Whisper is a machine learning model for speech recognition and transcription, created by OpenAI and first released as open-source software in September 2022. It is capable of transcribing speech in English and multiple other languages, and can translate several non-English languages into English. Whisper is a weakly-supervised deep learning acoustic model, made using an encoder-decoder transformer architecture. OpenAI claims that the combination of different training data and post-training filtering used in its development has led to improved recognition of accents, background noise, and jargon compared to previous approaches. While the model does not outperform larger, more specialized models and still experiences AI hallucination, it has been showed to be useful for general sound recognition and has many applications across different industries. == Background == Speech recognition has had a long history in research; the first approaches made use of statistical methods, such as dynamic time warping, and later hidden Markov models. At around the 2010s, deep neural network approaches became more common for speech recognition models, which were enabled by the availability of large datasets ("big data") and increased computational performance. Early approaches to deep learning in speech recognition included convolutional neural networks, which were limited due to their inability to capture sequential data, which later led to developments of Seq2seq approaches, which include recurrent neural networks, which made use of long short-term memory. Transformers, introduced in 2017 by Google, displaced many prior state-of-the-art approaches across a wide range in machine learning, and started becoming the core neural architecture in fields such as language modeling and computer vision. Weakly-supervised approaches to training acoustic models were recognized in the early 2020s as promising for speech recognition approaches using deep neural networks. According to a NYT report, in 2021 OpenAI believed they exhausted sources of higher-quality data to train their large language models and decided to complement scraped web text with transcriptions of YouTube videos and podcasts, and developed Whisper to solve this task. Whisper Large V2 was released on December 8, 2022, followed by Whisper Large V3 being released in November 2023, during the OpenAI Dev Day. In March 2025, OpenAI released new transcription models based on GPT-4o and GPT-4o mini, both of which have lower error rates than Whisper. == Architecture == The Whisper architecture is based on an encoder-decoder transformer. Input audio is resampled to 16,000 Hertz (Hz) and converted to an 80-channel Log-magnitude Mel spectrogram using 25 ms windows with a 10 ms stride. The spectrogram is then normalized to a [-1, 1] range with near-zero mean. The encoder takes this Mel spectrogram as input and processes it. It first passes through two convolutional layers. Sinusoidal positional embeddings are added. It is then processed by a series of Transformer encoder blocks (with pre-activation residual connections). The encoder's output is layer normalized. The decoder is a standard transformer decoder. It has the same width and Transformer blocks as the encoder. It uses learned positional embeddings and tied input-output token representations (using the same weight matrix for both the input and output embeddings). It uses a byte-pair encoding tokenizer, of the same kind as used in GPT-2. English-only models use the GPT-2 vocabulary, while multilingual models employ a re-trained multilingual vocabulary with the same number of words. Special tokens are used to allow the decoder to perform multiple tasks: Tokens that denote language (one unique token per language). Tokens that specify task (<|transcribe|> or <|translate|>). Tokens that specify if no timestamps are present (<|notimestamps|>). If the token is not present, then the decoder predicts timestamps relative to the segment, and quantized to 20 ms intervals. <|nospeech|> for voice activity detection. <|startoftranscript|>, and <|endoftranscript|> . Any text that appears before <|startoftranscript|> is not generated by the decoder, but given to the decoder as context. Loss is only computed over non-contextual parts of the sequence, i.e. tokens between these two special tokens. == Training data == The training dataset consists of 680,000 hours of labeled audio-transcript pairs sourced from the internet using semi-supervised learning. This includes 117,000 hours in 96 non-English languages and 125,000 hours of X→English translation data, where X stands for any non-English language. Preprocessing involved standardization of transcripts, filtering to remove machine-generated transcripts using heuristics (e.g., punctuation, capitalization), language identification and matching with transcripts, fuzzy deduplication, and deduplication with evaluation datasets to avoid data contamination. Speechless segments were also included to allow voice activity detection training. For the files still remaining after the filtering process, audio files were then broken into 30-second segments paired with the subset of the transcript that occurs within that time. If this predicted spoken language differed from the language of the text transcript associated with the audio, that audio-transcript pair was not used for training the speech recognition models, but instead for training translation. The model was trained using the AdamW optimizer with gradient norm clipping and a linear learning rate decay with warmup, with batch size 256 segments. Training proceeded for 1 million updates (approximately 2-3 epochs). No data augmentation or regularization, except for the Large V2 model, which used SpecAugment, Stochastic Depth, and BPE Dropout. The training used data parallelism with float16, dynamic loss scaling, and activation checkpointing. === Post-training filtering === After training the first model, researchers ran it on different subsets of the training data, each representing a distinct source. Data sources were ranked by a combination of their error rate and size. Manual inspection of the top-ranked sources (high error, large size) helped determine if the source was low quality (e.g., partial transcriptions, inaccurate alignment). After training, it was fine-tuned to suppress the prediction of speaker names and low-quality sources were then removed. == Capacity == While Whisper does not outperform models which specialize in the LibriSpeech dataset, when tested across many datasets, it is more robust and makes 55.2% fewer errors than other models. Whisper has a differing error rate with respect to transcribing different languages, with a higher word error rate in languages not well-represented in the training data. The authors found that multi-task learning improved overall performance compared to models specialized to one task. They conjectured that the best Whisper model trained is still underfitting the dataset, and larger models and longer training can result in better models. Third-party evaluations have found varying levels of AI hallucination. A study of transcripts of public meetings found hallucinations in eight out of every 10 transcripts, while an engineer discovered hallucinations in "about half" of 100 hours of transcriptions and a developer identified them in "nearly every one" of 26,000 transcripts. A study of 13,140 short audio segments (averaging 10 seconds) found 187 hallucinations (1.4%), 38% of which generated text that could be harmful because it inserted false references to things like race, non-existent medications, or violent events that were not in the audio. == Applications == The model has been used as the base for many applications, such as a unified model for speech recognition and more general sound recognition. Whisper has also been integrated into the workflow of biomedical research. In 2025, a study on Alzheimer's disease detection used the model to transcribe spontaneous speech recordings. The transcripts that were generated by the model were combined with LLM vector embeddings and traditional classifiers to help classify the patients' health. Another application is when OVALYTICS incorporated Whisper to transcribe YouTube videos and automate content moderation systems, which improved its detection of offensive content. The model has also been used in academic libraries and cultral heritage institutions to generate transcripts and captions for their digitized audiovisual collections. In a 2025 case study, Emory University Libraries found that Whisper reduced the labor used in transcription by around 30-35%, shifting work from text creation to text correction. However, human review is still necessary to make sure accuracy, formatting, and accessibility are all standard.
Graphics address remapping table
The graphics address remapping table (GART), also known as the graphics aperture remapping table, or graphics translation table (GTT), is an I/O memory management unit (IOMMU) used by Accelerated Graphics Port (AGP) and PCI Express (PCIe) graphics cards. The GART allows the graphics card direct memory access (DMA) to the host system memory, through which buffers of textures, polygon meshes and other data are loaded. AMD later reused the same mechanism for I/O virtualization with other peripherals including disk controllers and network adapters. A GART is used as a means of data exchange between the main memory and video memory through which buffers (i.e. paging/swapping) of textures, polygon meshes and other data are loaded, but can also be used to expand the amount of video memory available for systems with only integrated or shared graphics (i.e. no discrete or inbuilt graphics processor), such as Intel HD Graphics processors. However, this type of memory (expansion) remapping has a caveat that affects the entire system: specifically, any GART, pre-allocated memory becomes pooled and cannot be utilised for any other purposes but graphics memory and display rendering. Since PCI Express, the GART is extended to the GTT (Graphics Translation Table), which act as a buffer or cache between system memory and graphics card, and in PCI Express, the GTT buffer size is changeable by the GPU driver. == Operating system support == === Windows === Support for AGP GART was added since Windows 95 OSR2. Later, support for GTT was added since Windows XP SP2 and Windows Vista. === Linux === Jeff Hartmann served as the primary maintainer of the Linux kernel's agpgart driver, which began as part of Brian Paul's Utah GLX accelerated Mesa 3D driver project. The developers primarily targeted Linux 2.4.x kernels, but made patches available against older 2.2.x kernels. Dave Jones heavily reworked agpgart for the Linux 2.6.x kernels, along with more contributions from Jeff Hartmann. === FreeBSD === In FreeBSD, the agpgart driver appeared in its 4.1 release. === Solaris === AGPgart support was introduced into Solaris Express Developer Edition as of its 7/05 release.
Rendezvous hashing
Rendezvous or highest random weight (HRW) hashing is an algorithm that allows clients to achieve distributed agreement on a set of k {\displaystyle k} options out of a possible set of n {\displaystyle n} options. A typical application is when clients need to agree on which sites (or proxies) objects are assigned to. Consistent hashing addresses the special case k = 1 {\displaystyle k=1} using a different method. Rendezvous hashing is both much simpler and more general than consistent hashing (see below). == History == Rendezvous hashing was invented by David Thaler and Chinya Ravishankar at the University of Michigan in 1996. Consistent hashing appeared a year later in the literature. Given its simplicity and generality, rendezvous hashing is now being preferred to consistent hashing in real-world applications. Rendezvous hashing was used very early on in many applications including mobile caching, router design, secure key establishment, and sharding and distributed databases. Other examples of real-world systems that use Rendezvous Hashing include the GitHub load balancer, the Apache Ignite distributed database, the Tahoe-LAFS file store, the CoBlitz large-file distribution service, Apache Druid, IBM's Cloud Object Store, the Arvados Data Management System, Apache Kafka, and the Twitter EventBus pub/sub platform. One of the first applications of rendezvous hashing was to enable multicast clients on the Internet (in contexts such as the MBONE) to identify multicast rendezvous points in a distributed fashion. It was used in 1998 by Microsoft's Cache Array Routing Protocol (CARP) for distributed cache coordination and routing. Some Protocol Independent Multicast routing protocols use rendezvous hashing to pick a rendezvous point. == Problem definition and approach == === Algorithm === Rendezvous hashing solves a general version of the distributed hash table problem: We are given a set of n {\displaystyle n} sites (servers or proxies, say). How can any set of clients, given an object O {\displaystyle O} , agree on a k-subset of sites to assign to O {\displaystyle O} ? The standard version of the problem uses k = 1. Each client is to make its selection independently, but all clients must end up picking the same subset of sites. This is non-trivial if we add a minimal disruption constraint, and require that when a site fails or is removed, only objects mapping to that site need be reassigned to other sites. The basic idea is to give each site S j {\displaystyle S_{j}} a score (a weight) for each object O i {\displaystyle O_{i}} , and assign the object to the highest scoring site. All clients first agree on a hash function h ( ⋅ ) {\displaystyle h(\cdot )} . For object O i {\displaystyle O_{i}} , the site S j {\displaystyle S_{j}} is defined to have weight w i , j = h ( O i , S j ) {\displaystyle w_{i,j}=h(O_{i},S_{j})} . Each client independently computes these weights w i , 1 , w i , 2 … w i , n {\displaystyle w_{i,1},w_{i,2}\dots w_{i,n}} and picks the k sites that yield the k largest hash values. The clients have thereby achieved distributed k {\displaystyle k} -agreement. If a site S {\displaystyle S} is added or removed, only the objects mapping to S {\displaystyle S} are remapped to different sites, satisfying the minimal disruption constraint above. The HRW assignment can be computed independently by any client, since it depends only on the identifiers for the set of sites S 1 , S 2 … S n {\displaystyle S_{1},S_{2}\dots S_{n}} and the object being assigned. HRW easily accommodates different capacities among sites. If site S k {\displaystyle S_{k}} has twice the capacity of the other sites, we simply represent S k {\displaystyle S_{k}} twice in the list, say, as S k , 1 , S k , 2 {\displaystyle S_{k,1},S_{k,2}} . Clearly, twice as many objects will now map to S k {\displaystyle S_{k}} as to the other sites. === Properties === Consider the simple version of the problem, with k = 1, where all clients are to agree on a single site for an object O. Approaching the problem naively, it might appear sufficient to treat the n sites as buckets in a hash table and hash the object name O into this table. Unfortunately, if any of the sites fails or is unreachable, the hash table size changes, forcing all objects to be remapped. This massive disruption makes such direct hashing unworkable. Under rendezvous hashing, however, clients handle site failures by picking the site that yields the next largest weight. Remapping is required only for objects currently mapped to the failed site, and disruption is minimal. Rendezvous hashing has the following properties: Low overhead: The hash function used is efficient, so overhead at the clients is very low. Load balancing: Since the hash function is randomizing, each of the n sites is equally likely to receive the object O. Loads are uniform across the sites. Site capacity: Sites with different capacities can be represented in the site list with multiplicity in proportion to capacity. A site with twice the capacity of the other sites will be represented twice in the list, while every other site is represented once. High hit rate: Since all clients agree on placing an object O into the same site SO, each fetch or placement of O into SO yields the maximum utility in terms of hit rate. The object O will always be found unless it is evicted by some replacement algorithm at SO. Minimal disruption: When a site fails, only the objects mapped to that site need to be remapped. Disruption is at the minimal possible level. Distributed k-agreement: Clients can reach distributed agreement on k sites simply by selecting the top k sites in the ordering. == O(log n) running time via skeleton-based hierarchical rendezvous hashing == The standard version of Rendezvous Hashing described above works quite well for moderate n, but when n {\displaystyle n} is extremely large, the hierarchical use of Rendezvous Hashing achieves O ( log n ) {\displaystyle O(\log n)} running time. This approach creates a virtual hierarchical structure (called a "skeleton"), and achieves O ( log n ) {\displaystyle O(\log n)} running time by applying HRW at each level while descending the hierarchy. The idea is to first choose some constant m {\displaystyle m} and organize the n {\displaystyle n} sites into c = ⌈ n / m ⌉ {\displaystyle c=\lceil n/m\rceil } clusters C 1 = { S 1 , S 2 … S m } , C 2 = { S m + 1 , S m + 2 … S 2 m } … {\displaystyle C_{1}=\left\{S_{1},S_{2}\dots S_{m}\right\},C_{2}=\left\{S_{m+1},S_{m+2}\dots S_{2m}\right\}\dots } Next, build a virtual hierarchy by choosing a constant f {\displaystyle f} and imagining these c {\displaystyle c} clusters placed at the leaves of a tree T {\displaystyle T} of virtual nodes, each with fanout f {\displaystyle f} . In the accompanying diagram, the cluster size is m = 4 {\displaystyle m=4} , and the skeleton fanout is f = 3 {\displaystyle f=3} . Assuming 108 sites (real nodes) for convenience, we get a three-tier virtual hierarchy. Since f = 3 {\displaystyle f=3} , each virtual node has a natural numbering in octal. Thus, the 27 virtual nodes at the lowest tier would be numbered 000 , 001 , 002 , . . . , 221 , 222 {\displaystyle 000,001,002,...,221,222} in octal (we can, of course, vary the fanout at each level - in that case, each node will be identified with the corresponding mixed-radix number). The easiest way to understand the virtual hierarchy is by starting at the top, and descending the virtual hierarchy. We successively apply Rendezvous Hashing to the set of virtual nodes at each level of the hierarchy, and descend the branch defined by the winning virtual node. We can in fact start at any level in the virtual hierarchy. Starting lower in the hierarchy requires more hashes, but may improve load distribution in the case of failures. For example, instead of applying HRW to all 108 real nodes in the diagram, we can first apply HRW to the 27 lowest-tier virtual nodes, selecting one. We then apply HRW to the four real nodes in its cluster, and choose the winning site. We only need 27 + 4 = 31 {\displaystyle 27+4=31} hashes, rather than 108. If we apply this method starting one level higher in the hierarchy, we would need 9 + 3 + 4 = 16 {\displaystyle 9+3+4=16} hashes to get to the winning site. The figure shows how, if we proceed starting from the root of the skeleton, we may successively choose the virtual nodes ( 2 ) 3 {\displaystyle (2)_{3}} , ( 20 ) 3 {\displaystyle (20)_{3}} , and ( 200 ) 3 {\displaystyle (200)_{3}} , and finally end up with site 74. The virtual hierarchy need not be stored, but can be created on demand, since the virtual nodes names are simply prefixes of base- f {\displaystyle f} (or mixed-radix) representations. We can easily create appropriately sorted strings from the digits, as required. In the example, we would be working with the strings 0 , 1 , 2 {\displaystyle 0,1,2} (at tier 1), 20 , 21 , 22 {\displaystyle 20,21,22} (at tier 2), and 200 , 201 , 202
Certifying algorithm
In theoretical computer science, a certifying algorithm is an algorithm that outputs, together with a solution to the problem it solves, a proof that the solution is correct. A certifying algorithm is said to be efficient if the combined runtime of the algorithm and a proof checker is slower by at most a constant factor than the best known non-certifying algorithm for the same problem. The proof produced by a certifying algorithm should be in some sense simpler than the algorithm itself, for otherwise any algorithm could be considered certifying (with its output verified by running the same algorithm again). Sometimes this is formalized by requiring that a verification of the proof take less time than the original algorithm, while for other problems (in particular those for which the solution can be found in linear time) simplicity of the output proof is considered in a less formal sense. For instance, the validity of the output proof may be more apparent to human users than the correctness of the algorithm, or a checker for the proof may be more amenable to formal verification. Implementations of certifying algorithms that also include a checker for the proof generated by the algorithm may be considered to be more reliable than non-certifying algorithms. For, whenever the algorithm is run, one of three things happens: it produces a correct output (the desired case), it detects a bug in the algorithm or its implication (undesired, but generally preferable to continuing without detecting the bug), or both the algorithm and the checker are faulty in a way that masks the bug and prevents it from being detected (undesired, but unlikely as it depends on the existence of two independent bugs). == Examples == Many examples of problems with checkable algorithms come from graph theory. For instance, a classical algorithm for testing whether a graph is bipartite would simply output a Boolean value: true if the graph is bipartite, false otherwise. In contrast, a certifying algorithm might output a 2-coloring of the graph in the case that it is bipartite, or a cycle of odd length if it is not. Any graph is bipartite if and only if it can be 2-colored, and non-bipartite if and only if it contains an odd cycle. Both checking whether a 2-coloring is valid and checking whether a given odd-length sequence of vertices is a cycle may be performed more simply than testing bipartiteness. Analogously, it is possible to test whether a given directed graph is acyclic by a certifying algorithm that outputs either a topological order or a directed cycle. It is possible to test whether an undirected graph is a chordal graph by a certifying algorithm that outputs either an elimination ordering (an ordering of all vertices such that, for every vertex, the neighbors that are later in the ordering form a clique) or a chordless cycle. And it is possible to test whether a graph is planar by a certifying algorithm that outputs either a planar embedding or a Kuratowski subgraph. The extended Euclidean algorithm for the greatest common divisor of two integers x and y is certifying: it outputs three integers g (the divisor), a, and b, such that ax + by = g. This equation can only be true of multiples of the greatest common divisor, so testing that g is the greatest common divisor may be performed by checking that g divides both x and y and that this equation is correct.
Informetrics
Informetrics is the study of quantitative aspects of information, it is an extension and evolution of traditional bibliometrics and scientometrics. Informetrics uses bibliometrics and scientometrics methods to study mainly the problems of literature information management and evaluation of science and technology. Informetrics is an independent discipline that uses quantitative methods from mathematics and statistics to study the process, phenomena, and law of informetrics. Informetrics has gained more attention as it is a common scientific method for academic evaluation, research hotspots in discipline, and trend analysis. Informetrics includes the production, dissemination, and use of all forms of information, regardless of its form or origin. Informetrics encompasses the following fields: Scientometrics, which studies quantitative aspects of science Webometrics, which studies quantitative aspects of the World Wide Web Bibliometrics, which studies quantitative aspects of recorded information Cybermetrics, which is similar to webometrics, but broadens its definition to include electronic resources == Origin and Development == The term informetrics (French: informétrie) was coined by German scholar Otto Nacke in 1979, and came from the German word 'informetrie’. The corresponding English terminology soon appeared in the subsequent literature. In September 1980, Professor Otto Nacke introduced the term 'informetrics' at the first seminar on Informetrics in Frankfurt, Germany. Later, Committee on Informetrics has established through The International Federation for Information and Documentation (FID). In 1987, informetrics started to be officially recognized by the international information community and several foreign information scientists. In 1988, at First International Conference on Bibliometrics and Theoretical Aspects of Information Retrieval Archived 2022-05-23 at the Wayback Machine, Brooks suggested bibliometrics and scientometrics can be included in the field of informetrics. In 1990, Leo Egghe and Ronald Rousseau proposed the formation of the discipline of informetrics: statistical bibliography (1923) to bibliometrics and scientometrics (1969) and then to informetrics (1979). In 1993, the International Society for Scientometrics and Informetrics (ISSI) Archived 2023-11-05 at the Wayback Machine was founded at the International Conference on Bibliometrics, Informetrics and Scientometrics in Berlin, and the first one was held in Belgium and organized by Leo Egghe and Ronald Rousseau. The society was formally incorporated in 1994 in the Netherlands and plays a significant role in the development of informetrics. The ISSI aims to promote the "exchange and communication of professional information in the fields of scientometrics and informetrics, including improve standards, theory and practice, as well as promote research, education and training". In addition, to "engage in relevant public conversation and policy discussions". In the western world, 20th century's Informetrics is mostly based on Lotka's law, named after Alfred J. Lotka, Zipf's law, named after George Kingsley Zipf, Bradford's law named after Samuel C. Bradford and on the work of Derek J. de Solla Price, Gerard Salton, Leo Egghe, Ronald Rousseau, Tibor Braun, Olle Persson, Peter Ingwersen, Manfred Bonitz, and Eugene Garfield. == Difference Between Informetrics, Bibliometrics and Scientometrics == Since the 1960s, three similar terms have emerged in the fields of library science, philology and science of science, they are bibliometrics, scientometrics and informetrics, representing three very similar quantitative sub-disciplines. The three metrics terms can be confusing and often misused. Informetrics and bibliometrics interpenetrate each other but have different aspects in research object, research scope, and measuring unit. Informetrics and scientometrics are very different in their research purpose and research object, as well as the research scope and application. Bibliometrics is categorised under the field of library science, it uses mathematical and statistical methods to describe, evaluate, and predict the current status and trends of science and technology. Also to study the "distribution structure, quantitative relationship, change law and quantitative management of literature information, quantitative relationships, patterns and quantitative management of literature and information". The term was first used by Alan Pritchard in 1969 in his paper Statistical Bibliography or Bibliometrics?. Scientometrics is a branch of science that quantitatively evaluates and predicts the process and management of scientific activities in order to reveal their development patterns and trends. The definition of scientometrics was described by Derek De Solla Price in his book Science to Science as the “quantitative study of science, communication in science, and science policy”. === Links between the three metrics terms === The most prominent connection between the three metrics terms is in their research objects. Since all three disciplines use literature information as their research object, therefore, they have some similarities and overlaps in their research methods and fields. Moreover, they all use mathematical methods as the basic research methods and they all apply the three basic laws, Bradford's law, Lotka's law and Zipf's law. === Distinctions between the three metrics terms === The distinction between the three metrics terms can tell from their research object and research purpose. The research of bibliometrics focuses on the analysis of "scientific output in the form of articles, publications, citations, and others". Scientometrics is to measure the basic characteristics and laws of scientific activities. Where informetrics is to investigate information sources and information distribution process. == Concept and System Structure == === Purpose of Informetrics Research === The main purpose of informetrics is to use its theocratical research to solve the methodological issues in the research process, and to discover and reveal the basic laws of information distribution through the study of information process and phenomenon. In this way, makes information management more scientific and provides a quantitative basis for information services and information management decisions. For informetrics, it is necessary to bring quantitative analysis methods to further reveal the structure of information units and the "quantitative change law of literature information”. Further to this, to improve the scientific accuracy of information science from a theoretical point of view. At the same time, to better solve the basic contradictions in the information service, overcome the information crisis, and make the information management work more effective to serve science and technology, economic and social development. Quantitative analysis of bibliographic data was pioneered by Robert K. Merton in an article called Science, Technology, and Society in Seventeenth Century England and originally published by Merton in 1938. === The Significance of Informetrics Research === The significance of informetrics research is to summarize various empirical laws from the theoretical point of view, at the same time test and modify the various empirical laws in the new information unit conditions, and explore its new applicability, therefore, the scientific nature of information science can be improved, but also to provide theoretical guidance for practical work. === The Objects of Informetrics Research === The object of informetrics is broader than the field of bibliometrics and scientometrics, including "messages, data, events, objects, text, and documents”. Informetrics is often used to inform policies and decisions across a broad range of fields, such as economy, politics, technology and social spheres that "influence the flow and use patterns of information". Tague-Sutcliffe describes the following uses of informetrics: Citation analysis; Characteristics of authors; Use of recorded information; Obsolescence of the literature; Concomitant growth of new concepts; Characteristics of publication sources; Definition and measurement o information; Growth of subject literature, databases, libraries; Types and characteristics of retrieval performance measures; Statistical aspects of language, word, and phrase frequencies. == Basic Laws == In the field of informetrics research, there are many outstanding contributors in the discipline with a solid knowledge of quantitative research methods. In the early 20th century, several scientists contributed empirical applications that have become the three basic laws of informetrics, Bradford's law, Lotka's law, and Zipf's law, which promote the development of informetrics. === Bradford's Law === The British documentalist and librarian Samuel C. Bradford first discovered the law of concentration and scattering of literature, and in 1934, it has be
Learning automaton
A learning automaton is one type of machine learning algorithm studied since 1970s. Learning automata select their current action based on past experiences from the environment. It will fall into the range of reinforcement learning if the environment is stochastic and a Markov decision process (MDP) is used. == History == Research in learning automata can be traced back to the work of Michael Lvovitch Tsetlin in the early 1960s in the Soviet Union. Together with some colleagues, he published a collection of papers on how to use matrices to describe automata functions. Additionally, Tsetlin worked on reasonable and collective automata behaviour, and on automata games. Learning automata were also investigated by researches in the United States in the 1960s. However, the term learning automaton was not used until Narendra and Thathachar introduced it in a survey paper in 1974. == Definition == A learning automaton is an adaptive decision-making unit situated in a random environment that learns the optimal action through repeated interactions with its environment. The actions are chosen according to a specific probability distribution which is updated based on the environment response the automaton obtains by performing a particular action. With respect to the field of reinforcement learning, learning automata are characterized as policy iterators. In contrast to other reinforcement learners, policy iterators directly manipulate the policy π. Another example for policy iterators are evolutionary algorithms. Formally, Narendra and Thathachar define a stochastic automaton to consist of: a set X of possible inputs, a set Φ = { Φ1, ..., Φs } of possible internal states, a set α = { α1, ..., αr } of possible outputs, or actions, with r ≤ s, an initial state probability vector p(0) = ≪ p1(0), ..., ps(0) ≫, a computable function A which after each time step t generates p(t+1) from p(t), the current input, and the current state, and a function G: Φ → α which generates the output at each time step. In their paper, they investigate only stochastic automata with r = s and G being bijective, allowing them to confuse actions and states. The states of such an automaton correspond to the states of a "discrete-state discrete-parameter Markov process". At each time step t=0,1,2,3,..., the automaton reads an input from its environment, updates p(t) to p(t+1) by A, randomly chooses a successor state according to the probabilities p(t+1) and outputs the corresponding action. The automaton's environment, in turn, reads the action and sends the next input to the automaton. Frequently, the input set X = { 0,1 } is used, with 0 and 1 corresponding to a nonpenalty and a penalty response of the environment, respectively; in this case, the automaton should learn to minimize the number of penalty responses, and the feedback loop of automaton and environment is called a "P-model". More generally, a "Q-model" allows an arbitrary finite input set X, and an "S-model" uses the interval [0,1] of real numbers as X. A visualised demo/ Art Work of a single Learning Automaton had been developed by μSystems (microSystems) Research Group at Newcastle University. == Finite action-set learning automata == Finite action-set learning automata (FALA) are a class of learning automata for which the number of possible actions is finite or, in more mathematical terms, for which the size of the action-set is finite.
Master data
Master data represents "data about the business entities that provide context for business transactions". The most commonly found categories of master data are parties (individuals and organisations, and their roles, such as customers, suppliers, employees), products, financial structures (such as ledgers and cost centres) and locational concepts. Master data should be distinguished from reference data. While both provide context for business transactions, reference data is concerned with classification and categorisation, while master data is concerned with business entities. Master data is, by its nature, almost always non-transactional in nature. There exist edge cases where an organization may need to treat certain transactional processes and operations as "master data". This arises, for example, where information about master data entities, such as customers or products, is only contained within transactional data such as orders and receipts and is not housed separately. ISO 8000 is the international standard for data quality and data portability in master data. == Alternative definition == An alternative definition of the term master data is that it represents the business objects that contain the most valuable, agreed upon information shared across an organization. In this sense, it gives context to business activities and transactions, answering questions like who, what, when and how as well as expanding the ability to make sense of these activities through categorizations, groupings and hierarchies. It can cover relatively static reference data, transactional, unstructured, analytical, hierarchical and metadata. What constitutes master data under this definition is therefore not about an essential quality of the data (e.g. it is a business entity that provides context for business transactions), but rather about the context in which the organisation has decided to treat the data. == Externally-defined master data == For most organisations, most or all master data is defined and managed within that organisation. Some master data, however, may be externally defined and managed. This represents the single source of basic business data used across a marketplace, regardless of organisation or location. Thus, it can be used by multiple enterprises within a value chain, facilitating "integration of multiple data sources and literally [putting] everyone in the market on the same page." An example of market master data is the Universal Product Code (UPC) found on consumer products. == Master data management == Curating and managing master data is key to ensuring its quality and thus fitness for purpose. All aspects of an organisation, operational and analytical, are greatly dependent on the quality of an organization's master data. Master Data is therefore the focus of the information technology (IT) discipline of master data management (MDM). Without this discipline in place, organisations commonly encounter difficulties with having multiple versions of "the truth" about a business entity, both within individual applications, and distributed across applications.