AI Coding Wiki

AI Coding Wiki — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

Mojito (framework)

Mojito is an environment agnostic, Model-View-Controller (MVC) web application framework. It was designed by Yahoo. == Features == Mojito supports agile development of web applications. Mojito has built-in support for unit testing, Internationalization, syntax and coding convention checks. Both server and client components are written in JavaScript. Mojito allows developers designing web applications to leverage the utilities of both configuration and MVC framework. Mojito is capable of running on both JavaScript-enabled web browsers and servers using Node.js because they both utilize JavaScript. Mojito applications mainly consist of two components: JSON Configuration files: these define relationships between code components, assets, routing paths, and framework defaults and are available at the application and mojit level. Directories: these reflect MVC architecture and are used to separate resources such as assets, libraries, middleware, etc. == Architecture == In Mojito, both server and "client" side scripting is done in JavaScript, allowing it to run on both client and server thereby breaking the "front-end back-end barrier." It has both client and server runtimes. === Server runtime === This block houses operations needed by server side components. Services include: Routing rules, HTTP Server, config loader and disk-based loader. === Client runtime === This block houses operations called upon while running client sides components. Services include local storage/cache access and JSON based /URL based loader === Core === Core function can be accessed on client or server. Services include Registry, Dispatcher, Front controller, Resource store. === Container === mojit object comes into the picture. This container also include the services used by mojits. API and Mojito services are the blocks which caters to services needed for execution of mojits. === API (Action Context) === Mojito services are a customizable service block. It offers mojits a range of services which might be needed by mojit to carry out certain actions. These services can be availed at both client and server side. Reusable services can be created and aggregated to the core here. == Mojits == Mojits are the modules of a Mojito application. An application consists of one or more mojits. A mojit encompasses a Model, Views and a Controller defined by JSON configuration files. It includes a View factory where views are created according to the model and a View cache that holds frequently requested views to aid performance. === Application Architecture === A Mojito application is a set of mojits facilitated by configurable JSON files which define the code for model, view and controller. This MVC structure works with API block and Mojito services, and can be deployed at both client and server side. While the application is deployed at client side, it can call server-side modules using binders. Binders are mojit codes that let mojits request services from each other. Mojit Proxy acts as an intermediary between binders and mojit's API (application context) block and other mojits. Controllers are command-issuing units of mojits. Models mirror the core logic and hold data. Applications can have multiple models. They can be centrally accessed from controllers. View files are created in accordance with controllers and models, and are marked-up before they are sent to users as output. === Application Directory Structure === Directory structure of a Mojito application with one mojit: [mojito_app]/ |-- application.json |-- assets/ | `-- favicon.icon |-- yui_modules/ | `-- .{affinity}.js |-- index.js |-- mojits/ | `-- [mojit_name | |-- assets/ | |-- yui_modules/ | | `-- .{affinity}.js | |-- binders/ | | `-- {view_name}.js | |-- controller.{affinity}.js | |-- defaults.json | |-- definition.json | |-- lang/ | | `-- {mojit_name}_{lang}.js | |-- models/ | | `-- {model_name}.{affinity}.js | |-- tests/ | | |-- yui_modules/ | | | `-- {module_name}.{affinity}-tests.js | | |-- controller.{affinity}-tests.js | | `-- models/ | | `-- {model_name}.{affinity}-tests.js | `-- views/ | |-- {view_name}.{view_engine}.html | `-- {view_name}.{device}.{view_engine}.html |-- package.json |-- routes.json (deprecated) |-- server.js == Model, View and Controller == The Model hosts data, which is accessed by the Controller and presented to the View. Controller also handles any client requests for data, in which case controller fetches data from the model and passes the data to the client. All three components are clustered in the mojit. Mojits are physically illustrated by directory structures and an application can have multiple mojits. Every mojit can have one controller, one or more views and zero or more models. === Model === The model it represents the application data and is independent of view or controller. Model contains code to manipulate the data. They are found in the models directory of each mojit. Functions include: Storing information for access by controller. Validation and error handling. Metadata required by the view === Controller === The controller acts like a connecting agent between model and view. It supplies input to Model and after fetching data from model, passes it to View. Functions include Redirection Monitors authentication Web safety Encoding === View === The view acts as a presentation filter by highlighting some model attributes and suppressing others. A view can be understood as a visual permutation of the model. The view renders data received from controller and displays it to the end user.
Read more →
Information extraction

Information extraction (IE) is the task of automatically extracting structured information from unstructured and/or semi-structured machine-readable documents and other electronically represented sources. Typically, this involves processing human language texts by means of natural language processing (NLP). Recent activities in multimedia document processing like automatic annotation and content extraction out of images/audio/video/documents could be seen as information extraction. Recent advances in NLP techniques have allowed for significantly improved performance compared to previous years. An example is the extraction from newswire reports of corporate mergers, such as denoted by the formal relation: MergerBetween ⁡ ( c o m p a n y 1 , c o m p a n y 2 , d a t e ) {\displaystyle \operatorname {MergerBetween} (\mathrm {company} _{1},\mathrm {company} _{2},\mathrm {date} )} , from an online news sentence such as: "Yesterday, New York based Foo Inc. announced their acquisition of Bar Corp." A broad goal of IE is to allow computation to be done on the previously unstructured data. A more specific goal is to allow automated reasoning about the logical form of the input data. Structured data is semantically well-defined data from a chosen target domain, interpreted with respect to category and context. Information extraction is the part of a greater puzzle which deals with the problem of devising automatic methods for text management, beyond its transmission, storage and display. The discipline of information retrieval (IR) has developed automatic methods, typically of a statistical flavor, for indexing large document collections and classifying documents. Another complementary approach is that of natural language processing (NLP) which has solved the problem of modelling human language processing with considerable success when taking into account the magnitude of the task. In terms of both difficulty and emphasis, IE deals with tasks in between both IR and NLP. In terms of input, IE assumes the existence of a set of documents in which each document follows a template, i.e. describes one or more entities or events in a manner that is similar to those in other documents but differing in the details. An example, consider a group of newswire articles on Latin American terrorism with each article presumed to be based upon one or more terroristic acts. We also define for any given IE task a template, which is a(or a set of) case frame(s) to hold the information contained in a single document. For the terrorism example, a template would have slots corresponding to the perpetrator, victim, and weapon of the terroristic act, and the date on which the event happened. An IE system for this problem is required to "understand" an attack article only enough to find data corresponding to the slots in this template. == History == Information extraction dates back to the late 1970s in the early days of NLP. An early commercial system from the mid-1980s was JASPER built for Reuters by the Carnegie Group Inc with the aim of providing real-time financial news to financial traders. Beginning in 1987, IE was spurred by a series of Message Understanding Conferences. MUC is a competition-based conference that focused on the following domains: MUC-1 (1987), MUC-3 (1989): Naval operations messages. MUC-3 (1991), MUC-4 (1992): Terrorism in Latin American countries. MUC-5 (1993): Joint ventures and microelectronics domain. MUC-6 (1995): News articles on management changes. MUC-7 (1998): Satellite launch reports. Considerable support came from the U.S. Defense Advanced Research Projects Agency (DARPA), who wished to automate mundane tasks performed by government analysts, such as scanning newspapers for possible links to terrorism. == Present significance == The present significance of IE pertains to the growing amount of information available in unstructured form. Tim Berners-Lee, inventor of the World Wide Web, refers to the existing Internet as the web of documents and advocates that more of the content be made available as a web of data. Until this transpires, the web largely consists of unstructured documents lacking semantic metadata. Knowledge contained within these documents can be made more accessible for machine processing by means of transformation into relational form, or by marking-up with XML tags. An intelligent agent monitoring a news data feed requires IE to transform unstructured data into something that can be reasoned with. A typical application of IE is to scan a set of documents written in a natural language and populate a database with the information extracted. == Tasks and subtasks == Applying information extraction to text is linked to the problem of text simplification in order to create a structured view of the information present in free text. The overall goal being to create a more easily machine-readable text to process the sentences. Typical IE tasks and subtasks include: Template filling: Extracting a fixed set of fields from a document, e.g. extract perpetrators, victims, time, etc. from a newspaper article about a terrorist attack. Event extraction: Given an input document, output zero or more event templates. For instance, a newspaper article might describe multiple terrorist attacks. Knowledge Base Population: Fill a database of facts given a set of documents. Typically the database is in the form of triplets, (entity 1, relation, entity 2), e.g. (Barack Obama, Spouse, Michelle Obama) Named entity recognition: recognition of known entity names (for people and organizations), place names, temporal expressions, and certain types of numerical expressions, by employing existing knowledge of the domain or information extracted from other sentences. Typically the recognition task involves assigning a unique identifier to the extracted entity. A simpler task is named entity detection, which aims at detecting entities without having any existing knowledge about the entity instances. For example, in processing the sentence "M. Smith likes fishing", named entity detection would denote detecting that the phrase "M. Smith" does refer to a person, but without necessarily having (or using) any knowledge about a certain M. Smith who is (or, "might be") the specific person whom that sentence is talking about. Coreference resolution: detection of coreference and anaphoric links between text entities. In IE tasks, this is typically restricted to finding links between previously extracted named entities. For example, "International Business Machines" and "IBM" refer to the same real-world entity. If we take the two sentences "M. Smith likes fishing. But he doesn't like biking", it would be beneficial to detect that "he" is referring to the previously detected person "M. Smith". Relationship extraction: identification of relations between entities, such as: PERSON works for ORGANIZATION (extracted from the sentence "Bill works for IBM.") PERSON located in LOCATION (extracted from the sentence "Bill is in France.") Semi-structured information extraction which may refer to any IE that tries to restore some kind of information structure that has been lost through publication, such as: Table extraction: finding and extracting tables from documents. Table information extraction : extracting information in structured manner from the tables. This task is more complex than table extraction, as table extraction is only the first step, while understanding the roles of the cells, rows, columns, linking the information inside the table and understanding the information presented in the table are additional tasks necessary for table information extraction. Comments extraction : extracting comments from the actual content of articles in order to restore the link between authors of each of the sentences Language and vocabulary analysis Terminology extraction: finding the relevant terms for a given corpus Audio extraction Template-based music extraction: finding relevant characteristic in an audio signal taken from a given repertoire; for instance time indexes of occurrences of percussive sounds can be extracted in order to represent the essential rhythmic component of a music piece. Note that this list is not exhaustive and that the exact meaning of IE activities is not commonly accepted and that many approaches combine multiple sub-tasks of IE in order to achieve a wider goal. Machine learning, statistical analysis and/or natural language processing are often used in IE. IE on non-text documents is becoming an increasingly interesting topic in research, and information extracted from multimedia documents can now be expressed in a high level structure as it is done on text. This naturally leads to the fusion of extracted information from multiple kinds of documents and sources. == World Wide Web applications == IE has been the focus of the MUC conferences. The proliferation of the Web, however, intensified the need for developing IE systems that help people
Read more →
Rule-based machine translation

Rule-based machine translation (RBMT) is a classical approach of machine translation systems based on linguistic information about source and target languages. Such information is retrieved from (unilingual, bilingual or multilingual) dictionaries and grammars covering the main semantic, morphological, and syntactic regularities of each language. Having input sentences, an RBMT system generates output sentences on the basis of analysis of both the source and the target languages involved. RBMT has been progressively superseded by more efficient methods, particularly neural machine translation. == History == The first RBMT systems were developed in the early 1970s. The most important steps of this evolution were the emergence of the following RBMT systems: Systran Japanese MT systems Today, other common RBMT systems include: Apertium GramTrans == Types of RBMT == There are three different types of rule-based machine translation systems: Direct Systems (Dictionary Based Machine Translation) map input to output with basic rules. Transfer RBMT Systems (Transfer Based Machine Translation) employ morphological and syntactical analysis. Interlingual RBMT Systems (Interlingua) use an abstract meaning. RBMT systems can also be characterized as the systems opposite to Example-based Systems of Machine Translation (Example Based Machine Translation), whereas Hybrid Machine Translations Systems make use of many principles derived from RBMT. == Basic principles == The main approach of RBMT systems is based on linking the structure of the given input sentence with the structure of the demanded output sentence, necessarily preserving their unique meaning. The following example can illustrate the general frame of RBMT: A girl eats an apple. Source Language = English; Demanded Target Language = German Minimally, to get a German translation of this English sentence one needs: A dictionary that will map each English word to an appropriate German word. Rules representing regular English sentence structure. Rules representing regular German sentence structure. And finally, we need rules according to which one can relate these two structures together. Accordingly, we can state the following stages of translation: 1st: getting basic part-of-speech information of each source word: a = indef.article; girl = noun; eats = verb; an = indef.article; apple = noun 2nd: getting syntactic information about the verb "to eat": NP-eat-NP; here: eat – Present Simple, 3rd Person Singular, Active Voice 3rd: parsing the source sentence: (NP an apple) = the object of eat Often only partial parsing is sufficient to get to the syntactic structure of the source sentence and to map it onto the structure of the target sentence. 4th: translate English words into German a (category = indef.article) => ein (category = indef.article) girl (category = noun) => Mädchen (category = noun) eat (category = verb) => essen (category = verb) an (category = indef. article) => ein (category = indef.article) apple (category = noun) => Apfel (category = noun) 5th: Mapping dictionary entries into appropriate inflected forms (final generation): A girl eats an apple. => Ein Mädchen isst einen Apfel. == Ontologies == An ontology is a formal representation of knowledge that includes the concepts (such as objects, processes etc.) in a domain and some relations between them. If the stored information is of linguistic nature, one can speak of a lexicon. In NLP, ontologies can be used as a source of knowledge for machine translation systems. With access to a large knowledge base, rule-based systems can be enabled to resolve many (especially lexical) ambiguities on their own. In the following classic examples, as humans, we are able to interpret the prepositional phrase according to the context because we use our world knowledge, stored in our lexicons:I saw a man/star/molecule with a microscope/telescope/binoculars.Since the syntax does not change, a traditional rule-based machine translation system may not be able to differentiate between the meanings. With a large enough ontology as a source of knowledge however, the possible interpretations of ambiguous words in a specific context can be reduced. === Building ontologies === The ontology generated for the PANGLOSS knowledge-based machine translation system in 1993 may serve as an example of how an ontology for NLP purposes can be compiled: A large-scale ontology is necessary to help parsing in the active modules of the machine translation system. In the PANGLOSS example, about 50,000 nodes were intended to be subsumed under the smaller, manually-built upper (abstract) region of the ontology. Because of its size, it had to be created automatically. The goal was to merge the two resources LDOCE online and WordNet to combine the benefits of both: concise definitions from Longman, and semantic relations allowing for semi-automatic taxonomization to the ontology from WordNet. A definition match algorithm was created to automatically merge the correct meanings of ambiguous words between the two online resources, based on the words that the definitions of those meanings have in common in LDOCE and WordNet. Using a similarity matrix, the algorithm delivered matches between meanings including a confidence factor. This algorithm alone, however, did not match all meanings correctly on its own. A second hierarchy match algorithm was therefore created which uses the taxonomic hierarchies found in WordNet (deep hierarchies) and partially in LDOCE (flat hierarchies). This works by first matching unambiguous meanings, then limiting the search space to only the respective ancestors and descendants of those matched meanings. Thus, the algorithm matched locally unambiguous meanings (for instance, while the word seal as such is ambiguous, there is only one meaning of seal in the animal subhierarchy). Both algorithms complemented each other and helped constructing a large-scale ontology for the machine translation system. The WordNet hierarchies, coupled with the matching definitions of LDOCE, were subordinated to the ontology's upper region. As a result, the PANGLOSS MT system was able to make use of this knowledge base, mainly in its generation element. == Components == The RBMT system contains: a SL morphological analyser - analyses a source language word and provides the morphological information; a SL parser - is a syntax analyser which analyses source language sentences; a translator - used to translate a source language word into the target language; a TL morphological generator - works as a generator of appropriate target language words for the given grammatica information; a TL parser - works as a composer of suitable target language sentences; Several dictionaries - more specifically a minimum of three dictionaries: a SL dictionary - needed by the source language morphological analyser for morphological analysis, a bilingual dictionary - used by the translator to translate source language words into target language words, a TL dictionary - needed by the target language morphological generator to generate target language words. The RBMT system makes use of the following: a Source Grammar for the input language which builds syntactic constructions from input sentences; a Source Lexicon which captures all of the allowable vocabulary in the domain; Source Mapping Rules which indicate how syntactic heads and grammatical functions in the source language are mapped onto domain concepts and semantic roles in the interlingua; a Domain Model/Ontology which defines the classes of domain concepts and restricts the fillers of semantic roles for each class; Target Mapping Rules which indicate how domain concepts and semantic roles in the interlingua are mapped onto syntactic heads and grammatical functions in the target language; a Target Lexicon which contains appropriate target lexemes for each domain concept; a Target Grammar for the target language which realizes target syntactic constructions as linearized output sentences. == Advantages == No bilingual texts are required. This makes it possible to create translation systems for languages that have no texts in common, or even no digitized data whatsoever. Domain independent. Rules are usually written in a domain independent manner, so the vast majority of rules will "just work" in every domain, and only a few specific cases per domain may need rules written for them. No quality ceiling. Every error can be corrected with a targeted rule, even if the trigger case is extremely rare. This is in contrast to statistical systems where infrequent forms will be washed away by default. Total control. Because all rules are hand-written, you can easily debug a rule-based system to see exactly where a given error enters the system, and why. Reusability. Because RBMT systems are generally built from a strong source language analysis that is fed to a transfer step and target language generator, the source language analysis and targe
Read more →
Weight initialization

In deep learning, weight initialization or parameter initialization describes the initial step in creating a neural network. A neural network contains trainable parameters that are modified during training: weight initialization is the pre-training step of assigning initial values to these parameters. The choice of weight initialization method affects the speed of convergence, the scale of neural activation within the network, the scale of gradient signals during backpropagation, and the quality of the final model. Proper initialization is necessary for avoiding issues such as vanishing and exploding gradients and activation function saturation. Note that even though this article is titled "weight initialization", both weights and biases are used in a neural network as trainable parameters, so this article describes how both of these are initialized. Similarly, trainable parameters in convolutional neural networks (CNNs) are called kernels and biases, and this article also describes these. == Constant initialization == We discuss the main methods of initialization in the context of a multilayer perceptron (MLP). Specific strategies for initializing other network architectures are discussed in later sections. For an MLP, there are only two kinds of trainable parameters, called weights and biases. Each layer l {\displaystyle l} contains a weight matrix W ( l ) ∈ R n l − 1 × n l {\displaystyle W^{(l)}\in \mathbb {R} ^{n_{l-1}\times n_{l}}} and a bias vector b ( l ) ∈ R n l {\displaystyle b^{(l)}\in \mathbb {R} ^{n_{l}}} , where n l {\displaystyle n_{l}} is the number of neurons in that layer. A weight initialization method is an algorithm for setting the initial values for W ( l ) , b ( l ) {\displaystyle W^{(l)},b^{(l)}} for each layer l {\displaystyle l} . The simplest form is zero initialization: W ( l ) = 0 , b ( l ) = 0 {\displaystyle W^{(l)}=0,b^{(l)}=0} Zero initialization is usually used for initializing biases, but it is not used for initializing weights, as it leads to symmetry in the network, causing all neurons to learn the same features. In this page, we assume b = 0 {\displaystyle b=0} unless otherwise stated. Recurrent neural networks typically use activation functions with bounded range, such as sigmoid and tanh, since unbounded activation may cause exploding values. (Le, Jaitly, Hinton, 2015) suggested initializing weights in the recurrent parts of the network to identity and zero bias, similar to the idea of residual connections and LSTM with no forget gate. In most cases, the biases are initialized to zero, though some situations can use a nonzero initialization. For example, in multiplicative units, such as the forget gate of LSTM, the bias can be initialized to 1 to allow good gradient signal through the gate. For neurons with ReLU activation, one can initialize the bias to a small positive value like 0.1, so that the gradient is likely nonzero at initialization, avoiding the dying ReLU problem. == Random initialization == Random initialization means sampling the weights from a normal distribution or a uniform distribution, usually independently. === LeCun initialization === LeCun initialization, popularized in (LeCun et al., 1998), is designed to preserve the variance of neural activations during the forward pass. It samples each entry in W ( l ) {\displaystyle W^{(l)}} independently from a distribution with mean 0 and variance 1 / n l − 1 {\displaystyle 1/n_{l-1}} . For example, if the distribution is a continuous uniform distribution, then the distribution is U ( ± 3 / n l − 1 ) {\displaystyle {\mathcal {U}}(\pm {\sqrt {3/n_{l-1}}})} . === Glorot initialization === Glorot initialization (or Xavier initialization) was proposed by Xavier Glorot and Yoshua Bengio. It was designed as a compromise between two goals: to preserve activation variance during the forward pass and to preserve gradient variance during the backward pass. For uniform initialization, it samples each entry in W ( l ) {\displaystyle W^{(l)}} independently and identically from U ( ± 6 / ( n l + 1 + n l − 1 ) ) {\displaystyle {\mathcal {U}}(\pm {\sqrt {6/(n_{l+1}+n_{l-1})}})} . In the context, n l − 1 {\displaystyle n_{l-1}} is also called the "fan-in", and n l + 1 {\displaystyle n_{l+1}} the "fan-out". When the fan-in and fan-out are equal, then Glorot initialization is the same as LeCun initialization. === He initialization === As Glorot initialization performs poorly for ReLU activation, He initialization (or Kaiming initialization) was proposed by Kaiming He et al. for networks with ReLU activation. It samples each entry in W ( l ) {\displaystyle W^{(l)}} from N ( 0 , 2 / n l − 1 ) {\displaystyle {\mathcal {N}}(0,2/n_{l-1})} . === Orthogonal initialization === (Saxe et al. 2013) proposed orthogonal initialization: initializing weight matrices as uniformly random (according to the Haar measure) semi-orthogonal matrices, multiplied by a factor that depends on the activation function of the layer. It was designed so that if one initializes a deep linear network this way, then its training time until convergence is independent of depth. Sampling a uniformly random semi-orthogonal matrix can be done by initializing X {\displaystyle X} by IID sampling its entries from a standard normal distribution, then calculate ( X X ⊤ ) − 1 / 2 X {\displaystyle \left(XX^{\top }\right)^{-1/2}X} or its transpose, depending on whether X {\displaystyle X} is tall or wide. For CNN kernels with odd widths and heights, orthogonal initialization is done this way: initialize the central point by a semi-orthogonal matrix, and fill the other entries with zero. As an illustration, a kernel K {\displaystyle K} of shape 3 × 3 × c × c ′ {\displaystyle 3\times 3\times c\times c'} is initialized by filling K [ 2 , 2 , : , : ] {\displaystyle K[2,2,:,:]} with the entries of a random semi-orthogonal matrix of shape c × c ′ {\displaystyle c\times c'} , and the other entries with zero. (Balduzzi et al., 2017) used it with stride 1 and zero-padding. This is sometimes called the Orthogonal Delta initialization. Related to this approach, unitary initialization proposes to parameterize the weight matrices to be unitary matrices, with the result that at initialization they are random unitary matrices (and throughout training, they remain unitary). This is found to improve long-sequence modelling in LSTM. Orthogonal initialization has been generalized to layer-sequential unit-variance (LSUV) initialization. It is a data-dependent initialization method, and can be used in convolutional neural networks. It first initializes weights of each convolution or fully connected layer with orthonormal matrices. Then, proceeding from the first to the last layer, it runs a forward pass on a random minibatch, and divides the layer's weights by the standard deviation of its output, so that its output has variance approximately 1. === Fixup initialization === In 2015, the introduction of residual connections allowed very deep neural networks to be trained, much deeper than the ~20 layers of the previous state of the art (such as the VGG-19). Residual connections gave rise to their own weight initialization problems and strategies. These are sometimes called "normalization-free" methods, since using residual connection could stabilize the training of a deep neural network so much that normalizations become unnecessary. Fixup initialization is designed specifically for networks with residual connections and without batch normalization, as follows: Initialize the classification layer and the last layer of each residual branch to 0. Initialize every other layer using a standard method (such as He initialization), and scale only the weight layers inside residual branches by L − 1 2 m − 2 {\displaystyle L^{-{\frac {1}{2m-2}}}} . Add a scalar multiplier (initialized at 1) in every branch and a scalar bias (initialized at 0) before each convolution, linear, and element-wise activation layer. Similarly, T-Fixup initialization is designed for Transformers without layer normalization. === Others === Instead of initializing all weights with random values on the order of O ( 1 / n ) {\displaystyle O(1/{\sqrt {n}})} , sparse initialization initialized only a small subset of the weights with larger random values, and the other weights zero, so that the total variance is still on the order of O ( 1 ) {\displaystyle O(1)} . Random walk initialization was designed for MLP so that during backpropagation, the L2 norm of gradient at each layer performs an unbiased random walk as one moves from the last layer to the first. Looks linear initialization was designed to allow the neural network to behave like a deep linear network at initialization, since W R e L U ( x ) − W R e L U ( − x ) = W x {\displaystyle W\;\mathrm {ReLU} (x)-W\;\mathrm {ReLU} (-x)=Wx} . It initializes a matrix W {\displaystyle W} of shape R n 2 × m {\displaystyle \mathbb {R} ^{{\frac {n}{2}}\times m}} by any method, such as orthogonal initialization, t
Read more →
MoltenVK

MoltenVK is a software library which allows Vulkan applications to run on top of Metal on Apple's macOS, iOS, and tvOS operating systems. It is the first software component to be released for the Vulkan Portability Initiative, a project to have a subset of Vulkan run on platforms lacking native Vulkan drivers. There are some limitations compared with a native Vulkan implementation. == History == MoltenVK was first released as a proprietary and commercially licensed product by The Brenwill Workshop on July 27, 2016. On July 31, 2017, Khronos announced the formation of the Vulkan Portability Technical Subgroup. === Open source === On February 26, 2018, Khronos announced that Vulkan became available on macOS and iOS products through the MoltenVK library. Valve announced that Dota 2 will run on macOS using the Vulkan API with the aid of MoltenVK, and that they had made an arrangement with developer The Brenwill Workshop Ltd to release MoltenVK as open-source software under the Apache License version 2.0. On May 30, 2018, Qt was updated with Vulkan for Qt on macOS using MoltenVK. On May 31, 2018, optional Vulkan support for Dota 2 on macOS was released. Benchmarks for the game were available the following day, showing better performance using Vulkan and MoltenVK compared to OpenGL. On July 20, 2018, Wine was updated with Vulkan support on macOS using MoltenVK. On 29 July 2018, the first app using MoltenVK was accepted onto the App Store, after initially being rejected. On 6 August 2018, Google open-sourced Filament, a crossplatform real-time physically based rendering engine with MoltenVK for macOS/iOS. On November 28, 2018, Valve released Artifact, their first Vulkan-only game on macOS using MoltenVK. === Version 1.0 === On 29 January 2019, MoltenVK 1.0.32 was released with early prototype of Vulkan Portability Extensions. RPCS3 and Dolphin emulators were updated with Vulkan support on macOS using MoltenVK. On 13 April 2019, MoltenVK 1.0.34 was released with support for tessellation. On July 30, 2019, MoltenVK 1.0.36 was released targeting Metal 3.0. On July 31, 2020, MoltenVK 1.0.44 was released, adding support for the tvOS platform. On January 23, 2020, MoltenVK was updated to support for some of the new features of Vulkan 1.2, as of Vulkan SDK 1.2.121. === Version 1.1 === On October 1, 2020, MoltenVK 1.1.0 was released, adding full support for Vulkan 1.1, as of Vulkan SDK 1.2.154. On 9 December 2020, MoltenVK 1.1.1 was released, providing support for Vulkan on Apple silicon GPUs and support for the Mac Catalyst platform for porting iOS/iPadOS apps to macOS. === Version 1.2 === On October 18, 2022, MoltenVK 1.2.0 was released, adding full support for Vulkan 1.2 as of Vulkan SDK 1.3.231. In January 2023, MoltenVK 1.2.2 added support for Vulkan as of SDK 1.3.239, while this version of Vulkan SDK fixed some issues with the interconnectivity with Metal API, while version 1.2.3 supported some additional extensions. === Version 1.3 === On May 1, 2025, MoltenVK 1.3 was released with support for Vulkan 1.3. === Version 1.4 === On August 20, 2025, MoltenVK 1.4 was released with support for Vulkan 1.4.
Read more →
Randomized Hough transform

Hough transforms are techniques for object detection, a critical step in many implementations of computer vision, or data mining from images. Specifically, the Randomized Hough transform is a probabilistic variant to the classical Hough transform, and is commonly used to detect curves (straight line, circle, ellipse, etc.) The basic idea of Hough transform (HT) is to implement a voting procedure for all potential curves in the image, and at the termination of the algorithm, curves that do exist in the image will have relatively high voting scores. Randomized Hough transform (RHT) is different from HT in that it tries to avoid conducting the computationally expensive voting process for every nonzero pixel in the image by taking advantage of the geometric properties of analytical curves, and thus improve the time efficiency and reduce the storage requirement of the original algorithm. == Motivation == Although Hough transform (HT) has been widely used in curve detection, it has two major drawbacks: First, for each nonzero pixel in the image, the parameters for the existing curve and redundant ones are both accumulated during the voting procedure. Second, the accumulator array (or Hough space) is predefined in a heuristic way. The more accuracy needed, the higher parameter resolution should be defined. These two needs usually result in a large storage requirement and low speed for real applications. Therefore, RHT was brought up to tackle this problem. == Implementation == In comparison with HT, RHT takes advantage of the fact that some analytical curves can be fully determined by a certain number of points on the curve. For example, a straight line can be determined by two points, and an ellipse (or a circle) can be determined by three points. The case of ellipse detection can be used to illustrate the basic idea of RHT. The whole process generally consists of three steps: Fit ellipses with randomly selected points. Update the accumulator array and corresponding scores. Output the ellipses with scores higher than some predefined threshold. === Ellipse fitting === One general equation for defining ellipses is: a ( x − p ) 2 + 2 b ( x − p ) ( y − q ) + c ( y − q ) 2 = 1 {\displaystyle a(x-p)^{2}+2b(x-p)(y-q)+c(y-q)^{2}=1} with restriction: a c − b 2 > 0 {\displaystyle ac-b^{2}>0} However, an ellipse can be fully determined if one knows three points on it and the tangents in these points. RHT starts by randomly selecting three points on the ellipse. Let them be X 1 {\displaystyle X_{1}} , X 2 {\displaystyle X_{2}} and X 3 {\displaystyle X_{3}} . The first step is to find the tangents of these three points. They can be found by fitting a straight line using least squares technique for a small window of neighboring pixels. The next step is to find the intersection points of the tangent lines. This can be easily done by solving the line equations found in the previous step. Then let the intersection points be T 12 {\displaystyle T_{12}} and T 23 {\displaystyle T_{23}} , the midpoints of line segments X 1 X 2 {\displaystyle X_{1}X_{2}} and X 2 X 3 {\displaystyle X_{2}X_{3}} be M 12 {\displaystyle M_{12}} and M 23 {\displaystyle M_{23}} . Then the center of the ellipse will lie in the intersection of T 12 M 12 {\displaystyle T_{12}M_{12}} and T 23 M 23 {\displaystyle T_{23}M_{23}} . Again, the coordinates of the intersected point can be determined by solving line equations and the detailed process is skipped here for conciseness. Let the coordinates of ellipse center found in previous step be ( x 0 , y 0 ) {\displaystyle (x_{0},y_{0})} . Then the center can be translated to the origin with x ′ = x − x 0 {\displaystyle x'=x-x_{0}} and y ′ = y − y 0 {\displaystyle y'=y-y_{0}} so that the ellipse equation can be simplified to: a x ′ 2 + 2 b x ′ y ′ + c y ′ 2 = 1 {\displaystyle ax'^{2}+2bx'y'+cy'^{2}=1} Now we can solve for the rest of ellipse parameters: a {\displaystyle a} , b {\displaystyle b} and c {\displaystyle c} by substituting the coordinates of X 1 {\displaystyle X_{1}} , X 2 {\displaystyle X_{2}} and X 3 {\displaystyle X_{3}} into the equation above. === Accumulating === With the ellipse parameters determined from previous stage, the accumulator array can be updated correspondingly. Different from classical Hough transform, RHT does not keep "grid of buckets" as the accumulator array. Rather, it first calculates the similarities between the newly detected ellipse and the ones already stored in accumulator array. Different metrics can be used to calculate the similarity. As long as the similarity exceeds some predefined threshold, replace the one in the accumulator with the average of both ellipses and add 1 to its score. Otherwise, initialize this ellipse to an empty position in the accumulator and assign a score of 1. === Termination === Once the score of one candidate ellipse exceeds the threshold, it is determined as existing in the image (in other words, this ellipse is detected), and should be removed from the image and accumulator array so that the algorithm can detect other potential ellipses faster. The algorithm terminates when the number of iterations reaches a maximum limit or all the ellipses have been detected. Pseudo code for RHT: while (we find ellipses AND not reached the maximum epoch) { for (a fixed number of iterations) { Find a potential ellipse. if (the ellipse is similar to an ellipse in the accumulator) then Replace the one in the accumulator with the average of two ellipses and add 1 to the score; else Insert the ellipse into an empty position in the accumulator with a score of 1; } Select the ellipse with the best score and save it in a best ellipse table; Eliminate the pixels of the best ellipse from the image; Empty the accumulator; }
Read more →
Landmark point

In morphometrics, landmark point or shortly landmark is a point in a shape object in which correspondences between and within the populations of the object are preserved. In other disciplines, landmarks may be known as vertices, anchor points, control points, sites, profile points, 'sampling' points, nodes, markers, fiducial markers, etc. Landmarks can be defined either manually by experts or automatically by a computer program. There are three basic types of landmarks: anatomical landmarks, mathematical landmarks or pseudo-landmarks. An anatomical landmark is a biologically-meaningful point in an organism. Usually experts define anatomical points to ensure their correspondences within the same species. Examples of anatomical landmark in shape of a skull are the eye corner, tip of the nose, jaw, etc. Anatomical landmarks determine homologous parts of an organism, which share a common ancestry. Mathematical landmarks are points in a shape that are located according to some mathematical or geometrical property, for instance, a high curvature point or an extreme point. A computer program usually determines mathematical landmarks used for an automatic pattern recognition. Pseudo-landmarks are constructed points located between anatomical or mathematical landmarks. A typical example is an equally spaced set of points between two anatomical landmarks to get more sample points from a shape. Pseudo-landmarks are useful during shape matching, when the matching process requires a large number of points.
Read more →
Integreat

Integreat (former project name: Refguide+) is an open source mobile app that provides local information and services tailored to refugees and migrants coming to Germany. The content is maintained by local organizations, such as local governments or integration officers, and made available in locally relevant languages. It was developed by Tür an Tür - Digitalfabrik gGmbH (formerly Tür an Tür - Digital Factory gGmbH) in Augsburg together with a team of researchers and students from the Technical University of Munich. == History == In 1997, the Augsburg association "Tür an Tür", which has been working for refugees since 1992, published the brochure "First Steps", which answers local everyday questions. Since addresses and contact persons change quickly, some information is already outdated after a few weeks. Students of business informatics at the Technical University of Munich therefore developed the app Integreat within eight months together with the association and the social department of the city of Augsburg. The app was then also used by other cities and districts within months. As of February 3, 2022, information is available at 72 locations, including Munich, Dortmund, Nuremberg and Augsburg. == Mode of action == Refugees need information on areas such as registration, contact persons, health care, education, family, work and everyday life. Integreat seeks to provide refugees with this information by allowing them to select their geographic location and receive locally relevant information. This information is available offline once the app is opened so it can be used without an internet connection. In addition, the content is translated into the native languages of refugees and migrants to facilitate access. The content is licensed with a CC BY 4.0 license to facilitate collaboration and translation between content creators and dissemination of the content. Integreat is now being used for a broader migrant audience and says it can also support professionals, volunteers, and counseling centers. == Comparable mobile apps == Other mobile apps that are likewise intended to provide initial orientation for refugees include the app Ankommen, a joint project of the Federal Office for Migration and Refugees, the Goethe-Institut, the Federal Employment Agency and the Bavarian Broadcasting Corporation, which is intended as a companion for the first few weeks in Germany, and the Welcome App, a company-sponsored non-profit initiative for information about Germany and asylum procedures with a regional focus, and a book by the Konrad Adenauer Foundation (KAS) and Verlag Herder with a corresponding app Deutschland - Erste Informationen für Flüchtlinge (Germany - First Information for Refugees) as a companion for Arabic-speaking refugees in Germany.
Read more →
List of online database creator apps

This list of online database creator apps lists notable web apps where end users with minimal database administration expertise can create online databases to share with team members. Users need not have the coding skills to manage the solution stack themselves, because the web app already provides this predefined functionality. Such online database creator apps serve the gap between IT professionals (who can manage such a stack themselves) and people who would not create databases at all anyway. In other words, they provide a low-code way of doing database administration. As the concept of low-code development in general continues to evolve, some of the brands that began as online database creator apps are evolving into low-code development platforms for both the databases and the custom apps that use them. Airtable Bubble Caspio Coda.io Microsoft Access web apps plus SharePoint Oracle Application Express aka APEX Quickbase WaveMaker Rapid ZohoCreator
Read more →

Retrieval-augmented generation

Retrieval-augmented generation (RAG) is a technique that enables large language models (LLMs) to retrieve and incorporate new information from external data sources. With RAG, LLMs first refer to a specified set of documents, then respond to user queries. These documents supplement information from the LLM's pre-existing training data. This allows LLMs to use domain-specific and/or updated information that is not available in the training data. For example, this enables LLM-based chatbots to access internal company data or generate responses based on authoritative sources. RAG improves LLMs by incorporating information retrieval before generating responses. Unlike LLMs that rely on static training data, RAG pulls relevant text from databases, uploaded documents, or web sources. According to Ars Technica, "RAG is a way of improving LLM performance, in essence by blending the LLM process with a web search or other document look-up process to help LLMs stick to the facts." This method helps reduce AI hallucinations, which have caused chatbots to describe policies that don't exist, or recommend nonexistent legal cases to lawyers that are looking for citations to support their arguments. RAG also reduces the need to retrain LLMs with new data, saving on computational and financial costs. Beyond efficiency gains, RAG also allows LLMs to include sources in their responses, so users can verify the cited sources. This provides greater transparency, as users can cross-check retrieved content to ensure accuracy and relevance. The term retrieval-augmented generation (RAG) was introduced in a 2020 paper that described combining a parametric language model with a non-parametric external memory accessed through retrieval at inference time. == RAG and LLM limitations == LLMs can provide incorrect information. For example, when Google first demonstrated its LLM tool "Google Bard" (later re-branded to Gemini), the LLM provided incorrect information about the James Webb Space Telescope. This error contributed to a $100 billion decline in Google's stock value. RAG is used to prevent these errors, but it does not solve all the problems. For example, LLMs can generate misinformation even when pulling from factually correct sources if they misinterpret the context. MIT Technology Review gives the example of an AI-generated response stating, "The United States has had one Muslim president, Barack Hussein Obama." The model retrieved this from an academic book rhetorically titled Barack Hussein Obama: America's First Muslim President? The LLM did not "know" or "understand" the context of the title, generating a false statement. LLMs with RAG are programmed to prioritize new information. This technique has been called "prompt stuffing." Without prompt stuffing, the LLM's input is generated by a user; with prompt stuffing, additional relevant context is added to this input to guide the model's response. This approach provides the LLM with key information early in the prompt, encouraging it to prioritize the supplied data over pre-existing training knowledge. == Process == Retrieval-augmented generation (RAG) enhances large language models (LLMs) by incorporating an information-retrieval mechanism that allows models to access and utilize additional data beyond their original training set. Ars Technica notes that "when new information becomes available, rather than having to retrain the model, all that's needed is to augment the model's external knowledge base with the updated information" ("augmentation"). IBM states that "in the generative phase, the LLM draws from the augmented prompt and its internal representation of its training data to synthesize" an answer. === RAG key stages === Typically, the data to be referenced is converted into LLM embeddings, numerical representations in the form of a large vector space. RAG can be used on unstructured (usually text), semi-structured, or structured data (for example knowledge graphs). These embeddings are then stored in a vector database to allow for document retrieval. Given a user query, a document retriever is first called to select the most relevant documents that will be used to augment the query. This comparison can be done using a variety of methods, which depend in part on the type of indexing used. The model feeds this relevant retrieved information into the LLM via prompt engineering of the user's original query. Newer implementations (as of 2023) can also incorporate specific augmentation modules with abilities such as expanding queries into multiple domains and using memory and self-improvement to learn from previous retrievals. Finally, the LLM can generate output based on both the query and the retrieved documents. Some models incorporate extra steps to improve output, such as the re-ranking of retrieved information, context selection, and fine-tuning. == Applications == Retrieval-augmented generation is used in applications where generated responses need to be grounded in external or frequently updated information. Commonly cited use cases include search engines, question-answering systems, customer support chatbots, enterprise knowledge assistants, content generation, recommendation systems, retail and e-commerce, and industrial or manufacturing workflows. In healthcare, RAG has been studied as a way to ground large language model outputs in external medical knowledge sources, although reviews have noted continuing challenges around evaluation, ethics, and clinical reliability. == Improvements == Improvements to the basic process above can be applied at different stages in the RAG flow. === Encoder === These methods focus on the encoding of text as either dense or sparse vectors. Sparse vectors, which encode the identity of a word, are typically dictionary-length and contain mostly zeros. Dense vectors, which encode meaning, are more compact and contain fewer zeros. Various enhancements can improve the way similarities are calculated in the vector stores (databases). Performance improves by optimizing how vector similarities are calculated. Dot products enhance similarity scoring, while approximate nearest neighbor (ANN) searches improve retrieval efficiency over K-nearest neighbors (KNN) searches. Accuracy may be improved with Late Interactions, which allow the system to compare words more precisely after retrieval. This helps refine document ranking and improve search relevance. Hybrid vector approaches may be used to combine dense vector representations with sparse one-hot vectors, taking advantage of the computational efficiency of sparse dot products over dense vector operations. Other retrieval techniques focus on improving accuracy by refining how documents are selected. Some retrieval methods combine sparse representations, such as SPLADE, with query expansion strategies to improve search accuracy and recall. === Retriever-centric methods === These methods aim to enhance the quality of document retrieval in vector databases: Pre-training the retriever using the Inverse Cloze Task (ICT), a technique that helps the model learn retrieval patterns by predicting masked text within documents. Supervised retriever optimization aligns retrieval probabilities with the generator model's likelihood distribution. This involves retrieving the top-k vectors for a given prompt, scoring the generated response's perplexity, and minimizing KL divergence between the retriever's selections and the model's likelihoods to refine retrieval. Reranking techniques can refine retriever performance by prioritizing the most relevant retrieved documents during training. === Language model === By redesigning the language model with the retriever in mind, a 25-time smaller network can get comparable perplexity as its much larger counterparts. Because it is trained from scratch, this method (Retro) incurs the high cost of training runs that the original RAG scheme avoided. The hypothesis is that by giving domain knowledge during training, Retro needs less focus on the domain and can devote its smaller weight resources only to language semantics. The redesigned language model is shown here. It has been reported that Retro is not reproducible, so modifications were made to make it so. The more reproducible version is called Retro++ and includes in-context RAG. === Chunking === Chunking involves various strategies for breaking up the data into vectors so the retriever can find details in it. Three types of chunking strategies are: Fixed length with overlap. This is fast and easy. Overlapping consecutive chunks helps to maintain semantic context across chunks. Syntax-based chunks can break the document up into sentences. Libraries such as spaCy or NLTK can also help. File format-based chunking. Certain file types have natural chunks built in, and it's best to respect them. For example, code files are best chunked and vectorized as whole functions or classes. HTML files should leave

or base64 encoded elements

Statistical shape analysis

Statistical shape analysis is an analysis of the geometrical properties of some given set of shapes by statistical methods. For instance, it could be used to quantify differences between male and female gorilla skull shapes, normal and pathological bone shapes, leaf outlines with and without herbivory by insects, etc. Important aspects of shape analysis are to obtain a measure of distance between shapes, to estimate mean shapes from (possibly random) samples, to estimate shape variability within samples, to perform clustering and to test for differences between shapes. One of the main methods used is principal component analysis (PCA). Statistical shape analysis has applications in various fields, including medical imaging, computer vision, computational anatomy, sensor measurement, and geographical profiling. == Landmark-based techniques == In the point distribution model, a shape is determined by a finite set of coordinate points, known as landmark points. These landmark points often correspond to important identifiable features such as the corners of the eyes. Once the points are collected some form of registration is undertaken. This can be a baseline methods used by Fred Bookstein for geometric morphometrics in anthropology. Or an approach like Procrustes analysis which finds an average shape. David George Kendall investigated the statistical distribution of the shape of triangles, and represented each triangle by a point on a sphere. He used this distribution on the sphere to investigate ley lines and whether three stones were more likely to be co-linear than might be expected. Statistical distribution like the Kent distribution can be used to analyse the distribution of such spaces. Alternatively, shapes can be represented by curves or surfaces representing their contours, by the spatial region they occupy. == Shape deformations == Differences between shapes can be quantified by investigating deformations transforming one shape into another. In particular a diffeomorphism preserves smoothness in the deformation. This was pioneered in D'Arcy Thompson's On Growth and Form before the advent of computers. Deformations can be interpreted as resulting from a force applied to the shape. Mathematically, a deformation is defined as a mapping from a shape x to a shape y by a transformation function Φ {\displaystyle \Phi } , i.e., y = Φ ( x ) {\displaystyle y=\Phi (x)} . Given a notion of size of deformations, the distance between two shapes can be defined as the size of the smallest deformation between these shapes. Diffeomorphometry is the focus on comparison of shapes and forms with a metric structure based on diffeomorphisms, and is central to the field of Computational anatomy. Diffeomorphic registration, introduced in the 90's, is now an important player with existing codes bases organized around ANTS, DARTEL, DEMONS, LDDMM, StationaryLDDMM, and FastLDDMM are examples of actively used computational codes for constructing correspondences between coordinate systems based on sparse features and dense images. Voxel-based morphometry (VBM) is an important technology built on many of these principles. Methods based on diffeomorphic flows are also used. For example, deformations could be diffeomorphisms of the ambient space, resulting in the LDDMM (Large Deformation Diffeomorphic Metric Mapping) framework for shape comparison.

Text normalization

Text normalization is the process of transforming text into a single canonical form that it might not have had before. Normalizing text before storing or processing it allows for separation of concerns, since input is guaranteed to be consistent before operations are performed on it. Text normalization requires being aware of what type of text is to be normalized and how it is to be processed afterwards; there is no all-purpose normalization procedure. == Applications == Text normalization is frequently used when converting text to speech. Numbers, dates, acronyms, and abbreviations are non-standard "words" that need to be pronounced differently depending on context. For example: "$200" would be pronounced as "two hundred dollars" in English, but as "lua selau tālā" in Samoan. "vi" could be pronounced as "vie," "vee," or "the sixth" depending on the surrounding words. Text can also be normalized for storing and searching in a database. For instance, if a search for "resume" is to match the word "résumé," then the text would be normalized by removing diacritical marks; and if "john" is to match "John", the text would be converted to a single case. To prepare text for searching, it might also be stemmed (e.g. converting "flew" and "flying" both into "fly"), canonicalized (e.g. consistently using American or British English spelling), or have stop words removed. == Techniques == For simple, context-independent normalization, such as removing non-alphanumeric characters or diacritical marks, regular expressions would suffice. For example, the sed script sed ‑e "s/\s+/ /g" inputfile would normalize runs of whitespace characters into a single space. More complex normalization requires correspondingly complicated algorithms, including domain knowledge of the language and vocabulary being normalized. Among other approaches, text normalization has been modeled as a problem of tokenizing and tagging streams of text and as a special case of machine translation. == Textual scholarship == In the field of textual scholarship and the editing of historic texts, the term "normalization" implies a degree of modernization and standardization – for example in the extension of scribal abbreviations and the transliteration of the archaic glyphs typically found in manuscript and early printed sources. A normalized edition is therefore distinguished from a diplomatic edition (or semi-diplomatic edition), in which some attempt is made to preserve these features. The aim is to strike an appropriate balance between, on the one hand, rigorous fidelity to the source text (including, for example, the preservation of enigmatic and ambiguous elements); and, on the other, producing a new text that will be comprehensible and accessible to the modern reader. The extent of normalization is therefore at the discretion of the editor, and will vary. Some editors, for example, choose to modernize archaic spellings and punctuation, but others do not. An edition of a text might be normalized based on internal criteria, where orthography is standardized according to the language of the original, or external criteria, where the norms of a different time period are applied. For an example of the latter, a published edition of a medieval Icelandic manuscript might be normalized to the conventions of modern Icelandic, or it might be normalized to Classical Old Icelandic. Standards of normalization vary based on language of the edition as well as the specific conventions of the publisher.

ReRites

ReRites (also known as RERITES, ReadingRites, Big Data Poetry) is a literary work of "Human + A.I. poetry" by David Jhave Johnston that used neural network models trained to generate poetry which the author then edited. ReRites won the Robert Coover Award for a Work of Electronic Literature in 2022. == About the project == The ReRites project began as a daily rite of writing with a neural network, expanded into a series of performances from which video documentation has been published online, and concluded with a set of 12 books and an accompanying book of essays published by Anteism Books in 2019. In Electronic Literature, Scott Rettberg describes the early phases of the project in 2016, when it bore the preliminary name Big Data Poetry. Jhave (the artist name that David Jhave Johnston goes by) describes the process of writing ReRites as a rite: "Every morning for 2 hours (normally 6:30–8:30am) I get up and edit the poetic output of a neural net. Deleting, weaving, conjugating, lineating, cohering. Re-writing. Re-wiring authorship: hybrid augmented enhanced evolutionary". There is video documentation of the writing process. The human editing of the neural network's output is fundamental to this project, and Jhave gives examples of both unedited text extracts and his edited versions in publications about the project. Kyle Booten describes ReRites as "simultaneously dusty and outrageously verdant, monotonously sublime and speckled with beautiful and rare specimens". === Performances === ReRites was first shared with an audience through a series of performances where audience members and poets would participate in reading the automatically generated texts, which appeared on screen so fast that human readers could barely keep up. This has been described as allowing participants to "re-discover[..] the peculiar pleasures of being embodied", or, in Jhave's own words, as a space where human participants were "playing their wits and voices against an evocative infinite deep-learning muse". The first performance was at Brown University's Interrupt Festival in 2019. It has been performed many times since, including at the Barbican Centre in London and Anteism Books. === Print publications === For a single year Jhave published one book of poetry from the ReRites project each month. These twelve volumes are accompanied by a book of essays, all published by Anteism Books. The accompanying essays provide critical responses to the project from poets and scholars including Allison Parrish, Johanna Drucker, Kyle Booten, Stephanie Strickland, John Cayley, Lai-Tze Fan, Nick Montfort, Mairéad Byrne, and Chris Funkhouser. Allison Parrish notes elsewhere that these paratexts to ReRites serve a legitimising function for a genre of poetry that is not yet institutionally acknowledged. === Technical details === Starting in 2016 under the name Big Data Poetry, Jhave generated poems using, in his own words, "neural network code (..) adapted from three corporate github-hosted machine-learning libraries: TensorFlow (Google), PyTorch (Facebook), and AWD-LSTM (SalesForce)". He explains that the "models were trained on a customised corpus of 600,000 lines of poetry ranging from the romantic epoch to the 20th century avant garde". Jhave maintains a GitHub repository with some of the code supporting ReRites. == Reception == ReRites is described by John Cayley as "one of the most thorough and beautiful" poetic responses to machine learning. The work's influence on the field of electronic literature was acknowledged in 2022, when the work won the Electronic Literature Organization's Robert Coover Award for a Work of Electronic Literature. The jury described ReRites as particularly poignant in the time of the pandemic, as it was "a documentation of the performance of the private ritual of writing and the obsessive-compulsive need for writers to communicate — even when no one else is reading". The question of authorship and voice in ReRites has been raised by several critics. Although generated poetry is an established genre in electronic literature, Cayley notes that unlike the combinatory poems created by authors like Nick Montfort, where the author explicitly defines which words and phrases will be recombined, ReRites has "not been directed by literary preconceptions inscribed in the program itself, but only by patterns and rhythms pre-existing in the corpora". In an essay for the Australian journal TEXT, David Thomas Henry Wright asks how to understand authorship and authority in ReRites: "Who or what is the authority of the work? The original data fed into the machine, that is not currently retrievable or discernible from the final works? The code that was taken and adapted for his purposes? Or Jhave, the human editor?" Wright concludes that Jhave is the only actor with any intentionality and therefore the authority of the work. The centrality of the human editor is also emphasised by other scholars. In a chapter analysing ReRites Malthe Stavning Erslev argues that the machine learning misrepresents the dataset it is trained on. While ReRites uses 21st century neural networks, it has been compared to earlier literary traditions. Poet Victoria Stanton, who read at one of the ReRites performances, has compared ReRites to found poetry, while David Thomas Henry Wright compares it to the Oulipo movement and Mark Amerika to the cut-up technique. Scholars also position ReRites firmly within the long tradition of generative poetry both in electronic literature and print, stretching from the I Ching, Queneau's Cent Mille Milliards de Poemes and Nabokov's Pale Fire to computer-generated poems like Christopher Strachey's Love Letter Generator (1952) and more contemporary examples. Jhave describes the process of working with the output from the neural network as "carving". In his book My Life as an Artificial Creative Intelligence, Mark Amerika writes that the "method of carving the digital outputs provided by the language model as part of a collaborative remix jam session with GPT-2, where the language artist and the language model play off each other’s unexpected outputs as if caught in a live postproduction set, is one I share with electronic literature composer David Jhave Johnston, whose AI poetry experiments precede my own investigations."

Lexical substitution

Lexical substitution is the task of identifying a substitute for a word in the context of a clause. For instance, given the following text: "After the match, replace any remaining fluid deficit to prevent chronic dehydration throughout the tournament", a substitute of game might be given. Lexical substitution is strictly related to word sense disambiguation (WSD), in that both aim to determine the meaning of a word. However, while WSD consists of automatically assigning the appropriate sense from a fixed sense inventory, lexical substitution does not impose any constraint on which substitute to choose as the best representative for the word in context. By not prescribing the inventory, lexical substitution overcomes the issue of the granularity of sense distinctions and provides a level playing field for automatic systems that automatically acquire word senses (a task referred to as Word Sense Induction). == Evaluation == In order to evaluate automatic systems on lexical substitution, a task was organized at the Semeval-2007 evaluation competition held in Prague in 2007. A Semeval-2010 task on cross-lingual lexical substitution has also taken place. == Skip-gram model == The skip-gram model takes words with similar meanings into a vector space (collection of objects that can be added together and multiplied by numbers) that are found close to each other in N-dimensions (list of items). A variety of neural networks (computer system modeled after a human brain) are formed together as a result of the vectors and networks that are related together. This all occurs in the dimensions of the vocabulary that has been generated in a network. The model has been used in lexical substitution automation and prediction algorithms. One such algorithm developed by Oren Melamud, Omer Levy, and Ido Dagan uses the skip-gram model to find a vector for each word and its synonyms. Then, it calculates the cosine distance between vectors to determine which words will be the best substitutes. === Example === In a sentence like "The dog walked at a quick pace" each word has a specific vector in relation to the other. The vector for "The" would be [1,0,0,0,0,0,0] because the 1 is the word vocabulary and the 0s are the words surrounding that vocabulary, which create a vector.

XLNet

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