AI Assistant Zara

AI Assistant Zara — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

Spatial embedding

Spatial embedding is one of feature learning techniques used in spatial analysis where points, lines, polygons or other spatial data types. representing geographic locations are mapped to vectors of real numbers. Conceptually it involves a mathematical embedding from a space with many dimensions per geographic object to a continuous vector space with a much lower dimension. Such embedding methods allow complex spatial data to be used in neural networks and have been shown to improve performance in spatial analysis tasks == Embedded data types == Geographic data can take many forms: text, images, graphs, trajectories, polygons. Depending on the task, there may be a need to combine multimodal data from different sources. The next section describes examples of different types of data and their uses. === Text === Geolocated posts on social media can be used to acquire a library of documents bound to a given place that can be later transformed to embedded vectors using word embedding techniques. === Image === Satellites and aircraft collect digital spatial data acquired from remotely sensed images which can be used in machine learning. They are sometimes hard to analyse using basic image analysis methods and convolutional neural networks can be used to acquire an embedding of images bound to a given geographical object or a region. === Point === A single point of interest (POI) can be assigned multiple features that can be used in machine learning. These could be demographic, transportation, meteorological, or economic data, for example. When embedding single points, it is common to consider the entire set of available points as nodes in a graph. === Line / multiline === Among other things, motion trajectories are represented as lines (multilines). Individual trajectories are embedded taking into account travel time, distances and also features of points visited along the way. Embedding of trajectories allows to improve performance of such tasks as clustering and also categorization. === Polygon === The geographic areas analyzed in machine learning are defined by both administrative boundaries and top-down division into grids of regular shapes such as rectangles, for example. Both types are represented as polygons and, like points, can be assigned different demographic, transportation, or economic features. A polygon can also have features related to the size of the area or shape it represents. === Graph === An example domain where graph representation is used is the street layout in a city, where vertices can be intersections and edges can be roads. The vertices can also be destination points like public transport stops or important points in the city, and the edges represent the flow between them. Embedding graphs or single vertices allows to improve accuracy of analysis methods in which the treated geographical domain can be represented as a network. == Usage == POI recommendation - generating personalized point of interest recommendations based on user preferences. Next/future location prediction - prediction of the next location a person will go to based on their historical trajectory. Zone functions classification - based on different mobility of people or POI distribution a function of a given area in a city can be predicted. Crime prediction - estimation of crime rate in different regions of a city. Local event detection - studying spatio-temporal changes in embeddings can provide valuable information in detection of local event occurring in specific location. Regional mobility popularity prediction - analysis of mobility can show patterns in popularity of different regions in a city. Shape matching - finding a similar shape of given polygon, for example finding building with the same shape as input building. Travel time estimation - predicting estimated travel time given current traffic conditions and special occurring events. Time estimation for on-demand food delivery - estimation of delivery time when placing an order through the website. == Temporal aspect == Some of the data analyzed has a timestamp associated with it. In some cases of data analysis this information is omitted and in others it is used to divide the set into groups. The most common division is the separation of weekdays from weekends or division into hours of the day. This is particularly important in the analysis of mobility data, because the characteristics of mobility during the week and at different times of the day are very different from each other. Another area in which time division into, for example, individual months can be used is in the analysis of tourism of a given region. In order to take such a split into account, embedding methods treat the time stamp specifically or separate versions of the model are developed for different subgroups of the analyzed set.
Read more →
Second-order co-occurrence pointwise mutual information

In computational linguistics, second-order co-occurrence pointwise mutual information (SOC-PMI) is a method used to measure semantic similarity, or how close in meaning two words are. The method does not require the two words to appear together in a text. Instead, it works by analyzing the "neighbor" words that typically appear alongside each of the two target words in a large body of text (corpus). If the two target words frequently share the same neighbors, they are considered semantically similar. For example, the words "cemetery" and "graveyard" may not appear in the same sentence often, but they both frequently appear near words like "buried," "dead," and "funeral." SOC-PMI uses this shared context to determine that they have a similar meaning. The method is called "second-order" because it doesn't look at the direct co-occurrence of the target words (which would be first-order), but at the co-occurrence of their neighbors (a second level of association). The strength of these associations is quantified using pointwise mutual information (PMI). == History == The method builds on earlier work like the PMI-IR algorithm, which used the AltaVista search engine to calculate word association probabilities. The key advantage of a second-order approach like SOC-PMI is its ability to measure similarity between words that do not co-occur often, or at all. The British National Corpus (BNC) has been used as a source for word frequencies and contexts for this method. == Methodology == The SOC-PMI algorithm measures the similarity between two words, w 1 {\displaystyle w_{1}} and w 2 {\displaystyle w_{2}} , in several steps. === Step 1: Score neighboring words with PMI === First, for each target word ( w 1 {\displaystyle w_{1}} and w 2 {\displaystyle w_{2}} ), the algorithm identifies its "neighbor" words within a certain text window (e.g., within 5 words to the left or right) across a large corpus. The strength of the association between a target word t i {\displaystyle t_{i}} and its neighbor w {\displaystyle w} is calculated using pointwise mutual information (PMI). A higher PMI value means the two words appear together more often than would be expected by chance. The PMI between a target word t i {\displaystyle t_{i}} and a neighbor word w {\displaystyle w} is calculated as: f pmi ( t i , w ) = log 2 ⁡ f b ( t i , w ) × m f t ( t i ) f t ( w ) {\displaystyle f^{\text{pmi}}(t_{i},w)=\log _{2}{\frac {f^{b}(t_{i},w)\times m}{f^{t}(t_{i})f^{t}(w)}}} where: f b ( t i , w ) {\displaystyle f^{b}(t_{i},w)} is the number of times t i {\displaystyle t_{i}} and w {\displaystyle w} appear together in the context window. f t ( t i ) {\displaystyle f^{t}(t_{i})} is the total number of times t i {\displaystyle t_{i}} appears in the corpus. f t ( w ) {\displaystyle f^{t}(w)} is the total number of times w {\displaystyle w} appears in the corpus. m {\displaystyle m} is the total number of tokens (words) in the corpus. === Step 2: Create a semantic 'signature' for each word === For each target word ( w 1 {\displaystyle w_{1}} and w 2 {\displaystyle w_{2}} ), the algorithm creates a list of its most significant neighbors. This is done by taking the top β {\displaystyle \beta } neighbor words, sorted in descending order by their PMI score with the target word. This list of top neighbors, X w {\displaystyle X^{w}} , acts as a semantic "signature" for the word w {\displaystyle w} . X w = { X i w } {\displaystyle X^{w}=\{X_{i}^{w}\}} , for i = 1 , 2 , … , β {\displaystyle i=1,2,\ldots ,\beta } The size of this list, β {\displaystyle \beta } , is a parameter of the method. === Step 3: Compare the signatures === The algorithm then compares the signatures of w 1 {\displaystyle w_{1}} and w 2 {\displaystyle w_{2}} . It looks for words that are present in both signatures. The similarity of w 1 {\displaystyle w_{1}} to w 2 {\displaystyle w_{2}} is calculated by summing the PMI scores of w 2 {\displaystyle w_{2}} with every word in w 1 {\displaystyle w_{1}} 's signature list. The β {\displaystyle \beta } -PMI summation function defines this score. The score for w 1 {\displaystyle w_{1}} with respect to w 2 {\displaystyle w_{2}} is: f ( w 1 , w 2 , β ) = ∑ i = 1 β ( f pmi ( X i w 1 , w 2 ) ) γ {\displaystyle f(w_{1},w_{2},\beta )=\sum _{i=1}^{\beta }(f^{\text{pmi}}(X_{i}^{w_{1}},w_{2}))^{\gamma }} This sum only includes terms where the PMI value is positive. The exponent γ {\displaystyle \gamma } (with a value > 1) is used to give more weight to neighbors that are more strongly associated with w 2 {\displaystyle w_{2}} . This calculation is done in both directions: The similarity of w 1 {\displaystyle w_{1}} with respect to w 2 {\displaystyle w_{2}} : f ( w 1 , w 2 , β 1 ) = ∑ i = 1 β 1 ( f pmi ( X i w 1 , w 2 ) ) γ {\displaystyle f(w_{1},w_{2},\beta _{1})=\sum _{i=1}^{\beta _{1}}(f^{\text{pmi}}(X_{i}^{w_{1}},w_{2}))^{\gamma }} The similarity of w 2 {\displaystyle w_{2}} with respect to w 1 {\displaystyle w_{1}} : f ( w 2 , w 1 , β 2 ) = ∑ i = 1 β 2 ( f pmi ( X i w 2 , w 1 ) ) γ {\displaystyle f(w_{2},w_{1},\beta _{2})=\sum _{i=1}^{\beta _{2}}(f^{\text{pmi}}(X_{i}^{w_{2}},w_{1}))^{\gamma }} === Step 4: Calculate final similarity score === Finally, the total semantic similarity is the average of the two scores from the previous step. S i m ( w 1 , w 2 ) = f ( w 1 , w 2 , β 1 ) β 1 + f ( w 2 , w 1 , β 2 ) β 2 {\displaystyle \mathrm {Sim} (w_{1},w_{2})={\frac {f(w_{1},w_{2},\beta _{1})}{\beta _{1}}}+{\frac {f(w_{2},w_{1},\beta _{2})}{\beta _{2}}}} This score can be normalized to fall between 0 and 1. For example, using this method, the words cemetery and graveyard achieve a high similarity score of 0.986 (with specific parameter settings).
Read more →
Trigram

Trigrams are a special case of the n-gram, where n is 3. They are often used in natural language processing for performing statistical analysis of texts and in cryptography for control and use of ciphers and codes. See results of analysis of "Letter Frequencies in the English Language". == Frequency == Context is very important, varying analysis rankings and percentages are easily derived by drawing from different sample sizes, different authors; or different document types: poetry, science-fiction, technology documentation; and writing levels: stories for children versus adults, military orders, and recipes. Typical cryptanalytic frequency analysis finds that the 16 most common character-level trigrams in English are: Because encrypted messages sent by telegraph often omit punctuation and spaces, cryptographic frequency analysis of such messages includes trigrams that straddle word boundaries. This causes trigrams such as "edt" to occur frequently, even though it may never occur in any one word of those messages. == Examples == The sentence "the quick red fox jumps over the lazy brown dog" has the following word-level trigrams: the quick red quick red fox red fox jumps fox jumps over jumps over the over the lazy the lazy brown lazy brown dog And the word-level trigram "the quick red" has the following character-level trigrams (where an underscore "_" marks a space): the he_ e_q _qu qui uic ick ck_ k_r _re red
Read more →
GPTs

GPTs are custom versions of ChatGPT with added instructions and extra knowledge. GPTs can be used and created from the GPT Store. Any user can easily create them without any programming knowledge. GPTs can be tailored for specific writing styles, topics, or tasks. The ability to create GPTs was introduced in November 2023, and by January 2024, more than 3 million GPTs had been published. == Features and uses == GPTs can be configured to answer complex questions in specific fields, solve problems, provide image-based information, or create digital content. They can be programmed as educational tools, purchasing guides, or technical advisors, as well as for many others applications. GPTs are accessed from the GPT Store section of the ChatGPT web page. The “Explore GPT” link opens the store where the most popular GPTs in each section are highlighted. The GPTs are organized by categories. The store also uses a rating system based on user experiences similar to that used by other app stores such as Apple's App Store or Google Play. Those with the best ratings appear at the top of each category. According to La Vanguardia, the most popular categories are: Personal assistants Learning to program Image generation Creative writing Gaming Entertainment It is expected that in the future the creators of GPTs will be able to monetize them. Companies like Moderna are using GPTs to assist in various specific business tasks. The company has created 750 GPTs for its own internal use. == Configuration == Creating GPTs does not require prior programming knowledge. Free users can use existing GPTs but cannot create their own. Paying subscribers can use the editor on the ChatGPT site to configure the GPT's name, image and description, instructions and access to APIs, along with visibility options. == Criticism == The implementation and use of GPTs has not been without criticism. The GPT Store has been criticized for the proliferation of low-quality GPTs and spam due to a lack of effective moderation. There are also concerns about data privacy and security, as GPTs may collect and use personal information in ways that are not always transparent to users.
Read more →
Referring expression generation

Referring expression generation (REG) is the subtask of natural language generation (NLG) that received most scholarly attention. While NLG is concerned with the conversion of non-linguistic information into natural language, REG focuses only on the creation of referring expressions (noun phrases) that identify specific entities called targets. This task can be split into two sections. The content selection part determines which set of properties distinguish the intended target and the linguistic realization part defines how these properties are translated into natural language. A variety of algorithms have been developed in the NLG community to generate different types of referring expressions. == Types of referring expressions == A referring expression (RE), in linguistics, is any noun phrase, or surrogate for a noun phrase, whose function in discourse is to identify some individual object (thing, being, event...) The technical terminology for identify differs a great deal from one school of linguistics to another. The most widespread term is probably refer, and a thing identified is a referent, as for example in the work of John Lyons. In linguistics, the study of reference relations belongs to pragmatics, the study of language use, though it is also a matter of great interest to philosophers, especially those wishing to understand the nature of knowledge, perception and cognition more generally. Various devices can be used for reference: determiners, pronouns, proper names... Reference relations can be of different kinds; referents can be in a "real" or imaginary world, in discourse itself, and they may be singular, plural, or collective. === Pronouns === The simplest type of referring expressions are pronoun such as he and it. The linguistics and natural language processing communities have developed various models for predicting anaphor referents, such as centering theory, and ideally referring-expression generation would be based on such models. However most NLG systems use much simpler algorithms, for example using a pronoun if the referent was mentioned in the previous sentence (or sentential clause), and no other entity of the same gender was mentioned in this sentence. === Definite noun phrases === There has been a considerable amount of research on generating definite noun phrases, such as the big red book. Much of this builds on the model proposed by Dale and Reiter. This has been extended in various ways, for example Krahmer et al. present a graph-theoretic model of definite NP generation with many nice properties. In recent years a shared-task event has compared different algorithms for definite NP generation, using the TUNA corpus. === Spatial and temporal reference === Recently there has been more research on generating referring expressions for time and space. Such references tend to be imprecise (what is the exact meaning of tonight?), and also to be interpreted in different ways by different people. Hence it may be necessary to explicitly reason about false positive vs false negative tradeoffs, and even calculate the utility of different possible referring expressions in a particular task context. === Criteria for good expressions === Ideally, a good referring expression should satisfy a number of criteria: Referential success: It should unambiguously identify the referent to the reader. Ease of comprehension: The reader should be able to quickly read and understand it. Computational complexity: The generation algorithm should be fast No false inferences: The expression should not confuse or mislead the reader by suggesting false implicatures or other pragmatic inferences. For example, a reader may be confused if he is told Sit by the brown wooden table in a context where there is only one table. == History == === Pre-2000 era === REG goes back to the early days of NLG. One of the first approaches was done by Winograd in 1972 who developed an "incremental" REG algorithm for his SHRDLU program. Afterwards researchers started to model the human abilities to create referring expressions in the 1980s. This new approach to the topic was influenced by the researchers Appelt and Kronfeld who created the programs KAMP and BERTRAND and considered referring expressions as parts of bigger speech acts. Some of their most interesting findings were the fact that referring expressions can be used to add information beyond the identification of the referent as well as the influence of communicative context and the Gricean maxims on referring expressions. Furthermore, its skepticism concerning the naturalness of minimal descriptions made Appelt and Kronfeld's research a foundation of later work on REG. The search for simple, well-defined problems changed the direction of research in the early 1990s. This new approach was led by Dale and Reiter who stressed the identification of the referent as the central goal. Like Appelt they discuss the connection between the Gricean maxims and referring expressions in their culminant paper in which they also propose a formal problem definition. Furthermore, Reiter and Dale discuss the Full Brevity and Greedy Heuristics algorithms as well as their Incremental Algorithm(IA) which became one of the most important algorithms in REG. === Later developments === After 2000 the research began to lift some of the simplifying assumptions, that had been made in early REG research in order to create more simple algorithms. Different research groups concentrated on different limitations creating several expanded algorithms. Often these extend the IA in a single perspective for example in relation to: Reference to Sets like "the t-shirt wearers" or "the green apples and the banana on the left" Relational Descriptions like "the cup on the table" or "the woman who has three children" Context Dependency, Vagueness and Gradeability include statements like "the older man" or "the car on the left" which are often unclear without a context Salience and Generation of Pronouns are highly discourse dependent making for example "she" a reference to "the (most salient) female person" Many simplifying assumptions are still in place or have just begun to be worked on. Also a combination of the different extensions has yet to be done and is called a "non-trivial enterprise" by Krahmer and van Deemter. Another important change after 2000 was the increasing use of empirical studies in order to evaluate algorithms. This development took place due to the emergence of transparent corpora. Although there are still discussions about what the best evaluation metrics are, the use of experimental evaluation has already led to a better comparability of algorithms, a discussion about the goals of REG and more task-oriented research. Furthermore, research has extended its range to related topics such as the choice of Knowledge Representation(KR) Frameworks. In this area the main question, which KR framework is most suitable for the use in REG remains open. The answer to this question depends on how well descriptions can be expressed or found. A lot of the potential of KR frameworks has been left unused so far. Some of the different approaches are the usage of: Graph search which treats relations between targets in the same way as properties. Constraint Satisfaction which allows for a separation between problem specification and the implementation. Modern Knowledge Representation which offers logical inference in for example Description Logic or Conceptual Graphs. == Problem definition == Dale and Reiter (1995) think about referring expressions as distinguishing descriptions. They define: The referent as the entity that should be described The context set as set of salient entities The contrast set or potential distractors as all elements of the context set except the referent A property as a reference to a single attribute–value pair Each entity in the domain can be characterised as a set of attribute–value pairs for example ⟨ {\displaystyle \langle } type, dog ⟩ {\displaystyle \rangle } , ⟨ {\displaystyle \langle } gender, female ⟩ {\displaystyle \rangle } or ⟨ {\displaystyle \langle } age, 10 years ⟩ {\displaystyle \rangle } . The problem then is defined as follows: Let r {\displaystyle r} be the intended referent, and C {\displaystyle C} be the contrast set. Then, a set L {\displaystyle L} of attribute–value pairs will represent a distinguishing description if the following two conditions hold: Every attribute–value pair in L {\displaystyle L} applies to r {\displaystyle r} : that is, every element of L {\displaystyle L} specifies an attribute–value that r {\displaystyle r} possesses. For every member c {\displaystyle c} of C {\displaystyle C} , there is at least one element l {\displaystyle l} of L {\displaystyle L} that does not apply to c {\displaystyle c} : that is, there is an l {\displaystyle l} in L {\displaystyle L} that specifies an attribute–value that c {\displaystyle c} does not possess. l {\displaystyle l} is said
Read more →
Trigram

Trigrams are a special case of the n-gram, where n is 3. They are often used in natural language processing for performing statistical analysis of texts and in cryptography for control and use of ciphers and codes. See results of analysis of "Letter Frequencies in the English Language". == Frequency == Context is very important, varying analysis rankings and percentages are easily derived by drawing from different sample sizes, different authors; or different document types: poetry, science-fiction, technology documentation; and writing levels: stories for children versus adults, military orders, and recipes. Typical cryptanalytic frequency analysis finds that the 16 most common character-level trigrams in English are: Because encrypted messages sent by telegraph often omit punctuation and spaces, cryptographic frequency analysis of such messages includes trigrams that straddle word boundaries. This causes trigrams such as "edt" to occur frequently, even though it may never occur in any one word of those messages. == Examples == The sentence "the quick red fox jumps over the lazy brown dog" has the following word-level trigrams: the quick red quick red fox red fox jumps fox jumps over jumps over the over the lazy the lazy brown lazy brown dog And the word-level trigram "the quick red" has the following character-level trigrams (where an underscore "_" marks a space): the he_ e_q _qu qui uic ick ck_ k_r _re red
Read more →
Tesla Dojo

Tesla Dojo is a series of supercomputers designed and built by Tesla for computer vision video processing and recognition. It was used for training Tesla's machine learning models to improve its Full Self-Driving (FSD) advanced driver-assistance system. It went into production in July 2023. Dojo's goal was to efficiently process millions of terabytes of video data captured from real-life driving situations from Tesla's 4+ million cars. This goal led to a considerably different architecture than conventional supercomputer designs. In August 2025, Bloomberg News reported that the Dojo project had been disbanded, though it was restarted in January 2026. == History == Tesla operates several massively parallel computing clusters for developing its Autopilot advanced driver assistance system. Its primary unnamed cluster using 5,760 Nvidia A100 graphics processing units (GPUs) was touted by Andrej Karpathy in 2021 at the fourth International Joint Conference on Computer Vision and Pattern Recognition (CCVPR 2021) to be "roughly the number five supercomputer in the world" at approximately 81.6 petaflops, based on scaling the performance of the Nvidia Selene supercomputer, which uses similar components. However, the performance of the primary Tesla GPU cluster has been disputed, as it was not clear if this was measured using single-precision or double-precision floating point numbers (FP32 or FP64). Tesla also operates a second 4,032 GPU cluster for training and a third 1,752 GPU cluster for automatic labeling of objects. The primary unnamed Tesla GPU cluster has been used for processing one million video clips, each ten seconds long, taken from Tesla Autopilot cameras operating in Tesla cars in the real world, running at 36 frames per second. Collectively, these video clips contained six billion object labels, with depth and velocity data; the total size of the data set was 1.5 petabytes. This data set was used for training a neural network intended to help Autopilot computers in Tesla cars understand roads. By August 2022, Tesla had upgraded the primary GPU cluster to 7,360 GPUs. Dojo was first mentioned by Elon Musk in April 2019 during Tesla's "Autonomy Investor Day". In August 2020, Musk stated it was "about a year away" due to power and thermal issues. Dojo was officially announced at Tesla's Artificial Intelligence (AI) Day on August 19, 2021. Tesla revealed details of the D1 chip and its plans for "Project Dojo", a datacenter that would house 3,000 D1 chips; the first "Training Tile" had been completed and delivered the week before. In October 2021, Tesla released a "Dojo Technology" whitepaper describing the Configurable Float8 (CFloat8) and Configurable Float16 (CFloat16) floating point formats and arithmetic operations as an extension of Institute of Electrical and Electronics Engineers (IEEE) standard 754. At the follow-up AI Day in September 2022, Tesla announced it had built several System Trays and one Cabinet. During a test, the company stated that Project Dojo drew 2.3 megawatts (MW) of power before tripping a local San Jose, California power substation. At the time, Tesla was assembling one Training Tile per day. In August 2023, Tesla powered on Dojo for production use as well as a new training cluster configured with 10,000 Nvidia H100 GPUs. In January 2024, Musk described Dojo as "a long shot worth taking because the payoff is potentially very high. But it's not something that is a high probability." In June 2024, Musk explained that ongoing construction work at Gigafactory Texas is for a computing cluster claiming that it is planned to comprise an even mix of "Tesla AI" and Nvidia/other hardware with a total thermal design power of at first 130 MW and eventually exceeding 500 MW. In August 2025, Bloomberg News reported that the Dojo project was disbanded, though Musk announced it would be restarted in January 2026 with a new chip iteration. == Technical architecture == The fundamental unit of the Dojo supercomputer is the D1 chip, designed by a team at Tesla led by ex-AMD CPU designer Ganesh Venkataramanan, including Emil Talpes, Debjit Das Sarma, Douglas Williams, Bill Chang, and Rajiv Kurian. The D1 chip is manufactured by the Taiwan Semiconductor Manufacturing Company (TSMC) using 7 nanometer (nm) semiconductor nodes, has 50 billion transistors and a large die size of 645 mm2 (1.0 square inch). Updating at Artificial Intelligence (AI) Day in 2022, Tesla announced that Dojo would scale by deploying multiple ExaPODs, in which there would be: 10 Cabinets per ExaPOD (1,062,000 cores, 3,000 D1 chips) 2 System Trays per Cabinet (106,200 cores, 300 D1 chips) 6 Training Tiles per System Tray (53,100 cores, along with host interface hardware) 25 D1 chips per Training Tile (8,850 cores) 354 computing cores per D1 chip According to Venkataramanan, Tesla's senior director of Autopilot hardware, Dojo will have more than an exaflop (a million teraflops) of computing power. For comparison, according to Nvidia, in August 2021, the (pre-Dojo) Tesla AI-training center used 720 nodes, each with eight Nvidia A100 Tensor Core GPUs for 5,760 GPUs in total, providing up to 1.8 exaflops of performance. === D1 chip === Each node (computing core) of the D1 processing chip is a general purpose 64-bit CPU with a superscalar core. It supports internal instruction-level parallelism, and includes simultaneous multithreading (SMT). It doesn't support virtual memory and uses limited memory protection mechanisms. Dojo software/applications manage chip resources. The D1 instruction set supports both 64-bit scalar and 64-byte single instruction, multiple data (SIMD) vector instructions. The integer unit mixes reduced instruction set computer (RISC-V) and custom instructions, supporting 8, 16, 32, or 64 bit integers. The custom vector math unit is optimized for machine learning kernels and supports multiple data formats, with a mix of precisions and numerical ranges, many of which are compiler composable. Up to 16 vector formats can be used simultaneously. ==== Node ==== Each D1 node uses a 32-byte fetch window holding up to eight instructions. These instructions are fed to an eight-wide decoder which supports two threads per cycle, followed by a four-wide, four-way SMT scalar scheduler that has two integer units, two address units, and one register file per thread. Vector instructions are passed further down the pipeline to a dedicated vector scheduler with two-way SMT, which feeds either a 64-byte SIMD unit or four 8×8×4 matrix multiplication units. The network on-chip (NOC) router links cores into a two-dimensional mesh network. It can send one packet in and one packet out in all four directions to/from each neighbor node, along with one 64-byte read and one 64-byte write to local SRAM per clock cycle. Hardware native operations transfer data, semaphores and barrier constraints across memories and CPUs. System-wide double data rate 4 (DDR4) synchronous dynamic random-access memory (SDRAM) memory works like bulk storage. ==== Memory ==== Each core has a 1.25 megabytes (MB) of SRAM main memory. Load and store speeds reach 400 gigabytes (GB) per second and 270 GB/sec, respectively. The chip has explicit core-to-core data transfer instructions. Each SRAM has a unique list parser that feeds a pair of decoders and a gather engine that feeds the vector register file, which together can directly transfer information across nodes. ==== Die ==== Twelve nodes (cores) are grouped into a local block. Nodes are arranged in an 18×20 array on a single die, of which 354 cores are available for applications. The die runs at 2 gigahertz (GHz) and totals 440 MB of SRAM (360 cores × 1.25 MB/core). It reaches 376 teraflops using 16-bit brain floating point (BF16) numbers or using configurable 8-bit floating point (CFloat8) numbers, which is a Tesla proposal, and 22 teraflops at FP32. Each die comprises 576 bi-directional serializer/deserializer (SerDes) channels along the perimeter to link to other dies, and moves 8 TB/sec across all four die edges. Each D1 chip has a thermal design power of approximately 400 watts. === Training Tile === The water-cooled Training Tile packages 25 D1 chips into a 5×5 array. Each tile supports 36 TB/sec of aggregate bandwidth via 40 input/output (I/O) chips - half the bandwidth of the chip mesh network. Each tile supports 10 TB/sec of on-tile bandwidth. Each tile has 11 GB of SRAM memory (25 D1 chips × 360 cores/D1 × 1.25 MB/core). Each tile achieves 9 petaflops at BF16/CFloat8 precision (25 D1 chips × 376 TFLOP/D1). Each tile consumes 15 kilowatts; 288 amperes at 52 volts. === System Tray === Six tiles are aggregated into a System Tray, which is integrated with a host interface. Each host interface includes 512 x86 cores, providing a Linux-based user environment. Previously, the Dojo System Tray was known as the Training Matrix, which includes six Training Tiles, 20 Dojo Interface Processor cards across four host servers, and Ethernet-l
Read more →
Vector database

A vector database, vector store or vector search engine is a database that stores and retrieves embeddings of data in vector space. Vector databases typically implement approximate nearest neighbor algorithms so users can search for records semantically similar to a given input, unlike traditional databases which primarily look up records by exact match. Use-cases for vector databases include similarity search, semantic search, multi-modal search, recommendations engines, object detection, and retrieval-augmented generation (RAG). Vector embeddings are mathematical representations of data in a high-dimensional space. In this space, each dimension corresponds to a feature of the data, with the number of dimensions ranging from a few hundred to tens of thousands, depending on the complexity of the data being represented. Each data item is represented by one vector in this space. Words, phrases, or entire documents, as well as images, audio, and other types of data, can all be vectorized. These feature vectors may be computed from the raw data using machine learning methods such as feature extraction algorithms, word embeddings or deep learning networks. The goal is that semantically similar data items receive feature vectors close to each other. Vector retrieval can be combined with metadata filtering or lexical search to support filtered and hybrid retrieval workflows. == Techniques == Common techniques for similarity search on high-dimensional vectors include: Hierarchical Navigable Small World (HNSW) graphs Locality-sensitive hashing (LSH) and sketching Product quantization (PQ) Inverted files These techniques may also be combined in vector search systems. In recent benchmarks, HNSW-based implementations have been among the best performers. Conferences such as the International Conference on Similarity Search and Applications (SISAP) and the Conference on Neural Information Processing Systems (NeurIPS) have hosted competitions on vector search in large databases. == Applications == Vector databases are used in a wide range of machine learning applications including similarity search, semantic search, multi-modal search, recommendations engines, object detection, and retrieval-augmented generation. === Retrieval-augmented generation === An especially common use-case for vector databases is in retrieval-augmented generation (RAG), a method to improve domain-specific responses of large language models. The retrieval component of a RAG can be any search system, but is most often implemented as a vector database. Text documents describing the domain of interest are collected, and for each document or document section, a feature vector (known as an "embedding") is computed, typically using a deep learning network, and stored in a vector database along with a link to the document. Given a user prompt, the feature vector of the prompt is computed, and the database is queried to retrieve the most relevant documents. These are then automatically added into the context window of the large language model, and the large language model proceeds to create a response to the prompt given this context. == Implementations ==
Read more →
Jive (software)

Jive (formerly known as Clearspace, then Jive SBS, then Jive Engage) is a commercial Java EE-based Enterprise 2.0 collaboration and knowledge management tool produced by Jive Software. It was first released as "Clearspace" in 2006, then renamed SBS (for "Social Business Software") in March 2009, then renamed "Jive Engage" in 2011, and renamed simply to "Jive" in 2012. Jive integrates the functionality of online communities, microblogging, social networking, discussion forums, blogs, wikis, and IM under one unified user interface. Content placed into any of the systems (blog, wiki, documentation, etc.) can be found through a common search interface. Other features include RSS capability, email integration, a reputation and reward system for participation, personal user profiles, JAX-WS web service interoperability, and integration with the Spring Framework. The product is a pure-Java server-side web application and will run on any platform where Java (JDK 1.5 or higher) is installed. It does not require a dedicated server - users have reported successful deployment in both shared environments and multiple machine clusters. As of Jive 8, released March 30, 2015, there is a Jive-n version which is for internal use (hosted by the consumer or hosted by Jive as a service) and a Jive-x version which is an external version hosted as a service. Jive no longer supports wiki markup language. == Server requirements for Jive 8-n == The following are the server requirements for Jive 8-n Operating systems: RHEL version 6 or 7 for x86_64, CentOS version 6 or 7 for x86_64 or SuSE Enterprise Linux Server (SLES) 11 and 12 for x86_64 Application Servers: Jive ships with its own embedded Apache HTTPD and Tomcat servers as part of the install package. It is not possible to deploy the application onto other appservers. Databases: MySQL (5.1, 5.5, 5.6) Oracle (11gR2, 12c) Postgres (9.0, 9.1, 9.2, 9.3, 9.4 - 9.2 or higher recommended) Microsoft SQL Server (2008R2, 2012, 2014) Environment: Jive recommends a server with at least 4GB of RAM and a dual-core 2 GHz processor with x86_64 architecture The product integrates with an LDAP repository or Active Directory For optimal deployment with a large community Jive Software recommends: using dedicated cache and document-conversion servers hosting the application and database servers separately == Releases == Jive 8, released on March 30, 2015 Jive 7, released in October 2013 Jive 9.0.x, released in November 2016 Jive 9, released in November 2016, supported now
Read more →
Latent semantic analysis

Latent semantic analysis (LSA) is a technique in natural language processing, in particular distributional semantics, of analyzing relationships between a set of documents and the terms they contain by producing a set of concepts related to the documents and terms. LSA assumes that words that are close in meaning will occur in similar pieces of text (the distributional hypothesis). A matrix containing word counts per document (rows represent unique words and columns represent each document) is constructed from a large piece of text and a mathematical technique called singular value decomposition (SVD) is used to reduce the number of rows while preserving the similarity structure among columns. Documents are then compared by cosine similarity between any two columns. Values close to 1 represent very similar documents while values close to 0 represent very dissimilar documents. An information retrieval technique using latent semantic structure was patented in 1988 by Scott Deerwester, Susan Dumais, George Furnas, Richard Harshman, Thomas Landauer, Karen Lochbaum and Lynn Streeter. In the context of its application to information retrieval, it is sometimes called latent semantic indexing (LSI). == Overview == === Occurrence matrix === LSA can use a document-term matrix which describes the occurrences of terms in documents; it is a sparse matrix whose rows correspond to terms and whose columns correspond to documents. A typical example of the weighting of the elements of the matrix is tf-idf (term frequency–inverse document frequency): the weight of an element of the matrix is proportional to the number of times the terms appear in each document, where rare terms are upweighted to reflect their relative importance. This matrix is also common to standard semantic models, though it is not necessarily explicitly expressed as a matrix, since the mathematical properties of matrices are not always used. === Rank lowering === After the construction of the occurrence matrix, LSA finds a low-rank approximation to the term-document matrix. There could be various reasons for these approximations: The original term-document matrix is presumed too large for the computing resources; in this case, the approximated low rank matrix is interpreted as an approximation (a "least and necessary evil"). The original term-document matrix is presumed noisy: for example, anecdotal instances of terms are to be eliminated. From this point of view, the approximated matrix is interpreted as a de-noisified matrix (a better matrix than the original). The original term-document matrix is presumed overly sparse relative to the "true" term-document matrix. That is, the original matrix lists only the words actually in each document, whereas we might be interested in all words related to each document—generally a much larger set due to synonymy. The consequence of the rank lowering is that some dimensions are combined and depend on more than one term: {(car), (truck), (flower)} → {(1.3452 car + 0.2828 truck), (flower)} This mitigates the problem of identifying synonymy, as the rank lowering is expected to merge the dimensions associated with terms that have similar meanings. It also partially mitigates the problem with polysemy, since components of polysemous words that point in the "right" direction are added to the components of words that share a similar meaning. Conversely, components that point in other directions tend to either simply cancel out, or, at worst, to be smaller than components in the directions corresponding to the intended sense. === Derivation === Let X {\displaystyle X} be a matrix where element ( i , j ) {\displaystyle (i,j)} describes the occurrence of term i {\displaystyle i} in document j {\displaystyle j} (this can be, for example, the frequency). X {\displaystyle X} will look like this: d j ↓ t i T → [ x 1 , 1 … x 1 , j … x 1 , n ⋮ ⋱ ⋮ ⋱ ⋮ x i , 1 … x i , j … x i , n ⋮ ⋱ ⋮ ⋱ ⋮ x m , 1 … x m , j … x m , n ] {\displaystyle {\begin{matrix}&{\textbf {d}}_{j}\\&\downarrow \\{\textbf {t}}_{i}^{T}\rightarrow &{\begin{bmatrix}x_{1,1}&\dots &x_{1,j}&\dots &x_{1,n}\\\vdots &\ddots &\vdots &\ddots &\vdots \\x_{i,1}&\dots &x_{i,j}&\dots &x_{i,n}\\\vdots &\ddots &\vdots &\ddots &\vdots \\x_{m,1}&\dots &x_{m,j}&\dots &x_{m,n}\\\end{bmatrix}}\end{matrix}}} Now a row in this matrix will be a vector corresponding to a term, giving its relation to each document: t i T = [ x i , 1 … x i , j … x i , n ] {\displaystyle {\textbf {t}}_{i}^{T}={\begin{bmatrix}x_{i,1}&\dots &x_{i,j}&\dots &x_{i,n}\end{bmatrix}}} Likewise, a column in this matrix will be a vector corresponding to a document, giving its relation to each term: d j = [ x 1 , j ⋮ x i , j ⋮ x m , j ] {\displaystyle {\textbf {d}}_{j}={\begin{bmatrix}x_{1,j}\\\vdots \\x_{i,j}\\\vdots \\x_{m,j}\\\end{bmatrix}}} Now the dot product t i T t p {\displaystyle {\textbf {t}}_{i}^{T}{\textbf {t}}_{p}} between two term vectors gives the correlation between the terms over the set of documents. The matrix product X X T {\displaystyle XX^{T}} contains all these dot products. Element ( i , p ) {\displaystyle (i,p)} (which is equal to element ( p , i ) {\displaystyle (p,i)} ) contains the dot product t i T t p {\displaystyle {\textbf {t}}_{i}^{T}{\textbf {t}}_{p}} ( = t p T t i {\displaystyle ={\textbf {t}}_{p}^{T}{\textbf {t}}_{i}} ). Likewise, the matrix X T X {\displaystyle X^{T}X} contains the dot products between all the document vectors, giving their correlation over the terms: d j T d q = d q T d j {\displaystyle {\textbf {d}}_{j}^{T}{\textbf {d}}_{q}={\textbf {d}}_{q}^{T}{\textbf {d}}_{j}} . Now, from the theory of linear algebra, there exists a decomposition of X {\displaystyle X} such that U {\displaystyle U} and V {\displaystyle V} are orthogonal matrices and Σ {\displaystyle \Sigma } is a diagonal matrix. This is called a singular value decomposition (SVD): X = U Σ V T {\displaystyle {\begin{matrix}X=U\Sigma V^{T}\end{matrix}}} The matrix products giving us the term and document correlations then become X X T = ( U Σ V T ) ( U Σ V T ) T = ( U Σ V T ) ( V T T Σ T U T ) = U Σ V T V Σ T U T = U Σ Σ T U T X T X = ( U Σ V T ) T ( U Σ V T ) = ( V T T Σ T U T ) ( U Σ V T ) = V Σ T U T U Σ V T = V Σ T Σ V T {\displaystyle {\begin{matrix}XX^{T}&=&(U\Sigma V^{T})(U\Sigma V^{T})^{T}=(U\Sigma V^{T})(V^{T^{T}}\Sigma ^{T}U^{T})=U\Sigma V^{T}V\Sigma ^{T}U^{T}=U\Sigma \Sigma ^{T}U^{T}\\X^{T}X&=&(U\Sigma V^{T})^{T}(U\Sigma V^{T})=(V^{T^{T}}\Sigma ^{T}U^{T})(U\Sigma V^{T})=V\Sigma ^{T}U^{T}U\Sigma V^{T}=V\Sigma ^{T}\Sigma V^{T}\end{matrix}}} Since Σ Σ T {\displaystyle \Sigma \Sigma ^{T}} and Σ T Σ {\displaystyle \Sigma ^{T}\Sigma } are diagonal we see that U {\displaystyle U} must contain the eigenvectors of X X T {\displaystyle XX^{T}} , while V {\displaystyle V} must be the eigenvectors of X T X {\displaystyle X^{T}X} . Both products have the same non-zero eigenvalues, given by the non-zero entries of Σ Σ T {\displaystyle \Sigma \Sigma ^{T}} , or equally, by the non-zero entries of Σ T Σ {\displaystyle \Sigma ^{T}\Sigma } . Now the decomposition looks like this: X U Σ V T ( d j ) ( d ^ j ) ↓ ↓ ( t i T ) → [ x 1 , 1 … x 1 , j … x 1 , n ⋮ ⋱ ⋮ ⋱ ⋮ x i , 1 … x i , j … x i , n ⋮ ⋱ ⋮ ⋱ ⋮ x m , 1 … x m , j … x m , n ] = ( t ^ i T ) → [ [ u 1 ] … [ u l ] ] ⋅ [ σ 1 … 0 ⋮ ⋱ ⋮ 0 … σ l ] ⋅ [ [ v 1 ] ⋮ [ v l ] ] {\displaystyle {\begin{matrix}&X&&&U&&\Sigma &&V^{T}\\&({\textbf {d}}_{j})&&&&&&&({\hat {\textbf {d}}}_{j})\\&\downarrow &&&&&&&\downarrow \\({\textbf {t}}_{i}^{T})\rightarrow &{\begin{bmatrix}x_{1,1}&\dots &x_{1,j}&\dots &x_{1,n}\\\vdots &\ddots &\vdots &\ddots &\vdots \\x_{i,1}&\dots &x_{i,j}&\dots &x_{i,n}\\\vdots &\ddots &\vdots &\ddots &\vdots \\x_{m,1}&\dots &x_{m,j}&\dots &x_{m,n}\\\end{bmatrix}}&=&({\hat {\textbf {t}}}_{i}^{T})\rightarrow &{\begin{bmatrix}{\begin{bmatrix}\,\\\,\\{\textbf {u}}_{1}\\\,\\\,\end{bmatrix}}\dots {\begin{bmatrix}\,\\\,\\{\textbf {u}}_{l}\\\,\\\,\end{bmatrix}}\end{bmatrix}}&\cdot &{\begin{bmatrix}\sigma _{1}&\dots &0\\\vdots &\ddots &\vdots \\0&\dots &\sigma _{l}\\\end{bmatrix}}&\cdot &{\begin{bmatrix}{\begin{bmatrix}&&{\textbf {v}}_{1}&&\end{bmatrix}}\\\vdots \\{\begin{bmatrix}&&{\textbf {v}}_{l}&&\end{bmatrix}}\end{bmatrix}}\end{matrix}}} The values σ 1 , … , σ l {\displaystyle \sigma _{1},\dots ,\sigma _{l}} are called the singular values, and u 1 , … , u l {\displaystyle u_{1},\dots ,u_{l}} and v 1 , … , v l {\displaystyle v_{1},\dots ,v_{l}} the left and right singular vectors. Notice the only part of U {\displaystyle U} that contributes to t i {\displaystyle {\textbf {t}}_{i}} is the i 'th {\displaystyle i{\textrm {'th}}} row. Let this row vector be called t ^ i T {\displaystyle {\hat {\textrm {t}}}_{i}^{T}} . Likewise, the only part of V T {\displaystyle V^{T}} that contributes to d j {\displaystyle {\textbf {d}}_{j}} is the j 'th {\displaystyle j{\textrm {'th}}} column, d ^ j {\displaystyle {\hat {\textrm {d}}}_{j}} . These are not the eigenvectors, but depend on all the eigenvectors. I
Read more →
Spleak

Spleak was an IM platform where users could publish and rate content. It existed in the form of six bots covering as many subject areas: CelebSpleak, SportSpleak, VoteSpleak, TVSpleak, GameSpleak, and StyleSpleak. == Overview == Users can add a "multi-Spleak" (which contains all of the different Spleak bots in one) or add the separate bots to their IM buddy lists on MSN and AIM. Users are also allowed access to Spleak online by using a CelebSpleak, SportSpleak, or VoteSpleak widget, or through the CelebSpleak and SportSpleak applications with Facebook. Spleak was an alternate reality game and is moving to its own company, Spleak Media Network. "Celebrate Spleak" was introduced throughout 2007, launched in 2008, and was forced to retire in 2009. == Key people == Spleak was co-founded by Morten Lund and Nicolaj Reffstrup. The company's chief executive officer is Morrie Eisenburg; Josh Scott is Vice President in Product and Tyler Wells is Vice President in Engineering.
Read more →
Geometric hashing

In computer science, geometric hashing is a method for efficiently finding two-dimensional objects represented by discrete points that have undergone an affine transformation, though extensions exist to other object representations and transformations. In an off-line step, the objects are encoded by treating each pair of points as a geometric basis. The remaining points can be represented in an invariant fashion with respect to this basis using two parameters. For each point, its quantized transformed coordinates are stored in the hash table as a key, and indices of the basis points as a value. Then a new pair of basis points is selected, and the process is repeated. In the on-line (recognition) step, randomly selected pairs of data points are considered as candidate bases. For each candidate basis, the remaining data points are encoded according to the basis and possible correspondences from the object are found in the previously constructed table. The candidate basis is accepted if a sufficiently large number of the data points index a consistent object basis. Geometric hashing was originally suggested in computer vision for object recognition in 2D and 3D, but later was applied to different problems such as structural alignment of proteins. == Geometric hashing in computer vision == Geometric hashing is a method used for object recognition. Let’s say that we want to check if a model image can be seen in an input image. This can be accomplished with geometric hashing. The method could be used to recognize one of the multiple objects in a base, in this case the hash table should store not only the pose information but also the index of object model in the base. === Example === For simplicity, this example will not use too many point features and assume that their descriptors are given by their coordinates only (in practice local descriptors such as SIFT could be used for indexing). ==== Training Phase ==== Find the model's feature points. Assume that 5 feature points are found in the model image with the coordinates ( 12 , 17 ) ; {\displaystyle (12,17);} ( 45 , 13 ) ; {\displaystyle (45,13);} ( 40 , 46 ) ; {\displaystyle (40,46);} ( 20 , 35 ) ; {\displaystyle (20,35);} ( 35 , 25 ) {\displaystyle (35,25)} , see the picture. Introduce a basis to describe the locations of the feature points. For 2D space and similarity transformation the basis is defined by a pair of points. The point of origin is placed in the middle of the segment connecting the two points (P2, P4 in our example), the x ′ {\displaystyle x'} axis is directed towards one of them, the y ′ {\displaystyle y'} is orthogonal and goes through the origin. The scale is selected such that absolute value of x ′ {\displaystyle x'} for both basis points is 1. Describe feature locations with respect to that basis, i.e. compute the projections to the new coordinate axes. The coordinates should be discretised to make recognition robust to noise, we take the bin size 0.25. We thus get the coordinates ( − 0.75 , − 1.25 ) ; {\displaystyle (-0.75,-1.25);} ( 1.00 , 0.00 ) ; {\displaystyle (1.00,0.00);} ( − 0.50 , 1.25 ) ; {\displaystyle (-0.50,1.25);} ( − 1.00 , 0.00 ) ; {\displaystyle (-1.00,0.00);} ( 0.00 , 0.25 ) {\displaystyle (0.00,0.25)} Store the basis in a hash table indexed by the features (only transformed coordinates in this case). If there were more objects to match with, we should also store the object number along with the basis pair. Repeat the process for a different basis pair (Step 2). It is needed to handle occlusions. Ideally, all the non-colinear pairs should be enumerated. We provide the hash table after two iterations, the pair (P1, P3) is selected for the second one. Hash Table: Most hash tables cannot have identical keys mapped to different values. So in real life one won’t encode basis keys (1.0, 0.0) and (-1.0, 0.0) in a hash table. ==== Recognition Phase ==== Find interesting feature points in the input image. Choose an arbitrary basis. If there isn't a suitable arbitrary basis, then it is likely that the input image does not contain the target object. Describe coordinates of the feature points in the new basis. Quantize obtained coordinates as it was done before. Compare all the transformed point features in the input image with the hash table. If the point features are identical or similar, then increase the count for the corresponding basis (and the type of object, if any). For each basis such that the count exceeds a certain threshold, verify the hypothesis that it corresponds to an image basis chosen in Step 2. Transfer the image coordinate system to the model one (for the supposed object) and try to match them. If successful, the object is found. Otherwise, go back to Step 2. === Finding mirrored pattern === It seems that this method is only capable of handling scaling, translation, and rotation. However, the input image may contain the object in mirror transform. Therefore, geometric hashing should be able to find the object, too. There are two ways to detect mirrored objects. For the vector graph, make the left side positive, and the right side negative. Multiplying the x position by -1 will give the same result. Use 3 points for the basis. This allows detecting mirror images (or objects). Actually, using 3 points for the basis is another approach for geometric hashing. === Geometric hashing in higher-dimensions === Similar to the example above, hashing applies to higher-dimensional data. For three-dimensional data points, three points are also needed for the basis. The first two points define the x-axis, and the third point defines the y-axis (with the first point). The z-axis is perpendicular to the created axis using the right-hand rule. Notice that the order of the points affects the resulting basis
Read more →
Joox

Joox (stylised in all caps) is a music streaming service owned by Tencent, launched in January 2015. Joox is the biggest music streaming app in Asian markets such as Hong Kong, Macau, Indonesia, Malaysia, Myanmar, Thailand and also in South Africa before it was shut down in early 2022. Joox is a freemium service, providing most of its songs free, while some songs are only available for premium users, offered via paid subscriptions or by doing different tasks offered. In 2017, Joox launched their service in their first non-Asian market, South Africa, which for an unknown reason shut down five years later. The service now accounts for more than 50% of all music streaming app downloads in their Asian markets. The number of music-streaming users in Hong Kong, Macau, Malaysia, Thailand, Myanmar and Indonesia was expected to reach 87 million by 2020. == Background == Before the emergence of Joox, Tencent owned QQ Music, one of the largest music streaming and download service in China. In 2015, they introduced Joox as their expansion of music services to overseas market instead of mainland China, starting first in Hong Kong. Instead of providing free services by playing audio ads to users like Spotify, another major music service, Joox focused on banner ads, splash ads and other advertising methods such as category playlists and in-app skins. They claimed it as a success. Joox offered their premium VIP access to DStv subscribers free of charge. DStv is the sister company to Tencent and is the primary pay-TV provider in South Africa. In November 2021, it was announced that Joox will stop streaming in South Africa in March 2022.
Read more →
Neural network Gaussian process

A Neural Network Gaussian Process (NNGP) is a Gaussian process (GP) obtained as the limit of a certain type of sequence of neural networks. Specifically, a wide variety of network architectures converges to a GP in the infinitely wide limit, in the sense of distribution. The concept constitutes an intensional definition, i.e., a NNGP is just a GP, but distinguished by how it is obtained. == Motivation == Bayesian networks are a modeling tool for assigning probabilities to events, and thereby characterizing the uncertainty in a model's predictions. Deep learning and artificial neural networks are approaches used in machine learning to build computational models which learn from training examples. Bayesian neural networks merge these fields. They are a type of neural network whose parameters and predictions are both probabilistic. While standard neural networks often assign high confidence even to incorrect predictions, Bayesian neural networks can more accurately evaluate how likely their predictions are to be correct. Computation in artificial neural networks is usually organized into sequential layers of artificial neurons. The number of neurons in a layer is called the layer width. When we consider a sequence of Bayesian neural networks with increasingly wide layers (see figure), they converge in distribution to a NNGP. This large width limit is of practical interest, since the networks often improve as layers get wider. And the process may give a closed form way to evaluate networks. NNGPs also appears in several other contexts: It describes the distribution over predictions made by wide non-Bayesian artificial neural networks after random initialization of their parameters, but before training; it appears as a term in neural tangent kernel prediction equations; it is used in deep information propagation to characterize whether hyperparameters and architectures will be trainable. It is related to other large width limits of neural networks. === Scope === The first correspondence result had been established in the 1995 PhD thesis of Radford M. Neal, then supervised by Geoffrey Hinton at University of Toronto. Neal cites David J. C. MacKay as inspiration, who worked in Bayesian learning. Today the correspondence is proven for: Single hidden layer Bayesian neural networks; deep fully connected networks as the number of units per layer is taken to infinity; convolutional neural networks as the number of channels is taken to infinity; transformer networks as the number of attention heads is taken to infinity; recurrent networks as the number of units is taken to infinity. In fact, this NNGP correspondence holds for almost any architecture: Generally, if an architecture can be expressed solely via matrix multiplication and coordinatewise nonlinearities (i.e., a tensor program), then it has an infinite-width GP. This in particular includes all feedforward or recurrent neural networks composed of multilayer perceptron, recurrent neural networks (e.g., LSTMs, GRUs), (nD or graph) convolution, pooling, skip connection, attention, batch normalization, and/or layer normalization. === Illustration === Every setting of a neural network's parameters θ {\displaystyle \theta } corresponds to a specific function computed by the neural network. A prior distribution p ( θ ) {\displaystyle p(\theta )} over neural network parameters therefore corresponds to a prior distribution over functions computed by the network. As neural networks are made infinitely wide, this distribution over functions converges to a Gaussian process for many architectures. The notation used in this section is the same as the notation used below to derive the correspondence between NNGPs and fully connected networks, and more details can be found there. The figure to the right plots the one-dimensional outputs z L ( ⋅ ; θ ) {\displaystyle z^{L}(\cdot ;\theta )} of a neural network for two inputs x {\displaystyle x} and x ∗ {\displaystyle x^{}} against each other. The black dots show the function computed by the neural network on these inputs for random draws of the parameters from p ( θ ) {\displaystyle p(\theta )} . The red lines are iso-probability contours for the joint distribution over network outputs z L ( x ; θ ) {\displaystyle z^{L}(x;\theta )} and z L ( x ∗ ; θ ) {\displaystyle z^{L}(x^{};\theta )} induced by p ( θ ) {\displaystyle p(\theta )} . This is the distribution in function space corresponding to the distribution p ( θ ) {\displaystyle p(\theta )} in parameter space, and the black dots are samples from this distribution. For infinitely wide neural networks, since the distribution over functions computed by the neural network is a Gaussian process, the joint distribution over network outputs is a multivariate Gaussian for any finite set of network inputs. == Discussion == === Infinitely wide fully connected network === This section expands on the correspondence between infinitely wide neural networks and Gaussian processes for the specific case of a fully connected architecture. It provides a proof sketch outlining why the correspondence holds, and introduces the specific functional form of the NNGP for fully connected networks. The proof sketch closely follows the approach by Novak and coauthors. ==== Network architecture specification ==== Consider a fully connected artificial neural network with inputs x {\displaystyle x} , parameters θ {\displaystyle \theta } consisting of weights W l {\displaystyle W^{l}} and biases b l {\displaystyle b^{l}} for each layer l {\displaystyle l} in the network, pre-activations (pre-nonlinearity) z l {\displaystyle z^{l}} , activations (post-nonlinearity) y l {\displaystyle y^{l}} , pointwise nonlinearity ϕ ( ⋅ ) {\displaystyle \phi (\cdot )} , and layer widths n l {\displaystyle n^{l}} . For simplicity, the width n L + 1 {\displaystyle n^{L+1}} of the readout vector z L {\displaystyle z^{L}} is taken to be 1. The parameters of this network have a prior distribution p ( θ ) {\displaystyle p(\theta )} , which consists of an isotropic Gaussian for each weight and bias, with the variance of the weights scaled inversely with layer width. This network is illustrated in the figure to the right, and described by the following set of equations: x ≡ input y l ( x ) = { x l = 0 ϕ ( z l − 1 ( x ) ) l > 0 z i l ( x ) = ∑ j W i j l y j l ( x ) + b i l W i j l ∼ N ( 0 , σ w 2 n l ) b i l ∼ N ( 0 , σ b 2 ) ϕ ( ⋅ ) ≡ nonlinearity y l ( x ) , z l − 1 ( x ) ∈ R n l × 1 n L + 1 = 1 θ = { W 0 , b 0 , … , W L , b L } {\displaystyle {\begin{aligned}x&\equiv {\text{input}}\\y^{l}(x)&=\left\{{\begin{array}{lcl}x&&l=0\\\phi \left(z^{l-1}(x)\right)&&l>0\end{array}}\right.\\z_{i}^{l}(x)&=\sum _{j}W_{ij}^{l}y_{j}^{l}(x)+b_{i}^{l}\\W_{ij}^{l}&\sim {\mathcal {N}}\left(0,{\frac {\sigma _{w}^{2}}{n^{l}}}\right)\\b_{i}^{l}&\sim {\mathcal {N}}\left(0,\sigma _{b}^{2}\right)\\\phi (\cdot )&\equiv {\text{nonlinearity}}\\y^{l}(x),z^{l-1}(x)&\in \mathbb {R} ^{n^{l}\times 1}\\n^{L+1}&=1\\\theta &=\left\{W^{0},b^{0},\dots ,W^{L},b^{L}\right\}\end{aligned}}} ==== ==== z l | y l {\displaystyle z^{l}|y^{l}} is a Gaussian process We first observe that the pre-activations z l {\displaystyle z^{l}} are described by a Gaussian process conditioned on the preceding activations y l {\displaystyle y^{l}} . This result holds even at finite width. Each pre-activation z i l {\displaystyle z_{i}^{l}} is a weighted sum of Gaussian random variables, corresponding to the weights W i j l {\displaystyle W_{ij}^{l}} and biases b i l {\displaystyle b_{i}^{l}} , where the coefficients for each of those Gaussian variables are the preceding activations y j l {\displaystyle y_{j}^{l}} . Because they are a weighted sum of zero-mean Gaussians, the z i l {\displaystyle z_{i}^{l}} are themselves zero-mean Gaussians (conditioned on the coefficients y j l {\displaystyle y_{j}^{l}} ). Since the z l {\displaystyle z^{l}} are jointly Gaussian for any set of y l {\displaystyle y^{l}} , they are described by a Gaussian process conditioned on the preceding activations y l {\displaystyle y^{l}} . The covariance or kernel of this Gaussian process depends on the weight and bias variances σ w 2 {\displaystyle \sigma _{w}^{2}} and σ b 2 {\displaystyle \sigma _{b}^{2}} , as well as the second moment matrix K l {\displaystyle K^{l}} of the preceding activations y l {\displaystyle y^{l}} , z i l ∣ y l ∼ G P ( 0 , σ w 2 K l + σ b 2 ) K l ( x , x ′ ) = 1 n l ∑ i y i l ( x ) y i l ( x ′ ) {\displaystyle {\begin{aligned}z_{i}^{l}\mid y^{l}&\sim {\mathcal {GP}}\left(0,\sigma _{w}^{2}K^{l}+\sigma _{b}^{2}\right)\\K^{l}(x,x')&={\frac {1}{n^{l}}}\sum _{i}y_{i}^{l}(x)y_{i}^{l}(x')\end{aligned}}} The effect of the weight scale σ w 2 {\displaystyle \sigma _{w}^{2}} is to rescale the contribution to the covariance matrix from K l {\displaystyle K^{l}} , while the bias is shared for all inputs, and so σ b 2 {\displaystyle \sigma _{b}^{2}} makes the z i l {\displaystyle z_{i}^{l}} for different datapoints more similar and
Read more →
Controlled natural language

Controlled natural languages (CNLs) are subsets of natural languages that are obtained by restricting the grammar and vocabulary in order to reduce or eliminate ambiguity and complexity. Traditionally, controlled languages fall into two major types: those that improve readability for human readers (e.g. non-native speakers), and those that enable reliable automatic semantic analysis of the language. The first type of languages (often called "simplified" or "technical" languages), for example ASD Simplified Technical English, Caterpillar Technical English, IBM's Easy English, are used in the industry to increase the quality of technical documentation, and possibly simplify the semi-automatic translation of the documentation. These languages restrict the writer by general rules such as "Keep sentences short", "Avoid the use of pronouns", "Only use dictionary-approved words", and "Use only the active voice". The second type of languages have a formal syntax and formal semantics, and can be mapped to an existing formal language, such as first-order logic. Thus, those languages can be used as knowledge representation languages, and writing of those languages is supported by fully automatic consistency and redundancy checks, query answering, etc. == Languages == Existing controlled natural languages include: == Encoding == IETF has reserved simple as a BCP 47 variant subtag for simplified versions of languages.
Read more →