AI Face Fusion

AI Face Fusion — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  PhotoLine

  PhotoLine is a general purpose bitmap and vector graphics editor developed and published by Computerinsel GmbH for Windows, macOS, and Linux/Wine. It was originally created in 1995 by Gerhard Huber and Martin Huber. The program combines bitmap and vector graphics editing in one seamless working application unlike most graphics software which tend to focus on either bitmap or vector editing and output. PhotoLine is considered as a market competitor to Adobe Photoshop. == Features == PhotoLine edits and composes multi-layer raster and vector images with deep support for masking and alpha compositing and with full color management. Editing and color management in PhotoLine is mostly non-destructive. Image data in layers is preserved without loss of information regardless of the document's image mode or layer transformation. color depth, image resolution, color model, and ICC profile are preserved for each individual layer or group of layers. Layers can be cloned and reused anywhere in the layer stack, including repurposed as layer masks. Layer blending and compositing in PhotoLine supports common blend modes, and features a layer blend range of -200 to +200 percent. It is also possible to control which channels are blended for each layer, adjustment layer, and layer mask or group of layers. Filters, adjustment layers, and brushes have access to Lab and HIS color modes (HIS is a variant of HSL), separately of the color model of the underlying image layer. In Addition to raster and vector editing, PhotoLine can be used for small desktop publishing projects. Multi-page documents with page spreads and text flow between text frames and pages are supported. Character and paragraph styles can be defined. Spot colors, bleed settings, a baseline grid, a table of contents generator, and PDF/X support help with these projects. PhotoLine is however much more limited when compared to dedicated publishing software such as Adobe InDesign or QuarkXPress. PhotoLine incorporates the Open-source software library LibRaw to read raw images from digital cameras for import. Developing these files is non-destructive with a choice of embedding the RAW image data either in the PhotoLine document or link to the external RAW image file. PhotoLine can open raw files as linear unmodified and non color managed source images. Photoshop PSD files can be imported and exported. Core functionality of PhotoLine can be extended through standard Photoshop filter plugins, the G'MIC digital image processing framework, and PSP tubes. External programs can be linked for a seamless round-trip workflow and files can be sent directly for processing in third-party design applications. Custom functionality is further supported through scripting and macro recording. == Early history == Developed by two brothers, Gerhard Huber and Martin Huber, PhotoLine was first released in January 1996 on the Atari ST line of personal computers from Atari Corporation. Previously, Gerhard and Martin had worked on making graphics cards for Atari computers and writing drivers for image scanners. Atari's market share was declining, and the brothers considered developing a video game to expand the business. This led them to search for image editing software that would run on Atari computers and fit their game project. Only an image editor called tms Cranach came close to what Gerhard and Martin had in mind. tms Cranach was a Raster graphics editor running on Atari's MegaST/STe, TT030, and Falcon030 systems. However, Cranach turned out to be expensive software and complicated to use. The brothers contacted tms (Cranach's developers) and this resulted in an offer from tms to purchase Cranach and its source code, as tms intended to exit the Atari software market. After the purchase of Cranach and its source code Gerhard and Martin initially continued to sell Cranach, but sales were low. In 1995 the two decided to start developing a new graphics editor called "PhotoLine". PhotoLine was developed from scratch and written in C++. It nevertheless contained a lot of know-how from Cranach (which was written in C). PhotoLine first release was launched one year later in 1996. With the growing popularity of Microsoft Windows, the release of Windows 95, and the limiting graphics hardware on the Atari platforms, the developers switched development platforms and continued development of PhotoLine for Windows only. The first Windows version (PhotoLine 2.2) was released in the middle of 1997. Shortly after, the Atari version was discontinued and saw its final release as PhotoLine 2.30. The Huber brothers released this final Atari version into the public domain in 2012. The first Classic Mac OS version of PhotoLine 6 appeared in 1999 after many ex-Atari users who had switched to Mac OS pressured the PhotoLine developers to release an Apple port. == Linux Support == PhotoLine runs natively under Windows and MacOS. While a native Linux version of PhotoLine is not available, running PhotoLine under Wine is actively supported and maintained by the developers. Running PhotoLine under Linux/Wine PhotoLine enables the user to allow Little CMS to fully support color management under Linux instead of the native OS CMS. == File format == Native PhotoLine files have the extension .PLD, which is an abbreviation of "PhotoLine Document". It can contain embedded JPEG, PNG, or camera raw images. It contains a preview image in JPEG or PNG format, which is used by the operating system or third-party applications to display a thumbnail of its contents. Thumbnails are natively supported on MacOS X. During installation on Windows the user is presented with an option to install a PLD thumbnail preview driver which enables thumbnails of PLD content in Windows Explorer. Alternatively, the FastPictureViewer Standalone Codec Pack provides the ability to display PLD thumbnails in Windows Explorer. == Version History == PhotoLine was first developed for the Atari ST computer. Version 2 was the first version for Windows, and since version 6 PhotoLine is also available for MacOS.

  Read more →
• 

  Sayre's paradox

  Sayre's paradox is a dilemma encountered in the design of automated handwriting recognition systems. A standard statement of the paradox is that a cursively written word cannot be recognized without being segmented and cannot be segmented without being recognized. The paradox was first articulated in a 1973 publication by Kenneth M. Sayre, after whom it was named. == Nature of the problem == It is relatively easy to design automated systems capable of recognizing words inscribed in a printed format. Such words are segmented into letters by the very act of writing them on the page. Given templates matching typical letter shapes in a given language, individual letters can be identified with a high degree of probability. In cases of ambiguity, probable letter sequences can be compared with a selection of properly spelled words in that language (called a lexicon). If necessary, syntactic features of the language can be applied to render a generally accurate identification of the words in question. Printed-character recognition systems of this sort are commonly used in processing standardized government forms, in sorting mail by zip code, and so forth. In cursive writing, however, letters comprising a given word typically flow sequentially without gaps between them. Unlike a sequence of printed letters, cursively connected letters are not segmented in advance. Here is where Sayre's Paradox comes into play. Unless the word is already segmented into letters, template-matching techniques like those described above cannot be applied. That is, segmentation is a prerequisite for word recognition. But there are no reliable techniques for segmenting a word into letters unless the word itself has been identified. Word recognition requires letter segmentation, and letter segmentation requires word recognition. There is no way a cursive writing recognition system employing standard template-matching techniques can do both simultaneously. Advantages to be gained by use of automated cursive writing recognition systems include routing mail with handwritten addresses, reading handwritten bank checks, and automated digitalization of hand-written documents. These are practical incentives for finding ways of circumventing Sayre's Paradox. == Avoiding the paradox == One way of ameliorating the adverse effects of the paradox is to normalize the word inscriptions to be recognized. Normalization amounts to eliminating idiosyncrasies in the penmanship of the writer, such as unusual slope of the letters and unusual slant of the cursive line. This procedure can increase the probability of a correct match with a letter template, resulting in an incremental improvement in the success rate of the system. Since improvement of this sort still depends on accurate segmentation, however, it remains subject to the limitations of Sayre's Paradox. Researchers have come to realize that the only way to circumvent the paradox is by use of procedures that do not rely on accurate segmentation. == Directions of current research == Segmentation is accurate to the extent that it matches distinctions among letters in the actual inscriptions presented to the system for recognition (the input data). This is sometimes referred to as “explicit segmentation”. “Implicit segmentation,” by contrast, is division of the cursive line into more parts than the number of actual letters in the cursive line itself. Processing these “implicit parts” to achieve eventual word identification requires specific statistical procedures involving hidden Markov models (HMM). A Markov model is a statistical representation of a random process, which is to say a process in which future states are independent of states occurring before the present. In such a process, a given state is dependent only on the conditional probability of its following the state immediately before it. An example is a series of outcomes from successive casts of a die. An HMM is a Markov model, individual states of which are not fully known. Conditional probabilities between states are still determinate, but the identities of individual states are not fully disclosed. Recognition proceeds by matching HMMs of words to be recognized with previously prepared HMMs of words in the lexicon. The best match in a given case is taken to indicate the identity of the handwritten word in question. As with systems based on explicit segmentation, automated recognition systems based on implicit segmentation are judged more or less successful according to the percentage of correct identifications they accomplish. Instead of explicit segmentation techniques, most automated handwriting recognition systems today employ implicit segmentation in conjunction with HMM-based matching procedures. The constraints epitomized by Sayre's Paradox are largely responsible for this shift in approach.

  Read more →
• 

  Deep learning

  In machine learning, deep learning (DL) focuses on utilizing multilayered neural networks to perform tasks such as classification, regression, and representation learning. The field takes inspiration from biological neuroscience and revolves around stacking artificial neurons into layers and "training" them to process data. The adjective "deep" refers to the use of multiple layers (ranging from three to several hundred or thousands) in the network. Methods used can be supervised, semi-supervised or unsupervised. Some common deep learning network architectures include fully connected networks, deep belief networks, recurrent neural networks, convolutional neural networks, generative adversarial networks, transformers, and neural radiance fields. These architectures have been applied to fields including computer vision, speech recognition, natural language processing, machine translation, bioinformatics, drug design, medical image analysis, climate science, material inspection and board game programs, where they have produced results comparable to and in some cases surpassing human expert performance. Early forms of neural networks were inspired by information processing and distributed communication nodes in biological systems, particularly the human brain. However, current neural networks do not intend to model the brain function of organisms, and are generally seen as low-quality models for that purpose. == Overview == Most modern deep learning models are based on multi-layered neural networks such as convolutional neural networks and transformers, although they can also include propositional formulas or latent variables organized layer-wise in deep generative models such as the nodes in deep belief networks and deep Boltzmann machines. Fundamentally, deep learning refers to a class of machine learning algorithms in which a hierarchy of layers is used to transform input data into a progressively more abstract and composite representation. For example, in an image recognition model, the raw input may be an image (represented as a tensor of pixels). The first representational layer may attempt to identify basic shapes such as lines and circles, the second layer may compose and encode arrangements of edges, the third layer may encode a nose and eyes, and the fourth layer may recognize that the image contains a face. Importantly, a deep learning process can learn which features to optimally place at which level on its own. Prior to deep learning, machine learning techniques often involved hand-crafted feature engineering to transform the data into a more suitable representation for a classification algorithm to operate on. In the deep learning approach, features are not hand-crafted and the model discovers useful feature representations from the data automatically. This does not eliminate the need for hand-tuning; for example, varying numbers of layers and layer sizes can provide different degrees of abstraction. The word "deep" in "deep learning" refers to the number of layers through which the data is transformed. More precisely, deep learning systems have a substantial credit assignment path (CAP) depth. The CAP is the chain of transformations from input to output. CAPs describe potentially causal connections between input and output. For a feedforward neural network, the depth of the CAPs is that of the network and is the number of hidden layers plus one (as the output layer is also parameterized). For recurrent neural networks, in which a signal may propagate through a layer more than once, the CAP depth is potentially unlimited. No universally agreed-upon threshold of depth divides shallow learning from deep learning, but most researchers agree that deep learning involves CAP depth higher than two. CAP of depth two has been shown to be a universal approximator in the sense that it can emulate any function. Beyond that, more layers do not add to the function approximator ability of the network. Deep models (CAP > two) are able to extract better features than shallow models and hence, extra layers help in learning the features effectively. Deep learning architectures can be constructed with a greedy layer-by-layer method. Deep learning helps to disentangle these abstractions and pick out which features improve performance. Deep learning algorithms can be applied to unsupervised learning tasks. This is an important benefit because unlabeled data is more abundant than labeled data. Examples of deep structures that can be trained in an unsupervised manner are deep belief networks. The term deep learning was introduced to the machine learning community by Rina Dechter in 1986, and to artificial neural networks by Igor Aizenberg and colleagues in 2000, in the context of Boolean threshold neurons. The etymology of the term is more complicated. == Interpretations == Deep neural networks are generally interpreted in terms of the universal approximation theorem or probabilistic inference. The classic universal approximation theorem concerns the capacity of feedforward neural networks with a single hidden layer of finite size to approximate continuous functions. In 1989, the first proof was published by George Cybenko for sigmoid activation functions and was generalised to feed-forward multi-layer architectures in 1991 by Kurt Hornik. Recent work also showed that universal approximation also holds for non-bounded activation functions such as Kunihiko Fukushima's rectified linear unit. The universal approximation theorem for deep neural networks concerns the capacity of networks with bounded width but the depth is allowed to grow. Lu et al. proved that if the width of a deep neural network with ReLU activation is strictly larger than the input dimension, then the network can approximate any Lebesgue integrable function; if the width is smaller or equal to the input dimension, then a deep neural network is not a universal approximator. The probabilistic interpretation derives from the field of machine learning. It features inference, as well as the optimization concepts of training and testing, related to fitting and generalization, respectively. More specifically, the probabilistic interpretation considers the activation nonlinearity as a cumulative distribution function. The probabilistic interpretation led to the introduction of dropout as regularizer in neural networks. The probabilistic interpretation was introduced by researchers including Hopfield, Widrow and Narendra and popularized in surveys such as the one by Bishop. == History == === Before 1980 === There are two types of artificial neural network (ANN): feedforward neural network (FNN) or multilayer perceptron (MLP) and recurrent neural networks (RNN). RNNs have cycles in their connectivity structure, whereas FNNs do not. In the 1920s, Wilhelm Lenz and Ernst Ising created the Ising model which is essentially a non-learning RNN architecture consisting of neuron-like threshold elements. In 1972, Shun'ichi Amari made this architecture adaptive. His learning RNN was republished by John Hopfield in 1982. Other early recurrent neural networks were published by Kaoru Nakano in 1971. Already in 1948, Alan Turing produced work on "Intelligent Machinery" that was not published in his lifetime, containing "ideas related to artificial evolution and learning RNNs". Frank Rosenblatt (1958) proposed the perceptron, an MLP with 3 layers: an input layer, a hidden layer with randomized weights that did not learn, and an output layer. He later published a 1962 book that also introduced variants and computer experiments, including a version with four-layer perceptrons "with adaptive preterminal networks" where the last two layers have learned weights (here he credits H. D. Block and B. W. Knight). The book cites an earlier network by R. D. Joseph (1960) "functionally equivalent to a variation of" this four-layer system (the book mentions Joseph over 30 times). Should Joseph therefore be considered the originator of proper adaptive multilayer perceptrons with learning hidden units? Unfortunately, the learning algorithm was not a functional one, and fell into oblivion. The first working deep learning algorithm was the Group method of data handling, a method to train arbitrarily deep neural networks, published by Alexey Ivakhnenko and Lapa in 1965. They regarded it as a form of polynomial regression, or a generalization of Rosenblatt's perceptron to handle more complex, nonlinear, and hierarchical relationships. A 1971 paper described a deep network with eight layers trained by this method, which is based on layer by layer training through regression analysis. Superfluous hidden units are pruned using a separate validation set. Since the activation functions of the nodes are Kolmogorov-Gabor polynomials, these were also the first deep networks with multiplicative units or "gates". The first deep learning multilayer perceptron trained by stochastic gradient descent was published in 1967 by Shun'ichi

  Read more →
• 

  Contextual AI

  Contextual AI is an enterprise software company based in Mountain View, California. It develops a platform for building specialized Retrieval-Augmented Generation (RAG) agents for enterprise use. The company was founded in 2023 by Douwe Kiela and Amanpreet Singh, both former AI researchers at Facebook AI Research (FAIR) and Hugging Face. Douwe Kiela previously led the Meta research team that introduced the Retrieval-Augmented Generation (RAG) approach in 2020. Contextual AI focuses on enterprise generative AI applications using RAG 2.0 technology, with deployments primarily in the technology, banking, finance and media sectors. == History == In June 2023, Contextual AI announced it had raised $20 million in a seed funding round led by Bain Capital Ventures (BCV), with participation from Lightspeed Venture Partners, Greycroft, SV Angel, and several angel investors. In August 2024, the company raised $80 million in a Series A funding round led by Greycroft, with participation from previous investors including Bain Capital Ventures, Lightspeed, and Conviction Partners. The round also included new backers such as Bezos Expeditions, NVentures (Nvidia), HSBC Ventures, and Snowflake Ventures. == Features == Retrieval-Augmented Generation (RAG) is an artificial intelligence framework that integrates information retrieval with text generation to improve the performance of large language models (LLMs) on complex, knowledge-intensive tasks. It was introduced in 2020 by researchers at Meta AI, including Douwe Kiela, Patrick Lewis and others, in their paper Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. RAG enables language models to access and incorporate external information, such as proprietary databases or real-time web content, at query time, instead of relying solely on pre-trained, internal, static knowledge. This architecture addresses common limitations of standard LLMs, including hallucination, outdated information, and lack of attribution to source materials. RAG systems retrieve relevant context through a variety of techniques - including vector search, keyword search, text-to-SQL - and feeds this context into the language model to generate responses. The approach improves factual accuracy, supports domain-specific customization, enables citation of sources, and allows for more updated information without retraining the model itself. General Availability. In January 2025, Contextual AI announced the general availability of its enterprise platform for building specialized RAG agents. Early adopters included Qualcomm, which used the platform for their Customer Engineering team needs. Grounded Language Model. In March 2025, the company introduced a Grounded Language Model (GLM) for factual accuracy in enterprise AI applications. Reranker. In March 2025, Contextual AI released an instruction-following reranker that allows users to influence the ranking of retrieved documents through natural language instructions, such as prioritizing recent files, specific formats, or content from designated sources. == Applications == Contextual AI's platform has been adopted across a range of industries, including finance, technology, media and professional services. Clients include Fortune 500 companies such as Qualcomm and HSBC.

  Read more →
• 

  Turret lathe

  A turret lathe is a form of metalworking lathe that is used for repetitive production of duplicate parts, which by the nature of their cutting process are usually interchangeable. It evolved from earlier lathes with the addition of the turret, which is an indexable toolholder that allows multiple cutting operations to be performed, each with a different cutting tool, in easy, rapid succession, with no need for the operator to perform set-up tasks in between (such as installing or uninstalling tools) or to control the toolpath. The latter is due to the toolpath's being controlled by the machine, either in jig-like fashion, via the mechanical limits placed on it by the turret's slide and stops, or via digitally-directed servomechanisms for computer numerical control lathes. The name derives from the way early turrets took the general form of a flattened cylindrical block mounted to the lathe's cross-slide, capable of rotating about the vertical axis and with toolholders projecting out to all sides, and thus vaguely resembled a swiveling gun turret. Capstan lathe is the usual name in the UK and Commonwealth, though the two terms are also used in contrast: see below, Capstan versus turret. == History == Turret lathes became indispensable to the production of interchangeable parts and for mass production. The first turret lathe was built by Stephen Fitch in 1845 to manufacture screws for pistol percussion parts. In the mid-nineteenth century, the need for interchangeable parts for Colt revolvers enhanced the role of turret lathes in achieving this goal as part of the "American system" of manufacturing arms. Clock-making and bicycle manufacturing had similar requirements. Christopher Spencer invented the first fully automated turret lathe in 1873, which led to designs using cam action or hydraulic mechanisms. From the late-19th through mid-20th centuries, turret lathes, both manual and automatic (i.e., screw machines and chuckers), were one of the most important classes of machine tools for mass production. They were used extensively in the mass production for the war effort in World War II. The U.S. company Warner & Swasey was one of the premier brands in heavy turret lathes between the 1910s and 1960s; it became the world's largest manufacturer of such lathes by 1928. During World War II, it employed 7,000 people and produced half of the turret lathes manufactured in the United States. == Types == There are many variants of the turret lathe. They can be most generally classified by size (small, medium, or large); method of control (manual, automated mechanically, or automated via computer (numerical control (NC) or computer numerical control (CNC)); and bed orientation (horizontal or vertical). === Archetypical: horizontal, manual === In the late 1830s a "capstan lathe" with a turret was patented in Britain. The first American turret lathe was invented by Stephen Fitch in 1845. The archetypical turret lathe, and the first in order of historical appearance, is the horizontal-bed, manual turret lathe. The term "turret lathe" without further qualification is still understood to refer to this type. The formative decades for this class of machine were the 1840s through 1860s, when the basic idea of mounting an indexable turret on a bench lathe or engine lathe was born, developed, and disseminated from the originating shops to many other factories. Some important tool-builders in this development were Stephen Fitch; Gay, Silver & Co.; Elisha K. Root of Colt; J.D. Alvord of the Sharps Armory; Frederick W. Howe, Richard S. Lawrence, and Henry D. Stone of Robbins & Lawrence; J.R. Brown of Brown & Sharpe; and Francis A. Pratt of Pratt & Whitney. Various designers at these and other firms later made further refinements. === Semi-automatic === Sometimes machines similar to those above, but with power feeds and automatic turret-indexing at the end of the return stroke, are called "semi-automatic turret lathes". This nomenclature distinction is blurry and not consistently observed. The term "turret lathe" encompasses them all. During the 1860s, when semi-automatic turret lathes were developed, they were sometimes called "automatic". What we today would call "automatics", that is, fully automatic machines, had not been developed yet. During that era both manual and semi-automatic turret lathes were sometimes called "screw machines", although we today reserve that term for fully automatic machines. === Automatic === During the 1870s through 1890s, the mechanically automated "automatic" turret lathe was developed and disseminated. These machines can execute many part-cutting cycles without human intervention. Thus the duties of the operator, which were already greatly reduced by the manual turret lathe, were even further reduced, and productivity increased. These machines use cams to automate the sliding and indexing of the turret and the opening and closing of the chuck. Thus, they execute the part-cutting cycle somewhat analogously to the way in which an elaborate cuckoo clock performs an automated theater show. Small- to medium-sized automatic turret lathes are usually called "screw machines" or "automatic screw machines", while larger ones are usually called "automatic chucking lathes", "automatic chuckers", or "chuckers". Such machine tools of the "automatic" variety, which in the pre-computer era meant mechanically automated, had already reached a highly advanced state by World War I. === Computer numerical control === When World War II ended, the digital computer was poised to develop from a colossal laboratory curiosity into a practical technology that could begin to disseminate into business and industry. The advent of computer-based automation in machine tools via numerical control (NC) and then computer numerical control (CNC) displaced to a large extent, but not at all completely, the previously existing manual and mechanically automated machines. Numerically controlled turrets allow automated selection of tools on a turret. CNC lathes may be horizontal or vertical in orientation and mount six separate tools on one or more turrets. Such machine tools can work in two axes per turret, with up to six axes being feasible for complex work. === Vertical === Vertical turret lathes have the workpiece held vertically, which allows the headstock to sit on the floor and the faceplate to become a horizontal rotating table, analogous to a huge potter's wheel. This is useful for the handling of very large, heavy, short workpieces. Vertical lathes in general are also called "vertical boring mills" or often simply "boring mills"; therefore a vertical turret lathe is a vertical boring mill equipped with a turret. == Other variations == === Capstan versus turret === The term "capstan lathe" overlaps in sense with the term "turret lathe" to a large extent. In many times and places, it has been understood to be synonymous with "turret lathe". In other times and places it has been held in technical contradistinction to "turret lathe", with the difference being in whether the turret's slide is fixed to the bed (ram-type turret) or slides on the bed's ways (saddle-type turret). The difference in terminology is mostly a matter of United Kingdom and Commonwealth usage versus United States usage. === Flat === A subtype of horizontal turret lathe is the flat-turret lathe. Its turret is flat (and analogous to a rotary table), allowing the turret to pass beneath the part. Patented by James Hartness of Jones & Lamson, and first disseminated in the 1890s, it was developed to provide more rigidity via requiring less overhang in the tool setup, especially when the part is relatively long. === Hollow-hexagon === Hollow-hexagon turret lathes competed with flat-turret lathes by taking the conventional hexagon turret and making it hollow, allowing the part to pass into it during the cut, analogously to how the part would pass over the flat turret. In both cases, the main idea is to increase rigidity by allowing a relatively long part to be turned without the tool overhang that would be needed with a conventional turret, which is not flat or hollow. === Monitor lathe === The term "monitor lathe" formerly (1860s–1940s) referred to the class of small- to medium-sized manual turret lathes used on relatively small work. The name was inspired by the monitor-class warships, which the monitor lathe's turret resembled. Today, lathes of such appearance, such as the Hardinge DSM-59 and its many clones, are still common, but the name "monitor lathe" is no longer current in the industry. === Toolpost turrets and tailstock turrets === Turrets can be added to non-turret lathes (bench lathes, engine lathes, toolroom lathes, etc.) by mounting them on the toolpost, tailstock, or both. Often these turrets are not as large as a turret lathe's, and they usually do not offer the sliding and stopping that a turret lathe's turret does; but they do offer the ability to index through successive tool

  Read more →
• 

  Sentence extraction

  Sentence extraction is a technique used for automatic summarization of a text. In this shallow approach, statistical heuristics are used to identify the most salient sentences of a text. Sentence extraction is a low-cost approach compared to more knowledge-intensive deeper approaches which require additional knowledge bases such as ontologies or linguistic knowledge. In short, sentence extraction works as a filter that allows only meaningful sentences to pass. The major downside of applying sentence-extraction techniques to the task of summarization is the loss of coherence in the resulting summary. Nevertheless, sentence extraction summaries can give valuable clues to the main points of a document and are frequently sufficiently intelligible to human readers. == Procedure == Usually, a combination of heuristics is used to determine the most important sentences within the document. Each heuristic assigns a (positive or negative) score to the sentence. After all heuristics have been applied, the highest-scoring sentences are included in the summary. The individual heuristics are weighted according to their importance. === Early approaches and some sample heuristics === Seminal papers which laid the foundations for many techniques used today have been published by Hans Peter Luhn in 1958 and H. P Edmundson in 1969. Luhn proposed to assign more weight to sentences at the beginning of the document or a paragraph. Edmundson stressed the importance of title-words for summarization and was the first to employ stop-lists in order to filter uninformative words of low semantic content (e.g. most grammatical words such as of, the, a). He also distinguished between bonus words and stigma words, i.e. words that probably occur together with important (e.g. the word form significant) or unimportant information. His idea of using key-words, i.e. words which occur significantly frequently in the document, is still one of the core heuristics of today's summarizers. With large linguistic corpora available today, the tf–idf value which originated in information retrieval, can be successfully applied to identify the key words of a text: If for example the word cat occurs significantly more often in the text to be summarized (TF = "term frequency") than in the corpus (IDF means "inverse document frequency"; here the corpus is meant by document), then cat is likely to be an important word of the text; the text may in fact be a text about cats.

  Read more →
• 

  Semantic decomposition (natural language processing)

  A semantic decomposition is an algorithm that breaks down the meanings of phrases or concepts into less complex concepts. The result of a semantic decomposition is a representation of meaning. This representation can be used for tasks, such as those related to artificial intelligence or machine learning. Semantic decomposition is common in natural language processing applications. The basic idea of a semantic decomposition is taken from the learning skills of adult humans, where words are explained using other words. It is based on Meaning-text theory. Meaning-text theory is used as a theoretical linguistic framework to describe the meaning of concepts with other concepts. == Background == Given that an AI does not inherently have language, it is unable to think about the meanings behind the words of a language. An artificial notion of meaning needs to be created for a strong AI to emerge. Creating an artificial representation of meaning requires the analysis of what meaning is. Many terms are associated with meaning, including semantics, pragmatics, knowledge and understanding or word sense. Each term describes a particular aspect of meaning, and contributes to a multitude of theories explaining what meaning is. These theories need to be analyzed further to develop an artificial notion of meaning best fit for our current state of knowledge. == Graph representations == Representing meaning as a graph is one of the two ways that both an AI cognition and a linguistic researcher think about meaning (connectionist view). Logicians utilize a formal representation of meaning to build upon the idea of symbolic representation, whereas description logics describe languages and the meaning of symbols. This contention between 'neat' and 'scruffy' techniques has been discussed since the 1970s. Research has so far identified semantic measures and with that word-sense disambiguation (WSD) - the differentiation of meaning of words - as the main problem of language understanding. As an AI-complete environment, WSD is a core problem of natural language understanding. AI approaches that use knowledge-given reasoning creates a notion of meaning combining the state of the art knowledge of natural meaning with the symbolic and connectionist formalization of meaning for AI. The abstract approach is shown in Figure. First, a connectionist knowledge representation is created as a semantic network consisting of concepts and their relations to serve as the basis for the representation of meaning. This graph is built out of different knowledge sources like WordNet, Wiktionary, and BabelNET. The graph is created by lexical decomposition that recursively breaks each concept semantically down into a set of semantic primes. The primes are taken from the theory of Natural Semantic Metalanguage, which has been analyzed for usefulness in formal languages. Upon this graph marker passing is used to create the dynamic part of meaning representing thoughts. The marker passing algorithm, where symbolic information is passed along relations form one concept to another, uses node and edge interpretation to guide its markers. The node and edge interpretation model is the symbolic influence of certain concepts. Future work uses the created representation of meaning to build heuristics and evaluate them through capability matching and agent planning, chatbots or other applications of natural language understanding.

  Read more →
• 

  Lose It!

  Lose It! is an American health and wellness mobile app developed by FitNow, Inc. The app generates calorie budgets for users by tracking weight, exercise, food and calorie intake, and personal goals, primarily to assist them in achieving weight loss. == History == Lose It! was developed in Boston and debuted in 2008. The app and its associated company were founded by J.J. Allaire, Charles Teague and Paul Dicristina. Prior to founding Lose It!, Teague and Allaire had founded the online research tool Onfolio, which was acquired by Microsoft in 2006. The Lose It! app was originally released as an iOS app before being released as a website in 2010 and an Android app in 2011. In 2015, Lose It! announced plans to release the app internationally. Lose It! was also available as an app for Apple Watch at its launch in 2015. The app’s “Snap It” feature, which allows users to approximate calorie counts by taking pictures of their daily meals and snacks, was released in beta in 2016. Snap It was named an Innovation Awards Honoree at the 2017 Consumer Electronics Show in Las Vegas. In 2020, Patrick Wetherille, one of the company’s earliest employees, was appointed chief executive officer. == App == Lose It! is weight loss app. The app allows users to set goals such as increasing strength, overall health/maintenance, and weight loss. It provides users recommended calorie budgets based on data such as their current weight and their desired weight. Lose It! also tracks data such as exercise/activity level and food consumption and allows users to track calories consumed by scanning barcodes for food products then retrieving calorie information for products. The app can also estimate the amount of calories in a food products. Lose It! has integration features connecting it to other apps such as Fitbit and Runkeeper. It also has social features such as joining groups and sharing progress with friends. The Premium version of the app allows users to track foods according to specific diets like keto, heart healthy or Mediterranean.

  Read more →
• 

  Convolutional layer

  In artificial neural networks, a convolutional layer is a type of network layer that applies a convolution operation to the input. Convolutional layers are some of the primary building blocks of convolutional neural networks (CNNs), a class of neural network most commonly applied to images, video, audio, and other data that have the property of uniform translational symmetry. The convolution operation in a convolutional layer involves sliding a small window (called a kernel or filter) across the input data and computing the dot product between the values in the kernel and the input at each position. This process creates a feature map that represents detected features in the input. == Concepts == === Kernel === Kernels, also known as filters, are small matrices of weights that are learned during the training process. Each kernel is responsible for detecting a specific feature in the input data. The size of the kernel is a hyperparameter that affects the network's behavior. === Convolution === For a 2D input x {\displaystyle x} and a 2D kernel w {\displaystyle w} , the 2D convolution operation can be expressed as: y [ i , j ] = ∑ m = 0 k h − 1 ∑ n = 0 k w − 1 x [ i + m , j + n ] ⋅ w [ m , n ] {\displaystyle y[i,j]=\sum _{m=0}^{k_{h}-1}\sum _{n=0}^{k_{w}-1}x[i+m,j+n]\cdot w[m,n]} where k h {\displaystyle k_{h}} and k w {\displaystyle k_{w}} are the height and width of the kernel, respectively. This generalizes immediately to nD convolutions. Commonly used convolutions are 1D (for audio and text), 2D (for images), and 3D (for spatial objects, and videos). === Stride === Stride determines how the kernel moves across the input data. A stride of 1 means the kernel shifts by one pixel at a time, while a larger stride (e.g., 2 or 3) results in less overlap between convolutions and produces smaller output feature maps. === Padding === Padding involves adding extra pixels around the edges of the input data. It serves two main purposes: Preserving spatial dimensions: Without padding, each convolution reduces the size of the feature map. Handling border pixels: Padding ensures that border pixels are given equal importance in the convolution process. Common padding strategies include: No padding/valid padding. This strategy typically causes the output to shrink. Same padding: Any method that ensures the output size same as input size is a same padding strategy. Full padding: Any method that ensures each input entry is convolved over for the same number of times is a full padding strategy. Common padding algorithms include: Zero padding: Add zero entries to the borders of input. Mirror/reflect/symmetric padding: Reflect the input array on the border. Circular padding: Cycle the input array back to the opposite border, like a torus. The exact numbers used in convolutions is complicated, for which we refer to (Dumoulin and Visin, 2018) for details. == Variants == === Standard === The basic form of convolution as described above, where each kernel is applied to the entire input volume. === Depthwise separable === Depthwise separable convolution separates the standard convolution into two steps: depthwise convolution and pointwise convolution. The depthwise separable convolution decomposes a single standard convolution into two convolutions: a depthwise convolution that filters each input channel independently and a pointwise convolution ( 1 × 1 {\displaystyle 1\times 1} convolution) that combines the outputs of the depthwise convolution. This factorization significantly reduces computational cost. It was first developed by Laurent Sifre during an internship at Google Brain in 2013 as an architectural variation on AlexNet to improve convergence speed and model size. === Dilated === Dilated convolution, or atrous convolution, introduces gaps between kernel elements, allowing the network to capture a larger receptive field without increasing the kernel size. === Transposed === Transposed convolution, also known as deconvolution, fractionally strided convolution, and upsampling convolution, is a convolution where the output tensor is larger than its input tensor. It's often used in encoder-decoder architectures for upsampling. It's used in image generation, semantic segmentation, and super-resolution tasks. == History == The concept of convolution in neural networks was inspired by the visual cortex in biological brains. Early work by Hubel and Wiesel in the 1960s on the cat's visual system laid the groundwork for artificial convolution networks. An early convolution neural network was developed by Kunihiko Fukushima in 1969. It had mostly hand-designed kernels inspired by convolutions in mammalian vision. In 1979 he improved it to the Neocognitron, which learns all convolutional kernels by unsupervised learning (in his terminology, "self-organized by 'learning without a teacher'"). During the 1988 to 1998 period, a series of CNN were introduced by Yann LeCun et al., ending with LeNet-5 in 1998. It was an early influential CNN architecture for handwritten digit recognition, trained on the MNIST dataset, and was used in ATM. (Olshausen & Field, 1996) discovered that simple cells in the mammalian primary visual cortex implement localized, oriented, bandpass receptive fields, which could be recreated by fitting sparse linear codes for natural scenes. This was later found to also occur in the lowest-level kernels of trained CNNs. The field saw a resurgence in the 2010s with the development of deeper architectures and the availability of large datasets and powerful GPUs. AlexNet, developed by Alex Krizhevsky et al. in 2012, was a catalytic event in modern deep learning. In that year’s ImageNet competition, the AlexNet model achieved a 16% top-five error rate, significantly outperforming the next best entry, which had a 26% error rate. The network used eight trainable layers, approximately 650,000 neurons, and around 60 million parameters, highlighting the impact of deeper architectures and GPU acceleration on image recognition performance. From the 2013 ImageNet competition, most entries adopted deep convolutional neural networks, building on the success of AlexNet. Over the following years, performance steadily improved, with the top-five error rate falling from 16% in 2012 and 12% in 2013 to below 3% by 2017, as networks grew increasingly deep.

  Read more →
• 

  Convolutional neural network

  A convolutional neural network (CNN) is a type of feedforward neural network that learns features via filter (or kernel) optimization. This type of deep learning network has been applied to process and make predictions from many different types of data including text, images and audio. CNNs are the de-facto standard in deep learning-based approaches to computer vision and image processing, and have only recently been replaced—in some cases—by newer architectures such as the transformer. Vanishing gradients and exploding gradients, seen during backpropagation in earlier neural networks, are prevented by the regularization that comes from using shared weights over fewer connections. For example, for each neuron in the fully-connected layer, 10,000 weights would be required for processing an image sized 100 × 100 pixels. However, applying cascaded convolution (or cross-correlation) kernels, only 25 weights for each convolutional layer are required to process 5x5-sized tiles. Higher-layer features are extracted from wider context windows, compared to lower-layer features. Some applications of CNNs include: image and video recognition, recommender systems, image classification, image segmentation, medical image analysis, natural language processing, brain–computer interfaces, and financial time series. CNNs are also known as shift invariant or space invariant artificial neural networks, based on the shared-weight architecture of the convolution kernels or filters that slide along input features and provide translation-equivariant responses known as feature maps. Counter-intuitively, most convolutional neural networks are not invariant to translation, due to the downsampling operation they apply to the input. Feedforward neural networks are usually fully connected networks, that is, each neuron in one layer is connected to all neurons in the next layer. The "full connectivity" of these networks makes them prone to overfitting data. Typical ways of regularization, or preventing overfitting, include: penalizing parameters during training (such as weight decay) or trimming connectivity (skipped connections, dropout, etc.) Robust datasets also increase the probability that CNNs will learn the generalized principles that characterize a given dataset rather than the biases of a poorly-populated set. Convolutional networks were inspired by biological processes in that the connectivity pattern between neurons resembles the organization of the animal visual cortex. Individual cortical neurons respond to stimuli only in a restricted region of the visual field known as the receptive field. The receptive fields of different neurons partially overlap such that they cover the entire visual field. CNNs use relatively little pre-processing compared to other image classification algorithms. This means that the network learns to optimize the filters (or kernels) through automated learning, whereas in traditional algorithms these filters are hand-engineered. This simplifies and automates the process, enhancing efficiency and scalability overcoming human-intervention bottlenecks. == Architecture == A convolutional neural network consists of an input layer, hidden layers and an output layer. In a convolutional neural network, the hidden layers include one or more layers that perform convolutions. Typically this includes a layer that performs a dot product of the convolution kernel with the layer's input matrix. This product is usually the Frobenius inner product, and its activation function is commonly ReLU. As the convolution kernel slides along the input matrix for the layer, the convolution operation generates a feature map, which in turn contributes to the input of the next layer. This is followed by other layers such as pooling layers, fully connected layers, and normalization layers. Here it should be noted how close a convolutional neural network is to a matched filter. === Convolutional layers === In a CNN, the input is a tensor with shape: (number of inputs) × (input height) × (input width) × (input channels) After passing through a convolutional layer, the image becomes abstracted to a feature map, also called an activation map, with shape: (number of inputs) × (feature map height) × (feature map width) × (feature map channels). Convolutional layers convolve the input and pass its result to the next layer. This is similar to the response of a neuron in the visual cortex to a specific stimulus. Each convolutional neuron processes data only for its receptive field. Although fully connected feedforward neural networks can be used to learn features and classify data, this architecture is generally impractical for larger inputs (e.g., high-resolution images), which would require massive numbers of neurons because each pixel is a relevant input feature. A fully connected layer for an image of size 100 × 100 has 10,000 weights for each neuron in the second layer. Convolution reduces the number of free parameters, allowing the network to be deeper. For example, using a 5 × 5 tiling region, each with the same shared weights, requires only 25 neurons. Using shared weights means there are many fewer parameters, which helps avoid the vanishing gradients and exploding gradients problems seen during backpropagation in earlier neural networks. To speed processing, standard convolutional layers can be replaced by depthwise separable convolutional layers, which are based on a depthwise convolution followed by a pointwise convolution. The depthwise convolution is a spatial convolution applied independently over each channel of the input tensor, while the pointwise convolution is a standard convolution restricted to the use of 1 × 1 {\displaystyle 1\times 1} kernels. === Pooling layers === Convolutional networks may include local and/or global pooling layers along with traditional convolutional layers. Pooling layers reduce the dimensions of data by combining the outputs of neuron clusters at one layer into a single neuron in the next layer. Local pooling combines small clusters, tiling sizes such as 2 × 2 are commonly used. Global pooling acts on all the neurons of the feature map. There are two common types of pooling in popular use: max and average. Max pooling uses the maximum value of each local cluster of neurons in the feature map, while average pooling takes the average value. === Fully connected layers === Fully connected layers connect every neuron in one layer to every neuron in another layer. It is the same as a traditional multilayer perceptron neural network (MLP). Each neuron in the fully connected layer receives input from all the neurons in the previous layer. These inputs are weighted and summed with the corresponding biases, and then passed through an activation function to perform a nonlinear transformation, generating the output. The flattened matrix goes through a fully connected layer to classify the images. === Receptive field === In neural networks, each neuron receives input from some number of locations in the previous layer. In a convolutional layer, each neuron receives input from only a restricted area of the previous layer called the neuron's receptive field. Typically the area is a square (e.g. 5 by 5 neurons). Whereas, in a fully connected layer, the receptive field is the entire previous layer. Thus, in each convolutional layer, each neuron takes input from a larger area in the input than previous layers. This is due to applying the convolution over and over, which takes the value of a pixel into account, as well as its surrounding pixels. When using dilated layers, the number of pixels in the receptive field remains constant, but the field is more sparsely populated as its dimensions grow when combining the effect of several layers. To manipulate the receptive field size as desired, there are some alternatives to the standard convolutional layer. For example, atrous or dilated convolution expands the receptive field size without increasing the number of parameters by interleaving visible and blind regions. Moreover, a single dilated convolutional layer can comprise filters with multiple dilation ratios, thus having a variable receptive field size. === Weights === Each neuron in a neural network computes an output value by applying a specific function to the input values received from the receptive field in the previous layer. The function that is applied to the input values is determined by a vector of weights and a bias (typically real numbers). Learning consists of iteratively adjusting these biases and weights. The vectors of weights and biases are called filters and represent particular features of the input (e.g., a particular shape). A distinguishing feature of CNNs is that many neurons can share the same filter. This reduces the memory footprint because a single bias and a single vector of weights are used across all receptive fields that share that filter, as opposed to each receptive field having its own bias and vector

  Read more →
• 

  Apertus (LLM)

  Apertus is a public large language model, developed by the Swiss AI Initiative (a collaboration between EPFL, ETH Zurich, and the Swiss National Supercomputing Centre). It was released on September 2, 2025, under the free and open-source Apache 2.0 license. Designed initially for business and research use cases around the world, Apertus was trained on over 1800 languages, and comes in 8 billion or 70 billion parameter versions and is available on Hugging Face for download. The model was developed aiming to adhere to European copyright law, and is one of the first examples of AI as a public good in the vein of AI Sovereignty. It is also the first large model to comply with the European Union's Artificial Intelligence Act. At its launch, the model creators emphasized multilinguality, transparency, and auditability as priorities in contrast to commercial frontier model. While international reception was largely positive, the first iteration was significantly behind the capabilities of frontier models and needs adaptation for many use cases with chatbots being a secondary but not a primary use case. As of late 2025, it was considered the largest and most capable fully open model. The capability of future models will depend in part on how much more funding can be secured.

  Read more →
• 

  XLNet

  The XLNet was an autoregressive Transformer designed as an improvement over BERT, with 340M parameters and trained on 33 billion words. It was released on 19 June 2019, under the Apache 2.0 license. It achieved state-of-the-art results on a variety of natural language processing tasks, including language modeling, question answering, and natural language inference. == Architecture == The main idea of XLNet is to model language autoregressively like the GPT models, but allow for all possible permutations of a sentence. Concretely, consider the following sentence:My dog is cute.In standard autoregressive language modeling, the model would be tasked with predicting the probability of each word, conditioned on the previous words as its context: We factorize the joint probability of a sequence of words x 1 , … , x T {\displaystyle x_{1},\ldots ,x_{T}} using the chain rule: Pr ( x 1 , … , x T ) = Pr ( x 1 ) Pr ( x 2 | x 1 ) Pr ( x 3 | x 1 , x 2 ) … Pr ( x T | x 1 , … , x T − 1 ) . {\displaystyle \Pr(x_{1},\ldots ,x_{T})=\Pr(x_{1})\Pr(x_{2}|x_{1})\Pr(x_{3}|x_{1},x_{2})\ldots \Pr(x_{T}|x_{1},\ldots ,x_{T-1}).} For example, the sentence "My dog is cute" is factorized as: Pr ( My , dog , is , cute ) = Pr ( My ) Pr ( dog | My ) Pr ( is | My , dog ) Pr ( cute | My , dog , is ) . {\displaystyle \Pr({\text{My}},{\text{dog}},{\text{is}},{\text{cute}})=\Pr({\text{My}})\Pr({\text{dog}}|{\text{My}})\Pr({\text{is}}|{\text{My}},{\text{dog}})\Pr({\text{cute}}|{\text{My}},{\text{dog}},{\text{is}}).} Schematically, we can write it as → My → My dog → My dog is → My dog is cute . {\displaystyle {\texttt {}}{\texttt {}}{\texttt {}}{\texttt {}}\to {\text{My }}{\texttt {}}{\texttt {}}{\texttt {}}\to {\text{My dog }}{\texttt {}}{\texttt {}}\to {\text{My dog is }}{\texttt {}}\to {\text{My dog is cute}}.} However, for XLNet, the model is required to predict the words in a randomly generated order. Suppose we have sampled a randomly generated order 3241, then schematically, the model is required to perform the following prediction task: → is → dog is → dog is cute → My dog is cute {\displaystyle {\texttt {}}{\texttt {}}{\texttt {}}{\texttt {}}\to {\texttt {}}{\texttt {}}{\text{is }}{\texttt {}}\to {\texttt {}}{\text{dog is }}{\texttt {}}\to {\texttt {}}{\text{dog is cute}}\to {\text{My dog is cute}}} By considering all permutations, XLNet is able to capture longer-range dependencies and better model the bidirectional context of words. === Two-Stream Self-Attention === To implement permutation language modeling, XLNet uses a two-stream self-attention mechanism. The two streams are: Content stream: This stream encodes the content of each word, as in standard causally masked self-attention. Query stream: This stream encodes the content of each word in the context of what has gone before. In more detail, it is a masked cross-attention mechanism, where the queries are from the query stream, and the key-value pairs are from the content stream. The content stream uses the causal mask M causal = [ 0 − ∞ − ∞ … − ∞ 0 0 − ∞ … − ∞ 0 0 0 … − ∞ ⋮ ⋮ ⋮ ⋱ ⋮ 0 0 0 … 0 ] {\displaystyle M_{\text{causal}}={\begin{bmatrix}0&-\infty &-\infty &\dots &-\infty \\0&0&-\infty &\dots &-\infty \\0&0&0&\dots &-\infty \\\vdots &\vdots &\vdots &\ddots &\vdots \\0&0&0&\dots &0\end{bmatrix}}} permuted by a random permutation matrix to P M causal P − 1 {\displaystyle PM_{\text{causal}}P^{-1}} . The query stream uses the cross-attention mask P ( M causal − ∞ I ) P − 1 {\displaystyle P(M_{\text{causal}}-\infty I)P^{-1}} , where the diagonal is subtracted away specifically to avoid the model "cheating" by looking at the content stream for what the current masked token is. Like the causal masking for GPT models, this two-stream masked architecture allows the model to train on all tokens in one forward pass. == Training == Two models were released: XLNet-Large, cased: 110M parameters, 24-layer, 1024-hidden, 16-heads XLNet-Base, cased: 340M parameters, 12-layer, 768-hidden, 12-heads. It was trained on a dataset that amounted to 32.89 billion tokens after tokenization with SentencePiece. The dataset was composed of BooksCorpus, and English Wikipedia, Giga5, ClueWeb 2012-B, and Common Crawl. It was trained on 512 TPU v3 chips, for 5.5 days. At the end of training, it still under-fitted the data, meaning it could have achieved lower loss with more training. It took 0.5 million steps with an Adam optimizer, linear learning rate decay, and a batch size of 8192.

  Read more →
• 

  FlowVella

  FlowVella (formerly Flowboard) is an interactive presentation platform that includes an iPad/iPhone app, a Mac app and web site for viewing presentations, built first for the iPad and web. FlowVella allows users to create, publish and share presentations through their cloud-based SaaS system. FlowVella allows embedding of text, images, PDFs, video and gallery objects in easy linkable screens, defining modern interactive presentations. FlowVella grew out of Treemo Labs. == History == FlowVella launched as 'Flowboard' on April 18, 2013 after being built for almost a year. FlowVella was incubated out of Treemo Labs, which had years of experience building native apps for iPhone, iPad and Android devices. FlowVella is an iPad app and Mac app where users create, view, publish and share interactive presentations. Presentations are viewable on flowvella.com through a web-based viewer on any device or through the FlowVella native iPad app or Mac app. On December 18, 2014, Flowboard rebranded as FlowVella after a trademark dispute. == Presentation format == FlowVella is an interactive presentation format where instead of single directional slides, presentations are made up of linkable screens with embeddable media and content objects. While 'Flows' can be exported to PDF, they all have a web address and are meant to be viewed via a web browser or the FlowVella native applications. == Revenue model == FlowVella uses the freemium model for its presentation apps. Free users can make 4 public presentations with limited number of screens/slides, but most features are available to try out the software. In 2016, FlowVella introduced a second paid plan called PRO which includes team sharing, tracking and newly introduced 'Kiosk Mode' that launched in March of 2017. == Features == FlowVella is a native iPad app and Mac app which has advantages over web based tools. All downloaded presentations can be viewed offline, without an Internet connection. This includes videos which are enabled by caching the video files into memory. For students, teachers, sales people and all users, this is extremely important because this prevents having a presentation fail because of lack of an Internet connection. Beyond the offline capabilities, there is a trend to build native applications versus HTML5 as noted by Facebook and LinkedIn both rebuilding their mobile apps as 100% native applications.

  Read more →
• 

  Residual neural network

  A residual neural network (also referred to as a residual network or ResNet) is a deep learning architecture in which the layers learn residual functions with reference to the layer inputs. It was developed in 2015 for image recognition, and won the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) of that year. As a point of terminology, "residual connection" refers to the specific architectural motif of x ↦ f ( x ) + x {\displaystyle x\mapsto f(x)+x} , where f {\displaystyle f} is an arbitrary neural network module. The motif had been used previously (see §History for details). However, the publication of ResNet made it widely popular for feedforward networks, appearing in neural networks that are seemingly unrelated to ResNet. The residual connection stabilizes the training and convergence of deep neural networks with hundreds of layers, and is a common motif in deep neural networks, such as transformer models (e.g., BERT, and GPT models such as ChatGPT), the AlphaGo Zero system, the AlphaStar system, and the AlphaFold system. == Mathematics == === Residual connection === In a multilayer neural network model, consider a (non-residual) subnetwork with a certain number of stacked layers (e.g., 2 or 3). Let H ( x ; α ) {\displaystyle H(x;\alpha )} denote the subnetwork. Suppose H ∗ {\displaystyle H^{}} is the desired optimal output of this subnetwork. Residual learning simply adds x {\displaystyle x} directly to the output, such that the optimal learned output now becomes be H ∗ − x {\displaystyle H^{}-x} , which is interpreted as a "residual" with respect to x {\displaystyle x} . The operation of "adding x {\displaystyle x} " is implemented via a "skip connection" that performs an identity mapping to connect the input of the subnetwork with its output. This connection is referred to as a "residual connection" in later work. Let F ( x ; α ) = H ( x ; a ) + x {\displaystyle F(x;\alpha )=H(x;a)+x} . The function F {\displaystyle F} is often represented by matrix multiplication interlaced with activation functions and normalization operations (e.g., batch normalization or layer normalization). As a whole, one of these subnetworks is referred to as a "residual block". A deep residual network is constructed by simply stacking these blocks. Long short-term memory (LSTM) has a memory mechanism that serves as a residual connection. In an LSTM without a forget gate, an input x t {\displaystyle x_{t}} is processed by a function F {\displaystyle F} and added to a memory cell c t {\displaystyle c_{t}} , resulting in c t + 1 = c t + F ( x t ) {\displaystyle c_{t+1}=c_{t}+F(x_{t})} . An LSTM with a forget gate essentially functions as a highway network. To stabilize the variance of the layers' inputs, it is recommended to replace the residual connections x + f ( x ) {\displaystyle x+f(x)} with x / L + f ( x ) {\displaystyle x/L+f(x)} , where L {\displaystyle L} is the total number of residual layers. === Projection connection === If the function F {\displaystyle F} is of type F : R n → R m {\displaystyle F:\mathbb {R} ^{n}\to \mathbb {R} ^{m}} where n ≠ m {\displaystyle n\neq m} , then F ( x ) + x {\displaystyle F(x)+x} is undefined. To handle this special case, a projection connection is used: y = F ( x ) + P ( x ) {\displaystyle y=F(x)+P(x)} where P {\displaystyle P} is typically a linear projection, defined by P ( x ) = M x {\displaystyle P(x)=Mx} where M {\displaystyle M} is a m × n {\displaystyle m\times n} matrix. The matrix is trained via backpropagation, as is any other parameter of the model. === Signal propagation === The introduction of identity mappings facilitates signal propagation in both forward and backward paths. ==== Forward propagation ==== If the output of the ℓ {\displaystyle \ell } -th residual block is the input to the ( ℓ + 1 ) {\displaystyle (\ell +1)} -th residual block (assuming no activation function between blocks), then the ( ℓ + 1 ) {\displaystyle (\ell +1)} -th input is: x ℓ + 1 = F ( x ℓ ) + x ℓ {\displaystyle x_{\ell +1}=F(x_{\ell })+x_{\ell }} Applying this formulation recursively, e.g.: x ℓ + 2 = F ( x ℓ + 1 ) + x ℓ + 1 = F ( x ℓ + 1 ) + F ( x ℓ ) + x ℓ {\displaystyle {\begin{aligned}x_{\ell +2}&=F(x_{\ell +1})+x_{\ell +1}\\&=F(x_{\ell +1})+F(x_{\ell })+x_{\ell }\end{aligned}}} yields the general relationship: x L = x ℓ + ∑ i = ℓ L − 1 F ( x i ) {\displaystyle x_{L}=x_{\ell }+\sum _{i=\ell }^{L-1}F(x_{i})} where L {\textstyle L} is the index of a residual block and ℓ {\textstyle \ell } is the index of some earlier block. This formulation suggests that there is always a signal that is directly sent from a shallower block ℓ {\textstyle \ell } to a deeper block L {\textstyle L} . ==== Backward propagation ==== The residual learning formulation provides the added benefit of mitigating the vanishing gradient problem to some extent. However, it is crucial to acknowledge that the vanishing gradient issue is not the root cause of the degradation problem, which is tackled through the use of normalization. To observe the effect of residual blocks on backpropagation, consider the partial derivative of a loss function E {\displaystyle {\mathcal {E}}} with respect to some residual block input x ℓ {\displaystyle x_{\ell }} . Using the equation above from forward propagation for a later residual block L > ℓ {\displaystyle L>\ell } : ∂ E ∂ x ℓ = ∂ E ∂ x L ∂ x L ∂ x ℓ = ∂ E ∂ x L ( 1 + ∂ ∂ x ℓ ∑ i = ℓ L − 1 F ( x i ) ) = ∂ E ∂ x L + ∂ E ∂ x L ∂ ∂ x ℓ ∑ i = ℓ L − 1 F ( x i ) {\displaystyle {\begin{aligned}{\frac {\partial {\mathcal {E}}}{\partial x_{\ell }}}&={\frac {\partial {\mathcal {E}}}{\partial x_{L}}}{\frac {\partial x_{L}}{\partial x_{\ell }}}\\&={\frac {\partial {\mathcal {E}}}{\partial x_{L}}}\left(1+{\frac {\partial }{\partial x_{\ell }}}\sum _{i=\ell }^{L-1}F(x_{i})\right)\\&={\frac {\partial {\mathcal {E}}}{\partial x_{L}}}+{\frac {\partial {\mathcal {E}}}{\partial x_{L}}}{\frac {\partial }{\partial x_{\ell }}}\sum _{i=\ell }^{L-1}F(x_{i})\end{aligned}}} This formulation suggests that the gradient computation of a shallower layer, ∂ E ∂ x ℓ {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{\ell }}}} , always has a later term ∂ E ∂ x L {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{L}}}} that is directly added. Even if the gradients of the F ( x i ) {\displaystyle F(x_{i})} terms are small, the total gradient ∂ E ∂ x ℓ {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{\ell }}}} resists vanishing due to the added term ∂ E ∂ x L {\textstyle {\frac {\partial {\mathcal {E}}}{\partial x_{L}}}} . == Variants of residual blocks == === Basic block === A basic block is the simplest building block studied in the original ResNet. This block consists of two sequential 3x3 convolutional layers and a residual connection. The input and output dimensions of both layers are equal. === Bottleneck block === A bottleneck block consists of three sequential convolutional layers and a residual connection. The first layer in this block is a 1×1 convolution for dimension reduction (e.g., to 1/2 of the input dimension); the second layer performs a 3×3 convolution; the last layer is another 1×1 convolution for dimension restoration. The models of ResNet-50, ResNet-101, and ResNet-152 are all based on bottleneck blocks. === Pre-activation block === The pre-activation residual block applies activation functions before applying the residual function F {\displaystyle F} . Formally, the computation of a pre-activation residual block can be written as: x ℓ + 1 = F ( ϕ ( x ℓ ) ) + x ℓ {\displaystyle x_{\ell +1}=F(\phi (x_{\ell }))+x_{\ell }} where ϕ {\displaystyle \phi } can be any activation (e.g. ReLU) or normalization (e.g. LayerNorm) operation. This design reduces the number of non-identity mappings between residual blocks, and allows an identity mapping directly from the input to the output. This design was used to train models with 200 to over 1000 layers, and was found to consistently outperform variants where the residual path is not an identity function. The pre-activation ResNet with 200 layers took 3 weeks to train for ImageNet on 8 GPUs in 2016. Since GPT-2, transformer blocks have been mostly implemented as pre-activation blocks. This is often referred to as "pre-normalization" in the literature of transformer models. == Applications == Originally, ResNet was designed for computer vision. All transformer architectures include residual connections. Indeed, very deep transformers cannot be trained without them. The original ResNet paper made no claim on being inspired by biological systems. However, later research has related ResNet to biologically-plausible algorithms. A study published in Science in 2023 disclosed the complete connectome of an insect brain (specifically that of a fruit fly larva). This study discovered "multilayer shortcuts" that resemble the skip connections in artificial neural networks, including ResNets. == History == === Previous work === Residual connections were noticed in neu

  Read more →
• 

  Nanosemantics

  Nanosemantics Lab is a Russian IT company specializing in natural language processing (NLP), computer vision (CV), speech technologies (ASR/TTS) and creation of interactive dialog interfaces, particularly chatbots and virtual assistants, based on artificial intelligence (AI). The company uses neural network platforms, including its own-made platform PuzzleLib which works on Russian-made microprocessor architecture Elbrus and Russia-based Astra Linux operating system. The company was founded in 2005 by Igor Ashmanov and Natalya Kaspersky. == Profile == The company was one of the first on Russian market to develop dialog interfaces for different branches of businesses, as well as to support community of AI developers. The company's most demanded product, as for beginning of the 2020s, is the automated "online advisers", functioning as chat bots, made for helping customers with usage of commercial products. In 2009 the company released an online service called iii.ru, where visitors were able to create their own AI-based virtual personalities entitles "infs" (for free). A visitor was able to train its own "inf" and let them chat to other "live" visitors as well with other "infs". More than 2.3 million of "infs" were created and trained by visitors over several years. Nanosemantics Lab maintains its own linguistic programming language for AI development called Dialog Language (DL). Popular social networks and instant messaging services may be used as base platforms. Nanosemantics' AI bots support different types of businesses: banks and financial services, telecommunications, retail, travel and automobile industry, home appliances production, etc. Among its solutions, Nanosemantics lists projects for various companies and institutions, among them VTB, Beeline, MTS, Sberbank, Higher School of Economics, Webmoney, Gazpromneft, Rostelecom, Ford Motors, Ministry of Health of the Russian Federation and others. The company uses the term "inf" for naming its numerous types of chat bots. The term was coined by co-founder Igor Ashmanov, head of Ashmanov & Partners. A 2014 scholarly research at Higher School of Economics, called "Basics of Business Informatics", states that such "infs", when used at business, may lower load on employees, collect statistics useful for understanding market demand and also may increase customer loyalty by providing fast and informative answers due to usage of large databases. The same research describes Nanosemantics' project for Russian branch of Ford Motors company, when AI capabilities were used for promoting the car model Ford Kuga. The research pointed out that within 2 months since beginning, the promo-website conducted 47774 talks of visitors with the specialized "inf", which indicated several hundred thousand of questions and the longest chat lasted for 3 hours 10 minutes. One-year promo campaign showed that 28.6% of people who made pre-orders talked to an "inf". In 2016 Nanosemantics launched a SaaS platform aimed at creating customized virtual assistants by users. The company's flagship product is considered to be Dialog Operating System (DialogOS), a professional corporate platform for creating intellectual voice and textual bots. It has its own linguistic programming language for creation of flexible scenarios and ready-studied neural natural language processing modules that are able to understand human interlocutors. In 2021 the company presented technology called NLab Speech ASR which contains a set of neural-networking algorithms for processing audio signals and analysis of texts that were trained and calibrated using speech-based big data marked up manually. The technology allows speed of processing of data up to "6 real-time factor" and precision values in noisy audio data may exceed 82%. In March 2022 the technology was included in Russia's Joint Registry for Russian Programs for Computers and Databases. As well, another technology was included: NLab Speech TTS, which is text-to-speech system that produces synthesized speech from printed text. == Joint projects == Nanosemantics participates in Ashmanov & Partners' projects related to AI. Since 2014, it helps in development of hardware "personal assistant" called Lexy, a solution similar to Amazon Alexa and the analogues. In August 2019 it was announced that Nanosemantics is going to participate in creation of open operating system for creating automated voice assistants. The project was called SOVA (Smart Open Virtual Assistant) and received investment of 300 million roubles (~$4,6 million) from Russian state-maintained National Technological Initiative. The company maintains long-term partnerships with Skolkovo Innovation Center (resident of IT cluster), branch association "Neuronet" and Yandex. Together with USA-based startup Remedy Logic, Nanosemantics has developed a medical diagnostic system for finding, using AI, spinal pathologies in tomography images of human bodies. Among them: central, foraminal and lateral lumbar stenosis, hernias, arthrosis. The system offers options of treatment. Since August 2021 the company is the resident of Technology Valley of Moscow State University. Also in 2021, Nanosemantics became a member of Committee on Artificial Intelligence within the Russian Association of Software Developers "Native Soft". The company states as one of its missions support of initiatives aimed at preservation and development of the Russian language. In May 2021, together with Pushkin Institute, the company created a chat bot called Phil, that explains to Russian people meaning of different Russian neologisms, and offers synonyms for them. Bot's vocabulary contains more than 500 neologisms, as well the bot can give advice on jargonisms and other types of specific words. Also in 2021, Nanosemanics Lab has signed the first-ever Russian "Codex of ethics of artificial intelligence". It establishes guidelines for ethical behavior of businesses that implement AI-based solutions. === IT contests === The company regularly organizes All-Russian Turing Test competitions for IT developers. Some of these events are co-organized with Microsoft. During the competitions, judges randomly choose virtual interlocutor and have a short conversation with them. They have to determine if a human or a machine is talking to them. An interlocutor may be either a bot or its human creator or operator. The results are measured in per cent of judges that were successfully convinced by a machine that it was a human. In 2021 Nanosemantics took part in federal project "Artificial Intelligence" by National Technological Initiative. In December 2021 the company together with state enterprise "Resource Center of Universal Design and Rehabilitation Technologies" (RCUD-RT) held an all-Russian hackathon aimed at development of AI solutions for medicine. During 3 days, participants created several training programs for patients with speech disorders. In April 2022, another hackathon by Nanosemantics was held together with MIREA – Russian Technological University. Students were participating and trying to generate algorithms for voice deepfakes. 17 teams contested in creation of software that generated artificial voice of a certain person. == Recognition == Since its foundation, Nanosemantics Lab has received a number of recognitions and awards. Among them are several professional ROTOR awards for the website iii.ru (created in 2009). The website gives the general public the means to create and train virtual assistants, which can then be used on a website or integrated into social networks. In 2013, a virtual assistant called Dana, created for Beeline Kazakhstan, was awarded with professional prize "Crystal Headset" in nomination "the best applying of technology". In 2015, the RBTH international media service included Nanosemantics in its list of "Top 50 Startups" in Russia. In 2016, the company received Russian state-maintained award called Runet Prize in two nominations: "State and Society" and "Technology and Innovation". In 2021, in Velikiy Novgorod, Nanosemantics team has won a hackathon aimed at finding means of discovering corruption schemes in Russian laws. In February 2022 the company won another contest by National Technological Initiative, called "Prochtenie", aimed at creation of AI systems for checking schoolchildren's school essays. The Nanosemantics team was awarded 20 million rubles for "overcoming technological barrier" in contest dedicated to English language, and 12 million for 1st place in special nomination "Structure" in Russian-language essay contest.

  Read more →