AI Detector And Fixer

AI Detector And Fixer — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  Deep tomographic reconstruction

  Deep Tomographic Reconstruction is a set of methods for using deep learning methods to perform tomographic reconstruction of medical and industrial images. It uses artificial intelligence and machine learning, especially deep artificial neural networks or deep learning, to overcome challenges such as measurement noise, data sparsity, image artifacts, and computational inefficiency. This approach has been applied across various imaging modalities, including CT, MRI, PET, SPECT, ultrasound, and optical imaging == Historical background == Traditional tomographic reconstruction relies on analytic methods such as filtered back-projection, or iterative methods which incrementally compute inverse transformations from measurement data (e.g., Radon or Fourier transform data). However, these approaches are not sufficient for certain imaging techniques such as low-dose CT and fast MRI, or scenarios involving metal artifacts and patient motion. == Use in imaging modalities == === Computed tomography (CT) === In CT, deep learning models can be particularly effective in reducing radiation exposure while maintaining image quality. Deep neural networks can also be able to reconstruct images of fair quality from sparsely sampled data without sacrificing diagnostic performance. Deep learning-based generative AI models can reduce CT metal artifacts. === Magnetic resonance imaging (MRI) === In magnetic resonance imaging (MRI), deep learning can lead to reduced MRI motion artifacts, and increased acquisition speed, referred to as fast MRI. Despite suffering from disadvantages such as lower signal-to-noise ratio (SNR), deep learning can enhance image quality in low field MRI, making these systems clinically viable. === Positron emission tomography (PET) and single-photon emission CT (SPECT) === For PET imaging, deep learning models can provide substantial improvements in low-dose imaging and motion artifact correction. Also, deep learning can help SPECT for generation of attenuation background. A notable technique for PET denoising involves integrating MR data through multimodal networks, which use anatomical information from MRI to enhance PET image quality. === Ultrasound imaging === Deep learning can enhance ultrasound imaging by reducing speckle noise and motion blur. For ultrasound beamforming, deep neural networks can allow superior image quality with limited data at high speed. === Optical imaging and microscopy === Diffuse optical tomography, optical coherence tomography and microscopy can be improved by deep neural networks beyond traditional methods. Furthermore, deep learning can also enhance Photoacoustic imaging (see Deep learning in photoacoustic imaging), addressing challenges like high noise, low contrast, and limited resolution. Deep learning has also been applied to label-free live-cell imaging, where convolutional neural networks predict fluorescence labels from transmitted light images, a technique known as in silico labeling. This method can enable high-throughput, non-invasive cell analysis and phenotyping without the need for traditional fluorescent dyes.

  Read more →
• 

  Three-factor learning

  In neuroscience and machine learning, three-factor learning is the combination of Hebbian plasticity with a third modulatory factor to stabilise and enhance synaptic learning. This third factor can represent various signals such as reward, punishment, error, surprise, or novelty, often implemented through neuromodulators. == Description == Three-factor learning introduces the concept of eligibility traces, which flag synapses for potential modification pending the arrival of the third factor, and helps temporal credit assignement by bridging the gap between rapid neuronal firing and slower behavioral timescales, from which learning can be done. Biological basis for Three-factor learning rules have been supported by experimental evidence. This approach addresses the instability of classical Hebbian learning by minimizing autocorrelation and maximizing cross-correlation between inputs.

  Read more →
• 

  AI Overviews

  AI Overviews is an artificial intelligence (AI) feature integrated into Google Search that produces AI-generated summaries of search results. The feature has been criticized for its inaccuracy and for reducing website traffic. == History and development == AI Overviews were first introduced as part of Google's Search Generative Experience (SGE), which was unveiled at the Google I/O conference in May 2023. In May 2024 at Google I/O 2024, the feature was rebranded as AI Overviews and launched in the United States. The introduction of AI Overviews was seen as a strategic move to compete with other generative AI advancements, including OpenAI's ChatGPT. By August 2024, AI Overviews was rolled out to several other countries, including the United Kingdom, India, Japan, Brazil, Mexico, and Indonesia, with support for multiple languages. In October 2024, Google expanded the feature globally, making it available in over 100 countries. In December 2024, Botify x Demandsphere released findings stating that when AI Overviews and featured snippets appear together on the search engine results page, they take up approximately 67.1% of the screen on desktop and 75.7% on mobile. Even if content is ranking in the #1 position, it may not be visible to consumers if other visual elements on the results page are more prominent. In March 2025, Google started testing an "AI Mode", where the search results page is AI-generated. The company was also considering adding advertisements to the AI Mode, as they already exist in AI Overviews. As of May 2025, AI Overviews are available in over 200 countries and territories and in more than 40 languages. As of March 2026, Google AI Overviews appear on more than 48% of total Google Search queries, compared to just 6.49% in the previous year (58% year-over-year growth). == Functionality == The AI Overviews feature uses large language models to generate summaries from web content. The overviews are designed to be concise, providing a snapshot of relevant information about the queried topic. Google allows users to adjust the language complexity in summaries, offering both simplified and detailed options. The overviews also include links to sources. According to a June 2025 study by Semrush, the most cited source is Quora, followed by Reddit. == Reception == The feature has faced criticism for inaccuracies, including instances where erroneous or nonsensical content was generated. Depending on what is searched for, the overview may also consist of hallucinated content, such as when searching for idioms that do not exist. In May 2024, Google temporarily restricted the AI tool after it provided suggestions that were seen as nonsensical and harmful, such as telling users to eat rocks or apply glue on pizza. Concerns were also raised by content publishers, who feared a decline in web traffic as users relied on the summaries instead of visiting source websites. A Google patent from 2026 raised the concern of webmasters that Google could entirely replace the landing page of websites by an AI optimized copy of the website in its results. There is also apprehension about the ethical implications of AI-driven content aggregation, including its impact on intellectual property rights and the visibility of smaller content providers. The European Commission announced in December 2025 that they were investigating whether AI Overviews breached European competition law. In response, Google has stated its commitment to improve content validation and refine the algorithms used to filter unreliable information. Google implemented measures to prioritize link placement within AI Overviews, aiming to balance user convenience with the needs of content creators. In January 2026, Google restricted AI Overviews on certain health-related searches following an investigation by The Guardian. == Lawsuits == On February 24, 2025, Chegg sued Alphabet over the AI Overviews feature, claiming that it was leading to students preferring "low-quality, unverified AI summaries", thus violating antitrust law. Chegg also said it was considering either a sale or a take-private transaction. In September 2025, Penske Media Corporation, the publisher of Rolling Stone and The Hollywood Reporter, sued Google, claiming that AI Overviews illegally regurgitate content from their websites and drive off potential site visitors by always appearing on top of the search results while leaving little incentive to see the linked sources. The company stated that "the future of digital media and [...] its integrity [...] is threatened by Google's current actions", alleging that 20% of searches that link to Penske-owned websites show AI Overviews and that the figure is expected to rise. Google spokesperson José Castañeda called the claims "meritless" and stated that "AI Overviews send traffic to a greater diversity of sites." In 2026, Canadian musician Ashley MacIsaac filed a lawsuit against Google claiming that the AI Overview feature had wrongly stated that MacIsaac had been convicted of numerous criminal offences and was on the sex offender registry. He claims this incorrect information led to the cancellation of a December 2025 gig organized by the Sipekne'katik First Nation.

  Read more →
• 

  Hierarchical control system

  A hierarchical control system (HCS) is a form of control system in which a set of devices and governing software is arranged in a hierarchical tree. When the links in the tree are implemented by a computer network, then that hierarchical control system is also a form of networked control system. == Overview == A human-built system with complex behavior is often organized as a hierarchy. For example, a command hierarchy has among its notable features the organizational chart of superiors, subordinates, and lines of organizational communication. Hierarchical control systems are organized similarly to divide the decision making responsibility. Each element of the hierarchy is a linked node in the tree. Commands, tasks and goals to be achieved flow down the tree from superior nodes to subordinate nodes, whereas sensations and command results flow up the tree from subordinate to superior nodes. Nodes may also exchange messages with their siblings. The two distinguishing features of a hierarchical control system are related to its layers. Each higher layer of the tree operates with a longer interval of planning and execution time than its immediately lower layer. The lower layers have local tasks, goals, and sensations, and their activities are planned and coordinated by higher layers which do not generally override their decisions. The layers form a hybrid intelligent system in which the lowest, reactive layers are sub-symbolic. The higher layers, having relaxed time constraints, are capable of reasoning from an abstract world model and performing planning. A hierarchical task network is a good fit for planning in a hierarchical control system. Besides artificial systems, an animal's control systems are proposed to be organized as a hierarchy. In perceptual control theory, which postulates that an organism's behavior is a means of controlling its perceptions, the organism's control systems are suggested to be organized in a hierarchical pattern as their perceptions are constructed so. == Control system structure == The accompanying diagram is a general hierarchical model which shows functional manufacturing levels using computerised control of an industrial control system. Referring to the diagram; Level 0 contains the field devices such as flow and temperature sensors, and final control elements, such as control valves Level 1 contains the industrialised Input/Output (I/O) modules, and their associated distributed electronic processors. Level 2 contains the supervisory computers, which collate information from processor nodes on the system, and provide the operator control screens. Level 3 is the production control level, which does not directly control the process, but is concerned with monitoring production and monitoring targets Level 4 is the production scheduling level. == Applications == === Manufacturing, robotics and vehicles === Among the robotic paradigms is the hierarchical paradigm in which a robot operates in a top-down fashion, heavy on planning, especially motion planning. Computer-aided production engineering has been a research focus at NIST since the 1980s. Its Automated Manufacturing Research Facility was used to develop a five layer production control model. In the early 1990s DARPA sponsored research to develop distributed (i.e. networked) intelligent control systems for applications such as military command and control systems. NIST built on earlier research to develop its Real-Time Control System (RCS) and Real-time Control System Software which is a generic hierarchical control system that has been used to operate a manufacturing cell, a robot crane, and an automated vehicle. In November 2007, DARPA held the Urban Challenge. The winning entry, Tartan Racing employed a hierarchical control system, with layered mission planning, motion planning, behavior generation, perception, world modelling, and mechatronics. === Artificial intelligence === Subsumption architecture is a methodology for developing artificial intelligence that is heavily associated with behavior based robotics. This architecture is a way of decomposing complicated intelligent behavior into many "simple" behavior modules, which are in turn organized into layers. Each layer implements a particular goal of the software agent (i.e. system as a whole), and higher layers are increasingly more abstract. Each layer's goal subsumes that of the underlying layers, e.g. the decision to move forward by the eat-food layer takes into account the decision of the lowest obstacle-avoidance layer. Behavior need not be planned by a superior layer, rather behaviors may be triggered by sensory inputs and so are only active under circumstances where they might be appropriate. Reinforcement learning has been used to acquire behavior in a hierarchical control system in which each node can learn to improve its behavior with experience. James Albus, while at NIST, developed a theory for intelligent system design named the Reference Model Architecture (RMA), which is a hierarchical control system inspired by RCS. Albus defines each node to contain these components. Behavior generation is responsible for executing tasks received from the superior, parent node. It also plans for, and issues tasks to, the subordinate nodes. Sensory perception is responsible for receiving sensations from the subordinate nodes, then grouping, filtering, and otherwise processing them into higher level abstractions that update the local state and which form sensations that are sent to the superior node. Value judgment is responsible for evaluating the updated situation and evaluating alternative plans. World Model is the local state that provides a model for the controlled system, controlled process, or environment at the abstraction level of the subordinate nodes. At its lowest levels, the RMA can be implemented as a subsumption architecture, in which the world model is mapped directly to the controlled process or real world, avoiding the need for a mathematical abstraction, and in which time-constrained reactive planning can be implemented as a finite-state machine. Higher levels of the RMA however, may have sophisticated mathematical world models and behavior implemented by automated planning and scheduling. Planning is required when certain behaviors cannot be triggered by current sensations, but rather by predicted or anticipated sensations, especially those that come about as result of the node's actions.

  Read more →
• 

  Feature (machine learning)

  In machine learning and pattern recognition, a feature is an individual measurable property or characteristic of a data set. Choosing informative, discriminating, and independent features is crucial to producing effective algorithms for pattern recognition, classification, and regression tasks. Features are usually numeric, but other types such as strings and graphs are used in syntactic pattern recognition, after some pre-processing step such as one-hot encoding. The concept of "features" is related to that of explanatory variables used in statistical techniques such as linear regression. == Feature types == In feature engineering, two types of features are commonly used: numerical and categorical. Numerical features are continuous values that can be measured on a scale. Examples of numerical features include age, height, weight, and income. Numerical features can be used in machine learning algorithms directly. Categorical features are discrete values that can be grouped into categories. Examples of categorical features include gender, color, and zip code. Categorical features typically need to be converted to numerical features before they can be used in machine learning algorithms. This can be done using a variety of techniques, such as one-hot encoding, label encoding, and ordinal encoding. The type of feature that is used in feature engineering depends on the specific machine learning algorithm that is being used. Some machine learning algorithms, such as decision trees, can handle both numerical and categorical features. Other machine learning algorithms, such as linear regression, can only handle numerical features. == Classification == A numeric feature can be conveniently described by a feature vector. One way to achieve binary classification is using a linear predictor function (related to the perceptron) with a feature vector as input. The method consists of calculating the scalar product between the feature vector and a vector of weights, qualifying those observations whose result exceeds a threshold. Algorithms for classification from a feature vector include nearest neighbor classification, neural networks, and statistical techniques such as Bayesian approaches. == Examples == In character recognition, features may include histograms counting the number of black pixels along horizontal and vertical directions, number of internal holes, stroke detection and many others. In speech recognition, features for recognizing phonemes can include noise ratios, length of sounds, relative power, filter matches, logarithmic Mel-scale spectral vectors and Mel-frequency cepstral coefficients, which represent the frequency characteristics of audio signals. In spam detection algorithms, features may include the presence or absence of certain email headers, the email structure, the language, the frequency of specific terms, the grammatical correctness of the text. In computer vision, there are a large number of possible features, such as edges and objects. == Feature vectors == In pattern recognition and machine learning, a feature vector is an n-dimensional vector of numerical features that represent some object. Many algorithms in machine learning require a numerical representation of objects, since such representations facilitate processing and statistical analysis. When representing images, the feature values might correspond to the pixels of an image, while when representing texts the features might be the frequencies of occurrence of textual terms. Feature vectors are equivalent to the vectors of explanatory variables used in statistical procedures such as linear regression. Feature vectors are often combined with weights using a dot product in order to construct a linear predictor function that is used to determine a score for making a prediction. The vector space associated with these vectors is often called the feature space. In order to reduce the dimensionality of the feature space, a number of dimensionality reduction techniques can be employed. Higher-level features can be obtained from already available features and added to the feature vector; for example, for the study of diseases the feature 'Age' is useful and is defined as Age = 'Year of death' minus 'Year of birth' . This process is referred to as feature construction. Feature construction is the application of a set of constructive operators to a set of existing features resulting in construction of new features. Examples of such constructive operators include checking for the equality conditions {=, ≠}, the arithmetic operators {+,−,×, /}, the array operators {max(S), min(S), average(S)} as well as other more sophisticated operators, for example count(S, C) that counts the number of features in the feature vector S satisfying some condition C or, for example, distances to other recognition classes generalized by some accepting device. Feature construction has long been considered a powerful tool for increasing both accuracy and understanding of structure, particularly in high-dimensional problems. Applications include studies of disease and emotion recognition from speech. == Selection and extraction == The initial set of raw features can be redundant and large enough that estimation and optimization is made difficult or ineffective. Therefore, a preliminary step in many applications of machine learning and pattern recognition consists of selecting a subset of features, or constructing a new and reduced set of features to facilitate learning, and to improve generalization and interpretability. Extracting or selecting features is a combination of art and science; developing systems to do so is known as feature engineering. It requires the experimentation of multiple possibilities and the combination of automated techniques with the intuition and knowledge of the domain expert. Automating this process is feature learning, where a machine not only uses features for learning, but learns the features itself.

  Read more →
• 

  Hierarchical navigable small world

  Hierarchical navigable small world (HNSW) is an algorithm for approximate nearest neighbor search. It is used to find items that are similar to a query item in a large collection, without comparing the query with every item one by one. The algorithm is commonly used for searching vector data. In these systems, an item such as a document, image, song, or user profile is represented by a list of numbers called a vector. Items with similar vectors are treated as similar according to the model that produced the vectors. HNSW provides a way to search these vectors quickly, especially in large datasets. HNSW stores vectors in a graph. Each vector is a node, and links connect it to some nearby vectors. The graph has several layers: upper layers contain fewer nodes and act like a rough map, while the bottom layer contains all nodes and gives a more detailed view. A search starts in an upper layer, follows links toward nodes that are closer to the query, and then repeats the process in lower layers until it finds a set of likely nearest neighbors. == Background == The nearest neighbor search problem asks which items in a dataset are closest to a query item. A direct search can compare the query with every item in the dataset, but this becomes slow when the dataset is large. Exact search methods based on spatial trees, such as the k-d tree and R-tree, can also become less effective for high-dimensional data, a problem often associated with the curse of dimensionality. Approximate nearest neighbor methods trade some exactness for speed or lower resource use. Instead of always guaranteeing the exact closest item, they try to return close items quickly. Other approximate methods include locality-sensitive hashing and product quantization. HNSW builds on research into small-world networks and navigable graphs. In a small-world graph, most nodes can be reached from other nodes through a short chain of links. In a navigable graph, a search procedure can use local information to move toward a target. Jon Kleinberg's work on navigation in small-world networks is an important example of this research area. Later work studied ways to add links that make graphs easier to navigate greedily. The HNSW algorithm extends earlier navigable small world methods for similarity search by adding a hierarchy of graph layers. This hierarchy helps the algorithm find a good region of the graph before doing a more detailed search in the bottom layer. == Algorithm == HNSW is based on a proximity graph. In this graph, nearby vectors are connected by edges. The algorithm uses these edges to move through the dataset, rather than scanning every vector. The graph is hierarchical. Every vector appears in the bottom layer. Some vectors are also placed in higher layers, with fewer vectors appearing as the layers go upward. The upper layers allow long-range movement across the dataset, while the lower layers allow a more detailed search near promising candidates. A typical search proceeds as follows: The search begins from an entry point in the highest layer. At each step, the algorithm looks at neighboring nodes and moves to a neighbor that is closer to the query. When it cannot find a closer neighbor in that layer, it moves down to the next layer. In the bottom layer, it explores a wider set of candidate nodes and returns the nearest candidates found. This search strategy is often described as greedy navigation. The algorithm repeatedly chooses locally better nodes, using the graph structure to approach the query point. == Construction and parameters == The HNSW graph is built incrementally. When a new vector is inserted, the algorithm assigns it a maximum layer, searches for nearby existing nodes, and connects the new node to selected neighbors in each layer where it appears. Implementations usually expose parameters that control the trade-off between speed, accuracy, memory use, and construction time. A higher number of graph connections can improve recall but requires more memory. A larger search candidate list can improve accuracy but makes queries slower. A larger construction candidate list can improve the quality of the graph but makes index building slower. Because HNSW is approximate, its results are not always identical to a full exact search. Its practical performance depends on the dataset, distance measure, implementation, and parameter settings. Benchmarking studies have found HNSW-based libraries to be strong performers among approximate nearest neighbor methods, although worst-case performance can differ from performance on common benchmark datasets. == Use in vector search systems == HNSW is used as an index in systems that store and search high-dimensional vectors. These systems include vector databases, search engines, and database extensions. Typical uses include semantic search, recommender systems, image similarity search, and retrieval-augmented generation. Several software projects implement or support HNSW. Libraries include hnswlib, which is associated with the original HNSW authors, and FAISS. Database and search systems that document HNSW support include Apache Lucene, Chroma, ClickHouse, DuckDB, MariaDB, Milvus, pgvector, Qdrant, and Redis.

  Read more →
• 

  Rabbit r1

  The Rabbit r1 is an artificial intelligence personal assistant device developed by the American technology startup Rabbit Inc and co-designed by Teenage Engineering. It was announced at the 2024 Consumer Electronics Show as a handheld device intended to perform digital tasks through voice commands, touch interaction, and web-based AI agents. The r1 was marketed around Rabbit's concept of a "large action model" (LAM), which the company described as software able to operate websites and services on behalf of users. The device runs rabbitOS, an operating system based on the Android Open Source Project. Its services have included AI search, image recognition, voice interaction, music playback, rideshare and food-ordering integrations, and later experimental web-agent features such as LAM Playground and teach mode. Initial reviews were largely negative, with reviewers criticizing the device's limited functionality, bugs, and unclear advantages over a smartphone. Critics also questioned Rabbit's claims after the r1 software was shown to run on an Android phone. Rabbit continued to issue software updates after launch, including rabbitOS 2 in September 2025, which introduced a redesigned card-based interface, gesture navigation, and a "creations" feature for generating small software tools and experiences on the device. Rabbit Inc was founded by Jesse Lyu Cheng. == Hardware == Display: A 2.88-inch touchscreen for interactive user input. Input: push-to-talk button to activate voice commands; scroll wheel; Gyroscope; Magnetometer; Accelerometer; GPS. Camera: 8 MP single camera, with a resolution of 3264x2448, allowing for the connected external AI to use computer vision. Audio: Equipped with a speaker and dual microphones for audio interaction. Connectivity: Supports Wi-Fi and cellular connections via a SIM card slot to access internet services. Processor: Runs on a 2.3GHz MediaTek Helio P35 processor. Memory: Contains 4GB of RAM for operational tasks. Storage: Offers 128GB of internal storage for data. Ports: Utilizes a USB-C port for charging and data connections. == Software == The Rabbit r1 runs rabbitOS, which is based on the Android Open Source Project (AOSP), specifically Android 13. Rabbit founder Jesse Lyu described rabbitOS as a "very bespoke AOSP" after reports that the r1's software could be run on a conventional Android phone. Rabbit described the r1 as using a large action model (LAM), a type of AI agent intended to perform tasks across software interfaces rather than only answer questions. At launch, the device supported a limited set of services, including AI search, vision features, music playback, and some third-party integrations. Perplexity.ai was one of the AI services used to answer user queries. In 2024, Rabbit released several software updates that added features and attempted to address early criticism of the device. In July 2024, the company launched "beta rabbit", an advanced search and conversation mode for more complex queries. In October 2024, it released LAM Playground, a web-based agent feature intended to let the r1 operate websites on behalf of users. Reviewers found the feature experimental; Android Authority reported that it could perform some navigation tasks but struggled with CAPTCHAs, loops, and unintended behavior. In November 2024, Rabbit introduced a beta "teach mode", which allowed users to demonstrate web-based tasks in the Rabbithole web portal and later ask the r1 to repeat them. The company described teach mode as experimental, and The Verge noted that Rabbit warned users that results could be unpredictable and that CAPTCHA-protected sites could cause problems. Rabbit released rabbitOS 2 in September 2025. The update redesigned the interface around a card-based layout, added additional touchscreen gestures, and introduced "creations", a feature that lets users generate simple software tools, games, and interfaces through natural-language prompts. Coverage of the update described it as a major software overhaul rather than new hardware. == Reception == === Funding === Rabbit raised $20 million in funding from Khosla Ventures, Synergis Capital and Kakao Investment in October 2023. The company announced an additional $10 million in funding in December 2023. === Sales === Following its announcement at the 2024 Consumer Electronics Show, 130,000 units were sold. On August 13, 2024, Rabbit announced that sales of r1 had expanded to the entire European Union (except Malta) and United Kingdom. On August 21, 2024, sales of r1 expanded to Singapore. === Reviews === The r1 was met with strong criticism immediately after Rabbit began shipping the device. Some reviews questioned what the device was able to do that a smartphone could not, while comparing it to the similar Humane Ai Pin. YouTuber Marques Brownlee called the device "barely reviewable". Android Authority's Mishaal Rahman managed to install Rabbit r1's software on a Pixel 6a smartphone, after a tipster shared an APK file. The Verge echoed the claims made by Rahman. In response, Lyu published statements confirming its use of Android, but denying that the r1 is an Android app. Mashable called its Vision features impressive, but said that "these praise-worthy features are overshadowed by buggy performance". Ars Technica wrote a blog post claiming "the company is blocking access from bootleg APKs". TechCrunch gave a slightly more positive review, calling the device a "fun peep at a possible future", but could not "advise anyone to buy one now." Shortly after the launch of r1, Rabbit began a weekly cadence of software updates to address much of the criticism from the early reviews, including "battery and GPS performance, time zone selection, and more". Digital Trends said the Magic Camera feature "takes the most mundane, ordinary, and badly composed photos and makes something fun and eye-catching from them." Mashable said the "beta rabbit" feature "makes Rabbit R1 more conversational and intelligent". Later coverage noted that Rabbit continued to update the r1 after its poorly received launch. The Verge reported in September 2024 that about 5,000 of roughly 100,000 purchasers were using the device at any given moment, citing Lyu, and described the product as having launched before it was ready. In 2025, coverage of rabbitOS 2 described the update as an attempt to reset the device's software experience after the criticism of its original release. == Controversies == === GAMA project === Rabbit Inc has garnered attention due to allegations surrounding its funding and the company's past projects. The company came under scrutiny when Stephen Findeisen, known as Coffeezilla on YouTube, published a video in May 2024, alleging that Rabbit Incorporation was "built on a scam". Rabbit Incorporation, initially named Cyber Manufacturing Co, rebranded just two months before launching the Rabbit R1. The company, under its former name, raised $6 million in November 2021 for a project called GAMA, described as a "Next Generation NFT Project." Jesse Lyu, the CEO of Rabbit Incorporation, referred to GAMA as a "fun little project." Coffeezilla, who investigates influencer scams, highlighted old Clubhouse recordings of Jesse Lyu discussing the GAMA project. In these recordings, Lyu emphasized the substantial funding behind GAMA and its potential to be a revolutionary, carbon-negative cryptocurrency. Coffeezilla questioned the whereabouts of the funds raised for GAMA, estimating that approximately $1 million in refunds to investors remained unresolved. He suggested that the rebranding to Rabbit Incorporation and the shift to developing the Rabbit R1 were attempts to divert from the GAMA project's issues. In response to Coffeezilla's inquiries, Rabbit Incorporation stated that the $6 million raised was used for the GAMA project. The company said that NFTs cannot be refunded unless the owner agrees to "burn" them on the blockchain. Rabbit Incorporation also said that the GAMA project was open-sourced and returned to the community, aligning with community feedback. They also mentioned that efforts to buy back NFTs were made to counteract malicious trading and maintain market stability. === Security === In June 2024, Engadget reported that the Rabbitude team, a community reverse engineering project, had gained access to the r1's codebase revealing that r1's software contained several hardcoded API keys in its code for ElevenLabs, Microsoft Azure, Yelp, and Google Maps, potentially allowing unauthorized access to r1 responses, including those containing the users' personal information. For a short time, Rabbit immediately began revoking and rotating those secrets and confirmed that the code was leaked by an employee who had "been terminated and remains under investigation". In July 2024, the company revealed that all user chats and device pairing data were logged on the r1 with no ability to delete them. This meant that lost or stolen devices could be used to extract user

  Read more →
• 

  Empirical dynamic modeling

  Empirical dynamic modeling (EDM) is a framework for analysis and prediction of nonlinear dynamical systems. Applications include population dynamics, ecosystem service, medicine, neuroscience, dynamical systems, geophysics, and human-computer interaction. EDM was originally developed by Robert May and George Sugihara. It can be considered a methodology for data modeling, predictive analytics, dynamical system analysis, machine learning and time series analysis. == Description == Mathematical models have tremendous power to describe observations of real-world systems. They are routinely used to test hypothesis, explain mechanisms and predict future outcomes. However, real-world systems are often nonlinear and multidimensional, in some instances rendering explicit equation-based modeling problematic. Empirical models, which infer patterns and associations from the data instead of using hypothesized equations, represent a natural and flexible framework for modeling complex dynamics. Donald DeAngelis and Simeon Yurek illustrated that canonical statistical models are ill-posed when applied to nonlinear dynamical systems. A hallmark of nonlinear dynamics is state-dependence: system states are related to previous states governing transition from one state to another. EDM operates in this space, the multidimensional state-space of system dynamics rather than on one-dimensional observational time series. EDM does not presume relationships among states, for example, a functional dependence, but projects future states from localised, neighboring states. EDM is thus a state-space, nearest-neighbors paradigm where system dynamics are inferred from states derived from observational time series. This provides a model-free representation of the system naturally encompassing nonlinear dynamics. A cornerstone of EDM is recognition that time series observed from a dynamical system can be transformed into higher-dimensional state-spaces by time-delay embedding with Takens's theorem. The state-space models are evaluated based on in-sample fidelity to observations, conventionally with Pearson correlation between predictions and observations. == Methods == Primary EDM algorithms include Simplex projection, Sequential locally weighted global linear maps (S-Map) projection, Multivariate embedding in Simplex or S-Map, Convergent cross mapping (CCM), and Multiview Embeding, described below. Nearest neighbors are found according to: NN ( y , X , k ) = ‖ X N i E − y ‖ ≤ ‖ X N j E − y ‖ if 1 ≤ i ≤ j ≤ k {\displaystyle {\text{NN}}(y,X,k)=\|X_{N_{i}}^{E}-y\|\leq \|X_{N_{j}}^{E}-y\|{\text{ if }}1\leq i\leq j\leq k} === Simplex === Simplex projection is a nearest neighbor projection. It locates the k {\displaystyle k} nearest neighbors to the location in the state-space from which a prediction is desired. To minimize the number of free parameters k {\displaystyle k} is typically set to E + 1 {\displaystyle E+1} defining an E + 1 {\displaystyle E+1} dimensional simplex in the state-space. The prediction is computed as the average of the weighted phase-space simplex projected T p {\displaystyle Tp} points ahead. Each neighbor is weighted proportional to their distance to the projection origin vector in the state-space. Find k {\displaystyle k} nearest neighbor: N k ← NN ( y , X , k ) {\displaystyle N_{k}\gets {\text{NN}}(y,X,k)} Define the distance scale: d ← ‖ X N 1 E − y ‖ {\displaystyle d\gets \|X_{N_{1}}^{E}-y\|} Compute weights: For{ i = 1 , … , k {\displaystyle i=1,\dots ,k} } : w i ← exp ⁡ ( − ‖ X N i E − y ‖ / d ) {\displaystyle w_{i}\gets \exp(-\|X_{N_{i}}^{E}-y\|/d)} Average of state-space simplex: y ^ ← ∑ i = 1 k ( w i X N i + T p ) / ∑ i = 1 k w i {\displaystyle {\hat {y}}\gets \sum _{i=1}^{k}\left(w_{i}X_{N_{i}+T_{p}}\right)/\sum _{i=1}^{k}w_{i}} === S-Map === S-Map extends the state-space prediction in Simplex from an average of the E + 1 {\displaystyle E+1} nearest neighbors to a linear regression fit to all neighbors, but localised with an exponential decay kernel. The exponential localisation function is F ( θ ) = exp ( − θ d / D ) {\displaystyle F(\theta )={\text{exp}}(-\theta d/D)} , where d {\displaystyle d} is the neighbor distance and D {\displaystyle D} the mean distance. In this way, depending on the value of θ {\displaystyle \theta } , neighbors close to the prediction origin point have a higher weight than those further from it, such that a local linear approximation to the nonlinear system is reasonable. This localisation ability allows one to identify an optimal local scale, in-effect quantifying the degree of state dependence, and hence nonlinearity of the system. Another feature of S-Map is that for a properly fit model, the regression coefficients between variables have been shown to approximate the gradient (directional derivative) of variables along the manifold. These Jacobians represent the time-varying interaction strengths between system variables. Find k {\displaystyle k} nearest neighbor: N ← NN ( y , X , k ) {\displaystyle N\gets {\text{NN}}(y,X,k)} Sum of distances: D ← 1 k ∑ i = 1 k ‖ X N i E − y ‖ {\displaystyle D\gets {\frac {1}{k}}\sum _{i=1}^{k}\|X_{N_{i}}^{E}-y\|} Compute weights: For{ i = 1 , … , k {\displaystyle i=1,\dots ,k} } : w i ← exp ⁡ ( − θ ‖ X N i E − y ‖ / D ) {\displaystyle w_{i}\gets \exp(-\theta \|X_{N_{i}}^{E}-y\|/D)} Reweighting matrix: W ← diag ( w i ) {\displaystyle W\gets {\text{diag}}(w_{i})} Design matrix: A ← [ 1 X N 1 X N 1 − 1 … X N 1 − E + 1 1 X N 2 X N 2 − 1 … X N 2 − E + 1 ⋮ ⋮ ⋮ ⋱ ⋮ 1 X N k X N k − 1 … X N k − E + 1 ] {\displaystyle A\gets {\begin{bmatrix}1&X_{N_{1}}&X_{N_{1}-1}&\dots &X_{N_{1}-E+1}\\1&X_{N_{2}}&X_{N_{2}-1}&\dots &X_{N_{2}-E+1}\\\vdots &\vdots &\vdots &\ddots &\vdots \\1&X_{N_{k}}&X_{N_{k}-1}&\dots &X_{N_{k}-E+1}\end{bmatrix}}} Weighted design matrix: A ← W A {\displaystyle A\gets WA} Response vector at T p {\displaystyle Tp} : b ← [ X N 1 + T p X N 2 + T p ⋮ X N k + T p ] {\displaystyle b\gets {\begin{bmatrix}X_{N_{1}+T_{p}}\\X_{N_{2}+T_{p}}\\\vdots \\X_{N_{k}+T_{p}}\end{bmatrix}}} Weighted response vector: b ← W b {\displaystyle b\gets Wb} Least squares solution (SVD): c ^ ← argmin c ‖ A c − b ‖ 2 2 {\displaystyle {\hat {c}}\gets {\text{argmin}}_{c}\|Ac-b\|_{2}^{2}} Local linear model c ^ {\displaystyle {\hat {c}}} is prediction: y ^ ← c ^ 0 + ∑ i = 1 E c ^ i y i {\displaystyle {\hat {y}}\gets {\hat {c}}_{0}+\sum _{i=1}^{E}{\hat {c}}_{i}y_{i}} === Multivariate Embedding === Multivariate Embedding recognizes that time-delay embeddings are not the only valid state-space construction. In Simplex and S-Map one can generate a state-space from observational vectors, or time-delay embeddings of a single observational time series, or both. === Convergent Cross Mapping === Convergent cross mapping (CCM) leverages a corollary to the Generalized Takens Theorem that it should be possible to cross predict or cross map between variables observed from the same system. Suppose that in some dynamical system involving variables X {\displaystyle X} and Y {\displaystyle Y} , X {\displaystyle X} causes Y {\displaystyle Y} . Since X {\displaystyle X} and Y {\displaystyle Y} belong to the same dynamical system, their reconstructions (via embeddings) M x {\displaystyle M_{x}} , and M y {\displaystyle M_{y}} , also map to the same system. The causal variable X {\displaystyle X} leaves a signature on the affected variable Y {\displaystyle Y} , and consequently, the reconstructed states based on Y {\displaystyle Y} can be used to cross predict values of X {\displaystyle X} . CCM leverages this property to infer causality by predicting X {\displaystyle X} using the M y {\displaystyle M_{y}} library of points (or vice versa for the other direction of causality), while assessing improvements in cross map predictability as larger and larger random samplings of M y {\displaystyle M_{y}} are used. If the prediction skill of X {\displaystyle X} increases and saturates as the entire M y {\displaystyle M_{y}} is used, this provides evidence that X {\displaystyle X} is casually influencing Y {\displaystyle Y} . === Multiview Embedding === Multiview Embedding is a Dimensionality reduction technique where a large number of state-space time series vectors are combitorially assessed towards maximal model predictability. == Extensions == Extensions to EDM techniques include: Generalized Theorems for Nonlinear State Space Reconstruction Extended Convergent Cross Mapping Dynamic stability S-Map regularization Visual analytics with EDM Convergent Cross Sorting Expert system with EDM hybrid Sliding windows based on the extended convergent cross-mapping Empirical Mode Modeling Accounting for missing data and variable step sizes Accounting for observation noise Hierarchical Bayesian EDM via Gaussian processes Intelligent and Adaptive Control Optimal control via Empirical dynamic programming Multiview distance regularised S-map

  Read more →
• 

  List of chatbots

  A chatbot is a software application or web interface that is designed to mimic human conversation through text or voice interactions. Modern chatbots are typically online and use generative artificial intelligence systems that are capable of maintaining a conversation with a user in natural language and simulating the way a human would behave as a conversational partner. Such chatbots often use large language models (LLMs) and natural language processing, but simpler chatbots have existed for decades. == LLM chatbots == == General chatbots == == Historical chatbots ==

  Read more →
• 

  Semantic analysis (machine learning)

  In machine learning, semantic analysis of a text corpus is the task of building structures that approximate concepts from a large set of documents. It generally does not involve prior semantic understanding of the documents. Semantic analysis strategies include: Metalanguages based on first-order logic, which can analyze the speech of humans. Understanding the semantics of a text is symbol grounding: if language is grounded, it is equal to recognizing a machine-readable meaning. For the restricted domain of spatial analysis, a computer-based language understanding system was demonstrated. Latent semantic analysis (LSA), a class of techniques where documents are represented as vectors in a term space. A prominent example is probabilistic latent semantic analysis (PLSA). Latent Dirichlet allocation, which involves attributing document terms to topics. n-grams and hidden Markov models, which work by representing the term stream as a Markov chain, in which each term is derived from preceding terms. == Stochastic semantic analysis ==

  Read more →
• 

  Hardware for artificial intelligence

  Specialized computer hardware is often used to execute artificial intelligence (AI) programs faster, and with less energy, such as Lisp machines, neuromorphic engineering, event cameras, and physical neural networks. Since 2017, several consumer grade CPUs and SoCs have on-die NPUs. As of 2023, the market for AI hardware is dominated by GPUs. As of the 2020s, AI computation is dominated by graphics processing units (GPUs) and newer domain-specific accelerators such as Google's Tensor Processing Units (TPUs), AMD's Instinct MI300 series, and various on-device neural-processing units (NPUs) found in consumer hardware. == Scope == For the purposes of this article, AI hardware refers to computing components and systems specifically designed or optimized to accelerate artificial-intelligence workloads such as machine-learning training or inference. This includes general-purpose accelerators used for AI (for example, GPUs) and domain-specific accelerators (for example, TPUs, NPUs, and other AI ASICs). Event-based cameras are sometimes discussed in the context of neuromorphic computing, but they are input sensors rather than AI compute devices. Conversely, components such as memristors are basic circuit elements rather than specialized AI hardware when considered alone. == Lisp machines == Lisp machines were developed in the late 1970s and early 1980s to make artificial intelligence programs written in the programming language Lisp run faster. == Dataflow architecture == Dataflow architecture processors used for AI serve various purposes with varied implementations like the polymorphic dataflow Convolution Engine by Kinara (formerly Deep Vision), structure-driven dataflow by Hailo, and dataflow scheduling by Cerebras. == Component hardware == === AI accelerators === Since the 2010s, advances in computer hardware have led to more efficient methods for training deep neural networks that contain many layers of non-linear hidden units and a very large output layer. By 2019, graphics processing units (GPUs), often with AI-specific enhancements, had displaced central processing units (CPUs) as the dominant means to train large-scale commercial cloud AI. OpenAI estimated the hardware compute used in the largest deep learning projects from Alex Net (2012) to Alpha Zero (2017), and found a 300,000-fold increase in the amount of compute needed, with a doubling-time trend of 3.4 months. === General-purpose GPUs for AI === Since the 2010s, graphics processing units (GPUs) have been widely used to train and deploy deep learning models because of their highly parallel architecture and high memory bandwidth. Modern data-center GPUs include dedicated tensor or matrix-math units that accelerate neural-network operations. In 2022, NVIDIA introduced the Hopper-generation H100 GPU, adding FP8 precision support and faster interconnects for large-scale model training. AMD and other vendors have also developed GPUs and accelerators aimed at AI and high-performance computing workloads. === Domain-specific accelerators (ASICs / NPUs) === Beyond general-purpose GPUs, several companies have developed application-specific integrated circuits (ASICs) and neural processing units (NPUs) tailored for AI workloads. Google introduced the Tensor Processing Unit (TPU) in 2016 for deep-learning inference, with later generations supporting large-scale training through dense systolic-array designs and optical interconnects. Other vendors have released similar devices—such as Apple's Neural Engine and various on-device NPUs—that emphasize energy-efficient inference in mobile or edge computing environments. === Memory and interconnects === AI accelerators rely on fast memory and inter-chip links to manage the large data volumes of training and inference. High-bandwidth memory (HBM) stacks, standardized as HBM3 in 2022, provide terabytes-per-second throughput on modern GPUs and ASICs. These accelerators are often connected through dedicated fabrics such as NVIDIA's NVLink and NVSwitch or optical interconnects used in TPU systems to scale performance across thousands of chips.

  Read more →
• 

  United States Tech Force

  The U.S. Tech Force (also styled as US Tech Force, Tech Force, or Government Tech Force) is a federal hiring initiative launched by the second Donald Trump administration in December 2025. The program, administered by the Office of Personnel Management (OPM), aims to recruit about 1,000 early-career technology professionals into two-year government jobs to modernize federal IT systems, advance artificial intelligence (AI) capabilities, and address technological gaps in government operations. The initiative is an effort to plug capability gaps created by Trump-administration efforts to shrink the federal government, which led to the departure of some 220,000 federal employees, including many in IT. The initiative seeks early-career workers; officials said it would offer competitive salaries and opportunities to work on high-impact government technology projects. Major technology companies—including Amazon, Apple, Microsoft, Nvidia, Meta, Google, and OpenAI—agreed to help identify and refer candidates. Candidates are allowed to take Tech Force positions on leaves of absence and without divesting their stock, raising conflict-of-interest questions. In January 2026, OPM direction Scott Kupor said the deadline for applying to Tech Force was being extended because of "tremendous interest" without saying how many people had actually applied. Also in December 2025, news broke that the administration is planning another novel use of private-sector workers: hiring cybersecurity firms for offensive cyber operations.

  Read more →
• 

  Apache Kudu

  Apache Kudu is a free and open source column-oriented data store of the Apache Hadoop ecosystem. It is compatible with most of the data processing frameworks in the Hadoop environment. It provides completeness to Hadoop's storage layer to enable fast analytics on fast data. The open source project to build Apache Kudu began as internal project at Cloudera. The first version Apache Kudu 1.0 was released 19 September 2016. == Comparison with other storage engines == Kudu was designed and optimized for OLAP workloads. Like HBase, it is a real-time store that supports key-indexed record lookup and mutation. Kudu differs from HBase since Kudu's datamodel is a more traditional relational model, while HBase is schemaless. Kudu's "on-disk representation is truly columnar and follows an entirely different storage design than HBase/Bigtable".

  Read more →
• 

  Adversarial machine learning

  Adversarial machine learning is the study of the attacks on machine learning algorithms, and of the defenses against such attacks. Machine learning techniques are mostly designed to work on specific problem sets, under the assumption that the training and test data are generated from the same statistical distribution (IID). However, this assumption is often violated in practical high-stake applications, where users may intentionally supply fabricated data that violates the statistical assumption. Most common attacks in adversarial machine learning include evasion attacks, data poisoning attacks, Byzantine attacks and model extraction. == History == At the MIT Spam Conference in January 2004, John Graham-Cumming showed that a machine-learning spam filter could be used to defeat another machine-learning spam filter by automatically learning which words to add to a spam email to get the email classified as not spam. In 2004, Nilesh Dalvi and others noted that linear classifiers used in spam filters could be defeated by simple "evasion attacks" as spammers inserted "good words" into their spam emails. (Around 2007, some spammers added random noise to fuzz words within "image spam" in order to defeat OCR-based filters.) In 2006, Marco Barreno and others published "Can Machine Learning Be Secure?", outlining a broad taxonomy of attacks. As late as 2013 many researchers continued to hope that non-linear classifiers (such as support vector machines and neural networks) might be robust to adversaries, until Battista Biggio and others demonstrated the first gradient-based attacks on such machine-learning models (2012–2013). In 2012, deep neural networks began to dominate computer vision problems; starting in 2014, Christian Szegedy and others demonstrated that deep neural networks could be fooled by adversaries, again using a gradient-based attack to craft adversarial perturbations. Further work would show that adversarial attacks are harder to produce in uncontrolled environments, due to the different environmental constraints that cancel out the effect of noise. For example, any small rotation or slight illumination on an adversarial image can destroy the adversariality. In addition, researchers such as Google Brain's Nick Frosst point out that it is much easier to make self-driving cars miss stop signs by physically removing the sign itself, rather than creating adversarial examples. Frosst also believes that the adversarial machine learning community incorrectly assumes models trained on a certain data distribution will also perform well on a completely different data distribution. He suggests that a new approach to machine learning should be explored, and is currently working on a unique neural network that has characteristics more similar to human perception than state-of-the-art approaches. While adversarial machine learning continues to be heavily rooted in academia, large tech companies such as Google, Microsoft, and IBM have begun curating documentation and open source code bases to allow others to concretely assess the robustness of machine learning models and minimize the risk of adversarial attacks. === Examples === Examples include attacks in spam filtering, where spam messages are obfuscated through the misspelling of "bad" words or the insertion of "good" words; attacks in computer security, such as obfuscating malware code within network packets or modifying the characteristics of a network flow to mislead intrusion detection; attacks in biometric recognition where fake biometric traits may be exploited to impersonate a legitimate user; or to compromise users' template galleries that adapt to updated traits over time. Researchers showed that by changing only one-pixel it was possible to fool deep learning algorithms. Others 3-D printed a toy turtle with a texture engineered to make Google's object detection AI classify it as a rifle regardless of the angle from which the turtle was viewed. Creating the turtle required only low-cost commercially available 3-D printing technology. A machine-tweaked image of a dog was shown to look like a cat to both computers and humans. A 2019 study reported that humans can guess how machines will classify adversarial images. Researchers discovered methods for perturbing the appearance of a stop sign such that an autonomous vehicle classified it as a merge or speed limit sign. A data poisoning filter called Nightshade was released in 2023 by researchers at the University of Chicago. It was created for use by visual artists to put on their artwork to corrupt the data set of text-to-image models, which usually scrape their data from the internet without the consent of the image creator. McAfee attacked Tesla's former Mobileye system, fooling it into driving 50 mph over the speed limit, simply by adding a two-inch strip of black tape to a speed limit sign. Adversarial patterns on glasses or clothing designed to deceive facial-recognition systems or license-plate readers, have led to a niche industry of "stealth streetwear". An adversarial attack on a neural network can allow an attacker to inject algorithms into the target system. Researchers can also create adversarial audio inputs to disguise commands to intelligent assistants in benign-seeming audio; a parallel literature explores human perception of such stimuli. Clustering algorithms are used in security applications. Malware and computer virus analysis aims to identify malware families, and to generate specific detection signatures. In the context of malware detection, researchers have proposed methods for adversarial malware generation that automatically craft binaries to evade learning-based detectors while preserving malicious functionality. Optimization-based attacks such as GAMMA use genetic algorithms to inject benign content (for example, padding or new PE sections) into Windows executables, framing evasion as a constrained optimization problem that balances misclassification success with the size of the injected payload and showing transferability to commercial antivirus products. Complementary work uses generative adversarial networks (GANs) to learn feature-space perturbations that cause malware to be classified as benign; Mal-LSGAN, for instance, replaces the standard GAN loss with a least-squares objective and modified activation functions to improve training stability and produce adversarial malware examples that substantially reduce true positive rates across multiple detectors. == Challenges in applying machine learning to security == Researchers have observed that the constraints under which machine-learning techniques function in the security domain are different from those of common benchmark domains. Security data may change over time, include mislabeled samples, or reflect adversarial behavior, which complicates evaluation and reproducibility. === Data collection issues === Security datasets vary across formats, including binaries, network traces, and log files. Studies have reported that the process of converting these sources into features can introduce bias or inconsistencies. In addition, time-based leakage can occur when related malware samples are not properly separated across training and testing splits, which may lead to overly optimistic results. === Labeling and ground truth challenges === Malware labels are often unstable because different antivirus engines may classify the same sample in conflicting ways. Ceschin et al. note that families may be renamed or reorganized over time, causing further discrepancies in ground truth and reducing the reliability of benchmarks. === Concept drift === Because malware creators continuously adapt their techniques, the statistical properties of malicious samples also change. This form of concept drift has been widely documented and may reduce model performance unless systems are updated regularly or incorporate mechanisms for incremental learning. === Feature robustness === Researchers differentiate between features that can be easily manipulated and those that are more resistant to modification. For example, simple static attributes, such as header fields, may be altered by attackers, while structural features, such as control-flow graphs, are generally more stable but computationally expensive to extract. === Class imbalance === In realistic deployment environments, the proportion of malicious samples can be extremely low, ranging from 0.01% to 2% of total data. This unbalanced distribution causes models to develop a bias towards the majority class, achieving high accuracy but failing to identify malicious samples. Prior approaches to this problem have included both data-level solutions and sequence-specific models. Methods like n-gram and Long Short-Term Memory (LSTM) networks can model sequential data, but their performance has been shown to decline significantly when malware samples are realistically proportioned in the training set, demonstrating the limitations in

  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 →