educators, podcasters, and independent professionals, who use platforms as infrastructure to reach their audience without necessarily relying on traditional intermediaries in the cultural and media industry. According to Goldman Sachs Research, the ongoing growth of the creator economy will likely benefit companies that possess a combination of factors, including a large global user base, access to substantial capital, robust AI-powered recommendation engines, versatile monetization tools, comprehensive data analytics, and integrated e-commerce options. Examples of creator economy software platforms include YouTube, TikTok, Instagram, Facebook, Twitch, Spotify, Substack, OnlyFans and Patreon. == History == The term "creator" was coined by YouTube in 2011 to be used instead of "YouTube star", an expression that at the time could only apply to famous individuals on the platform. The term has since become omnipresent and is used to describe anyone creating any form of online content. A number of platforms such as TikTok, Snapchat, YouTube, and Facebook have set up funds with which to pay creators. == Criticism == The large majority of content creators derive no monetary gain for their creations, with most of the benefits accruing to the platforms who can make significant revenues from their uploads. As few as 0.1% of creators are able to earn a living through their channels.

  Read more →
• 

  Electronic submission

  Electronic submission refers to the submission of a document by electronic means: that is, via e-mail or a web form on the Internet, or on an electronic medium such as a compact disc, a hard disk or a USB flash drive. Traditionally, the term "manuscript" referred to anything that was explicitly "written by hand". However, in popular usage and especially in the context of computers and the internet, the term "manuscript" may even refer to documents (text or otherwise) typed out or prepared on typewriters and computers and can be extended to digital photographs and videos, and online surveys too. In other words, any manuscript prepared and submitted online can be considered to be an electronic submission. == History and early usage == There is no concrete data indicating when and by whom were electronic submissions used for the first time. However, research based universities in several countries have been encouraging the collection of course assignments and projects in the form of electronic submissions for almost a decade now. Several governments and organizations are also switching to electronic submissions for the collection of research papers, grant applications and government application forms. == Types of electronic submissions == Since modern computers can store and process information and data in virtually any format and with the Internet allowing easy transfer of this data, the number of scenarios in which submissions can be collected electronically has increased exponentially in the last few years. Some of these scenarios are described below. In most of these scenarios, submissions were collected on hard paper until the Information Technology revolution occurred. === Academic Submissions === Teachers, professors and teaching assistants often collect course assignments and projects electronically. Electronic submissions are usually collected using a web-based system which more often than not also helps in the management of submissions collected and stored on it. (Explained By Henny L, University of Lethbridge, AB, Canada) === Research Papers === In call-for-paper or academic conferences, prospective presenters are usually asked to submit a short abstract or a full paper on their presentation or research work electronically, which is reviewed before being accepted for the conference. === Proposals for Grants === Several grant-giving organizations like the NSA, W3C, NIA, NIH etc. require grant seekers to submit a proposal which if accepted result in the desired grants. A majority of these proposals are now submitted electronically on systems that also help in the managing and tracking the proposals submitted. === Articles for Publication === Magazines, newspapers and other publishing houses have begun accepting electronic submissions for articles from various sources - both internal (by journalists and writers hired by them) as well as external (by users and popular readers). The submitted articles are stored on a server hosted by the publication house or by a third-party Archived 2019-10-13 at the Wayback Machine vendor and are usually evaluated before being given a green signal. === Contests and Competition Entries === Almost every kind of contest or competition requires participants to submit an entry in a format described by the organizers of the contest. If the contest is an Internet-based one, then the entries or nominations for the contest are collected electronically using e-mail or other electronic means depending on feasibility and the choice of the organizers. === Government Applications === The governments of several countries are turning to electronic submission of applications and forms for various government procedures. Electronic submissions allow easier management of the applications and forms submitted. === Legal documents === Many legal documents may be submitted to the courts electronically. In England and Wales, the Civil Procedure Rules include a suitable "document exchange" as an acceptable "method of service". Case law in employment law cases has established that where a claim is submitted electronically, a prudent legal adviser should "check that it has been received and there must be systems in place for doing that". === Resumés and CVs === It has become commonplace for job-seekers to submit soft copies (electronic versions) of their resumés and CVs to recruiting agencies and online job portals. This is usually done over the Internet using e-mail or a pre-hosted web-based system. == Submission management systems == The art and science of collecting and managing electronic submissions is called Submission Management. Certain software vendors have begun developing submission management systems to assist in the collection, tracking and management of complex submission processes realized electronically. Most of these systems are web based and accessible from any device with a browser and an Internet connection. However, a majority of these systems are application specific and cannot be applied to all submission management scenarios. == Resistance to electronic submissions == Despite the easier management and tracking of electronic submissions compared to their paper-based counterparts, widespread adoption and use of electronic submissions and systems for managing them has been hampered by several facts, which include but are not limited to: Inconvenience while drawing figures, diagrams and equations on a computer Resistance to change and adoption of new technologies Lack of or limited access to the Internet.

  Read more →