In machine learning, a variational autoencoder (VAE) is an artificial neural network architecture introduced by Diederik P. Kingma and Max Welling in 2013. It is part of the families of probabilistic graphical models and variational Bayesian methods. In addition to being seen as an autoencoder neural network architecture, variational autoencoders can also be studied within the mathematical formulation of variational Bayesian methods, connecting a neural encoder network to its decoder through a probabilistic latent space (for example, as a multivariate Gaussian distribution) that corresponds to the parameters of a variational distribution. Thus, the encoder maps each point (such as an image) from a large complex dataset into a distribution within the latent space, rather than to a single point in that space. The decoder has the opposite function, which is to map from the latent space to the input space, again according to a distribution (although in practice, noise is rarely added during the decoding stage). By mapping a point to a distribution instead of a single point, the network can avoid overfitting the training data. Both networks are typically trained together with the usage of the reparameterization trick, although the variance of the noise model can be learned separately. Although this type of model was initially designed for unsupervised learning, its effectiveness has been proven for semi-supervised learning and supervised learning. == Overview of architecture and operation == A variational autoencoder is a generative model with a prior and noise distribution respectively. Usually such models are trained using the expectation-maximization meta-algorithm (e.g. probabilistic PCA, (spike & slab) sparse coding). Such a scheme optimizes a lower bound of the data likelihood, which is usually computationally intractable, and in doing so requires the discovery of q-distributions, or variational posteriors. These q-distributions are normally parameterized for each individual data point in a separate optimization process. However, variational autoencoders use a neural network as an amortized approach to jointly optimize across data points. In that way, the same parameters are reused for multiple data points, which can result in massive memory savings. The first neural network takes as input the data points themselves, and outputs parameters for the variational distribution. As it maps from a known input space to the low-dimensional latent space, it is called the encoder. The decoder is the second neural network of this model. It is a function that maps from the latent space to the input space, e.g. as the means of the noise distribution. It is possible to use another neural network that maps to the variance, however this can be omitted for simplicity. In such a case, the variance can be optimized with gradient descent. To optimize this model, one needs to know two terms: the "reconstruction error", and the Kullback–Leibler divergence (KL-D). Both terms are derived from the free energy expression of the probabilistic model, and therefore differ depending on the noise distribution and the assumed prior of the data, here referred to as p-distribution. For example, a standard VAE task such as IMAGENET is typically assumed to have a gaussianly distributed noise; however, tasks such as binarized MNIST require a Bernoulli noise. The KL-D from the free energy expression maximizes the probability mass of the q-distribution that overlaps with the p-distribution, which unfortunately can result in mode-seeking behaviour. The "reconstruction" term is the remainder of the free energy expression, and requires a sampling approximation to compute its expectation value. More recent approaches replace Kullback–Leibler divergence (KL-D) with various statistical distances, see "Statistical distance VAE variants" below. == Formulation == From the point of view of probabilistic modeling, one wants to maximize the likelihood of the data x {\displaystyle x} by their chosen parameterized probability distribution p θ ( x ) = p ( x | θ ) {\displaystyle p_{\theta }(x)=p(x|\theta )} . This distribution is usually chosen to be a Gaussian N ( x | μ , σ ) {\displaystyle N(x|\mu ,\sigma )} which is parameterized by μ {\displaystyle \mu } and σ {\displaystyle \sigma } respectively, and as a member of the exponential family it is easy to work with as a noise distribution. Simple distributions are easy enough to maximize, however distributions where a prior is assumed over the latents z {\displaystyle z} results in intractable integrals. Let us find p θ ( x ) {\displaystyle p_{\theta }(x)} via marginalizing over z {\displaystyle z} . p θ ( x ) = ∫ z p θ ( x , z ) d z , {\displaystyle p_{\theta }(x)=\int _{z}p_{\theta }({x,z})\,dz,} where p θ ( x , z ) {\displaystyle p_{\theta }({x,z})} represents the joint distribution under p θ {\displaystyle p_{\theta }} of the observable data x {\displaystyle x} and its latent representation or encoding z {\displaystyle z} . According to the chain rule, the equation can be rewritten as p θ ( x ) = ∫ z p θ ( x | z ) p θ ( z ) d z {\displaystyle p_{\theta }(x)=\int _{z}p_{\theta }({x|z})p_{\theta }(z)\,dz} In the vanilla variational autoencoder, z {\displaystyle z} is usually taken to be a finite-dimensional vector of real numbers, and p θ ( x | z ) {\displaystyle p_{\theta }({x|z})} to be a Gaussian distribution. Then p θ ( x ) {\displaystyle p_{\theta }(x)} is a mixture of Gaussian distributions. It is now possible to define the set of the relationships between the input data and its latent representation as Prior p θ ( z ) {\displaystyle p_{\theta }(z)} Likelihood p θ ( x | z ) {\displaystyle p_{\theta }(x|z)} Posterior p θ ( z | x ) {\displaystyle p_{\theta }(z|x)} Unfortunately, the computation of p θ ( z | x ) {\displaystyle p_{\theta }(z|x)} is expensive and in most cases intractable. To speed up the calculus to make it feasible, it is necessary to introduce a further function to approximate the posterior distribution as q ϕ ( z | x ) ≈ p θ ( z | x ) {\displaystyle q_{\phi }({z|x})\approx p_{\theta }({z|x})} with ϕ {\displaystyle \phi } defined as the set of real values that parametrize q {\displaystyle q} . This is sometimes called amortized inference, since by "investing" in finding a good q ϕ {\displaystyle q_{\phi }} , one can later infer z {\displaystyle z} from x {\displaystyle x} quickly without doing any integrals. In this way, the problem is to find a good probabilistic autoencoder, in which the conditional likelihood distribution p θ ( x | z ) {\displaystyle p_{\theta }(x|z)} is computed by the probabilistic decoder, and the approximated posterior distribution q ϕ ( z | x ) {\displaystyle q_{\phi }(z|x)} is computed by the probabilistic encoder. Parametrize the encoder as E ϕ {\displaystyle E_{\phi }} , and the decoder as D θ {\displaystyle D_{\theta }} . == Evidence lower bound (ELBO) == Like many deep learning approaches that use gradient-based optimization, VAEs require a differentiable loss function to update the network weights through backpropagation. For variational autoencoders, the idea is to jointly optimize the generative model parameters θ {\displaystyle \theta } to reduce the reconstruction error between the input and the output, and ϕ {\displaystyle \phi } to make q ϕ ( z | x ) {\displaystyle q_{\phi }({z|x})} as close as possible to p θ ( z | x ) {\displaystyle p_{\theta }(z|x)} . As reconstruction loss, mean squared error and cross entropy are often used. The Kullback–Leibler divergence D K L ( q ϕ ( z | x ) ∥ p θ ( z | x ) ) {\displaystyle D_{KL}(q_{\phi }({z|x})\parallel p_{\theta }({z|x}))} can be used as a loss function to squeeze q ϕ ( z | x ) {\displaystyle q_{\phi }({z|x})} under p θ ( z | x ) {\displaystyle p_{\theta }(z|x)} . This divergence loss expands to D K L ( q ϕ ( z | x ) ∥ p θ ( z | x ) ) = E z ∼ q ϕ ( ⋅ | x ) [ ln q ϕ ( z | x ) p θ ( z | x ) ] = E z ∼ q ϕ ( ⋅ | x ) [ ln q ϕ ( z | x ) p θ ( x ) p θ ( x , z ) ] = ln p θ ( x ) + E z ∼ q ϕ ( ⋅ | x ) [ ln q ϕ ( z | x ) p θ ( x , z ) ] . {\displaystyle {\begin{aligned}D_{KL}(q_{\phi }({z|x})\parallel p_{\theta }({z|x}))&=\mathbb {E} _{z\sim q_{\phi }(\cdot |x)}\left[\ln {\frac {q_{\phi }(z|x)}{p_{\theta }(z|x)}}\right]\\&=\mathbb {E} _{z\sim q_{\phi }(\cdot |x)}\left[\ln {\frac {q_{\phi }({z|x})p_{\theta }(x)}{p_{\theta }(x,z)}}\right]\\&=\ln p_{\theta }(x)+\mathbb {E} _{z\sim q_{\phi }(\cdot |x)}\left[\ln {\frac {q_{\phi }({z|x})}{p_{\theta }(x,z)}}\right].\end{aligned}}} Now, define the evidence lower bound (ELBO): L θ , ϕ ( x ) := E z ∼ q ϕ ( ⋅ | x ) [ ln p θ ( x , z ) q ϕ ( z | x ) ] = ln p θ ( x ) − D K L ( q ϕ ( ⋅ | x ) ∥ p θ ( ⋅ | x ) ) {\displaystyle L_{\theta ,\phi }(x):=\mathbb {E} _{z\sim q_{\phi }(\cdot |x)}\left[\ln {\frac {p_{\theta }(x,z)}{q_{\phi }({z|x})}}\right]=\ln p_{\theta }(x)-D_{KL}(q_{\phi }({\cdot |x})\parallel p_{\theta }({\cdot |x}))} Maximizing the ELBO θ ∗ , ϕ ∗ = argmax θ , ϕ L θ , ϕ ( x ) {\dis
Cloem
Cloem is a company based in Cannes, France, which applies natural language processing (NLP) technologies to assist patent applicants in creating variants of patent claims, called "cloems". According to the company, these "computer-generated claims can be published to keep potential competitors from attempting to file adjacent patent claims." == Technology == According to Cloem, dictionaries, ontologies and proprietary claim-drafting algorithms are used to draft alternative claims based on a client's original set of claims. In particular, the original set of claims is subject to various permutations and linguistic manipulations "by considering alternative definitions for terms as well as “synonyms, hyponyms, hyperonyms, meronyms, holonyms, and antonyms.”" == Possible uses == Cloem can optionally publish one or more created texts, as electronic publications or as paper-printed publications. These can potentially serve – through a defensive publication – as prior art to prevent another party for obtaining a patent on the subject-matter at stake. In other words, after an initial patent filing, an "improvement" patent (adjacent invention) can be applied for by another party, such as a competitor. By publishing variants of a patent claim, the risk of adverse patenting may potentially be decreased (improvement inventions may no longer be patentable). Cloems may also be potentially patentable. One of the issues of patentability, however, is that only a natural person can be a listed as an inventor on a patent. Since cloems are produced by a computer based on a person's input, it is not clear if the computer or the person is the inventor. The inventorship of Cloem texts is an open question.
DREAM Challenges
DREAM Challenges (Dialogue for Reverse Engineering Assessment and Methods) is a non-profit initiative for advancing biomedical and systems biology research via crowd-sourced competitions. Started in 2006, DREAM challenges collaborate with Sage Bionetworks to provide a platform for competitions run on the Synapse platform. Over 60 DREAM challenges have been conducted over the span of over 15 years. == Overview == DREAM Challenges were founded in 2006 by Gustavo Stolovizky from IBM Research and Andrea Califano from Columbia University. Current chair of the DREAM organization is Paul Boutros from University of California. Further organization spans emeritus chairs Justin Guinney and Gustavo Stolovizky, and multiple DREAM directors. Individual challenges focus on tackling a specific biomedical research question, typically narrowed down to a specific disease. A prominent disease focus has been on oncology, with multiple past challenges focused on breast cancer, acute myeloid leukemia, and prostate cancer or similar diseases. The data involved in an individual challenge reflects the disease context; while cancers typically involve data such as mutations in the human genome, gene expression and gene networks in transcriptomics, and large scale proteomics, newer challenges have shifted towards single cell sequencing technologies as well as emerging gut microbiome related research questions, thus reflecting trends in the wider research community. Motivation for DREAM Challenges is that via crowd-sourcing data to a larger audience via competitions, better models and insight is gained than if the analysis was conducted by a single entity. Past competitions have been published in such scientific venues as the flagship journals of the Nature Portfolio and PLOS publishing groups. Results of DREAM challenges are announced via web platforms, and the top performing participants are invited to present their results in the annual RECOMB/ISCB Conferences with RSG/DREAM organized by the ISCB. While DREAM Challenges have emphasized open science and data, in order to mitigate issues rising from highly sensitive data such as genomics in patient cohorts, "model to data" approaches have been adopted. In such challenges participants submit their models via containers such as Docker or Singularity. This allows retaining confidentiality of the original data as these containers are then run by the organizers on the confidential data. This differs from the more traditional open data model, where participants submit predictions directly based on the provided open data. == Challenge organization == DREAM challenge comprises a core DREAM/Sage Bionetworks organization group as well as an extended scientific expert group, who may have contributed to creation and conception of the challenge or by providing key data. Additionally, new DREAM challenges may be proposed by the wider research community. Pharmaceutical companies or other private entities may also be involved in DREAM challenges, for example in providing data. == Challenge structure == Timelines for key stages (such as introduction webinars, model submission deadlines, and final deadline for participation) are provided in advance. After the winners are announced, organizers start collaborating with the top performing participants to conduct post hoc analyses for a publication describing key findings from the competition. Challenges may be split into sub-challenges, each addressing a different subtopic within the research question. For example, regarding cancer treatment efficacy predictions, these may be separate predictions for progression-free survival, overall survival, best overall response according to RECIST, or exact time until event (progression or death). == Participation == During DREAM challenges, participants typically build models on provided data, and submit predictions or models that are then validated on held-out data by the organizers. While DREAM challenges avoid leaking validation data to participants, there are typically mid-challenge submission leaderboards available to assist participants in evaluating their performance on a sub-sampled or scrambled dataset. DREAM challenges are free for participants. During the open phase anybody can register via Synapse to participate either individually or as a team. A person may only register once and may not use any aliases. There are some exceptions, which disqualify an individual from participating, for example: Person has privileged access to the data for the particular challenge, thus providing them with an unfair advantage. Person has been caught or is under suspicion of cheating or abusing previous DREAM Challenges. Person is a minor (under age 18 or the age of majority in jurisdiction of residence). This may be alleviated via parental consent.
Drools
Drools is a business rule management system (BRMS) with a forward and backward chaining inference-based rules engine, more correctly known as a production rule system, using an enhanced implementation of the Rete algorithm. Drools supports the Java Rules Engine API (Java Specification Request 94) standard for its business rule engine and enterprise framework for the construction, maintenance, and enforcement of business policies in an organization, application, or service. == Drools in Apache Kie == Drools, as part of the Kie Community has entered Apache Incubator in January, 2023. == Red Hat Decision Manager == Red Hat Decision Manager (formerly Red Hat JBoss BRMS) is a business rule management system and reasoning engine for business policy and rules development, access, and change management. JBoss Enterprise BRMS is a productized version of Drools with enterprise-level support available. JBoss Rules is also a productized version of Drools, but JBoss Enterprise BRMS is the flagship product. Components of the enterprise version: JBoss Enterprise Web Platform – the software infrastructure, supported to run the BRMS components only JBoss Enterprise Application Platform or JBoss Enterprise SOA Platform – the software infrastructure, supported to run the BRMS components only Business Rules Engine – Drools Expert using the Rete algorithm and the Drools Rule Language (DRL) Business Rules Manager – Drools Guvnor - Guvnor is a centralized repository for Drools Knowledge Bases, with rich web-based GUIs, editors, and tools to aid in the management of large numbers of rules. Business Rules Repository – Drools Guvnor Drools and Guvnor are JBoss Community open source projects. As they are mature, they are brought into the enterprise-ready product JBoss Enterprise BRMS. Components of the JBoss Community version: Drools Guvnor (Business Rules Manager) – a centralized repository for Drools Knowledge Bases Drools Expert (rule engine) – uses the rules to perform reasoning Drools Flow (process/workflow), or jBPM 5 – provides for workflow and business processes Drools Fusion (event processing/temporal reasoning) – provides for complex event processing Drools Planner/OptaPlanner (automated planning) – optimizes automated planning, including NP-hard planning problems == Example == This example illustrates a simple rule to print out information about a holiday in July. It checks a condition on an instance of the Holiday class, and executes Java code if that condition is true. The purpose of dialect "mvel" is to point the getter and setters of the variables of your Plain Old Java Object (POJO) classes. Consider the above example, in which a Holiday class is used and inside the circular brackets (parentheses) "month" is used. So with the help of dialect "mvel" the getter and setters of the variable "month" can be accessed. Dialect "java" is used to help us write our Java code in our rules. There is one restriction or characteristic on this. We cannot use Java code inside the "when" part of the rule but we can use Java code in the "then" part. We can also declare a Reference variable $h1 without the $ symbol. There is no restriction on this. The main purpose of putting the $ symbol before the variable is to mark the difference between variables of POJO classes and Rules.
Library classification
A library classification is a system used within a library to organize materials, including books, sound and video recordings, electronic materials, etc., both on shelves and in catalogs and indexes. Each item is typically assigned a call number, which identifies the location of the item within the system. Materials can be arranged by many different factors, typically in either a hierarchical tree structure based on the subject or using a faceted classification system, which allows the assignment of multiple classifications to an object, enabling the classifications to be ordered in many ways. == Description == Library classification is an important and crucial aspect in library and information science. It is distinct from scientific classification in that it has as its goal to provide a useful ordering of documents rather than a theoretical organization of knowledge. Although it has the practical purpose of creating a physical ordering of documents, it does generally attempt to adhere to accepted scientific knowledge. Library classification helps to accommodate all the newly published literature in an already created order of arrangement in a filial sequence. Library classification can be defined as the arrangement of books on shelves, or description of them, in the manner which is most useful to those who read with the ultimate aim of grouping similar things together. Library classification is meant to achieve these four purposes: ordering the fields of knowledge in a systematic way, bring related items together in the most helpful sequence, provide orderly access on the shelf, and provide a location for an item on the shelf. Library classification is distinct from the application of subject headings in that classification organizes knowledge into a systematic order, while subject headings provide access to intellectual materials through vocabulary terms that may or may not be organized as a knowledge system. The characteristics that a bibliographic classification demands for the sake of reaching these purposes are: a useful sequence of subjects at all levels, a concise memorable notation, and a host of techniques and devices of number synthesis. == History == Library classifications were preceded by classifications used by bibliographers such as Conrad Gessner. The earliest library classification schemes organized books in broad subject categories. The earliest known library classification scheme is the Pinakes by Callimachus, a scholar at the Library of Alexandria during the third century BC. During the Renaissance and Reformation era, "Libraries were organized according to the whims or knowledge of individuals in charge." This changed the format in which various materials were classified. Some collections were classified by language and others by how they were printed. After the printing revolution in the sixteenth century, the increase in available printed materials made such broad classification unworkable, and more granular classifications for library materials had to be developed in the nineteenth century. In 1627 Gabriel Naudé published a book called Advice on Establishing a Library. At the time, he was working in the private library of Président à mortier Henri de Mesmes II. Mesmes had around 8,000 printed books and many more Greek, Latin and French written manuscripts. Although it was a private library, scholars with references could access it. The purpose of Advice on Establishing a Library was to identify rules for private book collectors to organize their collections in a more orderly way to increase the collection's usefulness and beauty. Naudé developed a classification system based on seven different classes: theology, medicine, jurisprudence, history, philosophy, mathematics, and the humanities. These seven classes would later be increased to twelve. Advice on Establishing a Library was about a private library, but within the same book, Naudé encouraged the idea of public libraries open to all people regardless of their ability to pay for access to the collection. One of the most famous libraries that Naudé helped improve was the Bibliothèque Mazarine in Paris. Naudé spent ten years there as a librarian. Because of Naudé's strong belief in free access to libraries to all people, the Bibliothèque Mazarine became the first public library in France around 1644. Although libraries created order within their collections from as early as the fifth century BC, the Paris Bookseller's classification, developed in 1842 by Jacques Charles Brunet, is generally seen as the first of the modern book classifications. Brunet provided five major classes: theology, jurisprudence, sciences and arts, belles-lettres, and history. Classification can now be seen as a provider of subject access to information in a networked environment. == Types == There are many standard systems of library classification in use, and many more have been proposed over the years. However, in general, classification systems can be divided into three types depending on how they are used: === Universal schemes === Covers all subjects, e.g. the Dewey Decimal Classification (DDC), Universal Decimal Classification (UDC), and Colon Classification (CC). === Specific classification schemes === Covers particular subjects or types of materials, e.g. Iconclass (art), British Catalogue of Music Classification, and Dickinson classification (music), or the NLM Classification (medicine). === National schemes === Specially created for certain countries, e.g. Swedish library classification system, SAB (Sveriges Allmänna Biblioteksförening). The Library of Congress Classification was designed around the collection of the US Library of Congress and has an American, European, and Christian bias. Nevertheless, it is used widely in large academic and research libraries. In terms of functionality, classification systems are often described as: === Enumerative === Subject headings are listed alphabetically, with numbers assigned to each heading in alphabetical order. === Hierarchical === Subjects are divided hierarchically, from most general to most specific. === Faceted/analytico-synthetic === Subjects are divided into mutually exclusive orthogonal facets. There are few completely enumerative systems or faceted systems; most systems are a blend but favouring one type or the other. The most common classification systems, LCC and DDC, are essentially enumerative, though with some hierarchical and faceted elements (more so for DDC), especially at the broadest and most general level. The first true faceted system was the colon classification of S. R. Ranganathan. == Methods or systems == Classification types denote the classification or categorization according to the form or characteristics or qualities of a classification scheme or schemes. Method and system has similar meaning. Method or methods or system means the classification schemes like Dewey Decimal Classification or Universal Decimal Classification. The types of classification is for identifying and understanding or education or research purposes while classification method means those classification schemes like DDC, UDC. === English language universal classification systems === The most common systems in English-speaking countries are: Dewey Decimal Classification (DDC) Library of Congress Classification (LCC) Universal Decimal Classification (UDC) Other systems include: Book Industry Standards and Communications (BISAC), originally developed for use by U.S. booksellers, has become increasingly popular in libraries. Bliss bibliographic classification used in some British libraries Colon classification (CC) Garside classification used in most libraries of University College London Gladstone Library Classification, devised by W.E. Gladstone and used exclusively at Gladstone's Library Harvard-Yenching Classification, an English classification system for Chinese language materials === Non-English universal classification systems === German Regensburger Verbundklassifikation (RVK) A system of book classification for Chinese libraries (Liu's Classification) library classification for user New Classification Scheme for Chinese Libraries Nippon Decimal Classification (NDC) Chinese Library Classification (CLC) Korean Decimal Classification (KDC) Russian Library-Bibliographical Classification (BBK) Swedish library classification system (SAB) === Universal classification systems that rely on synthesis (faceted systems) === Bliss bibliographic classification Colon classification Cutter Expansive Classification Universal Decimal Classification Newer classification systems tend to use the principle of synthesis (combining codes from different lists to represent the different attributes of a work) heavily, which is comparatively lacking in LC or DDC. == Practice == Library classification is associated with library (descriptive) cataloging under the rubric of cataloging and classification, sometimes grouped together as technical serv
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.
Six Little Dragons
Six Little Dragons (Chinese: 杭州六小龙), or Six Little Dragons of Hangzhou, are an informal grouping of the tech startups Game Science, DeepSeek, Unitree Robotics, DEEP Robotics, BrainCo and Manycore Tech. All six were established in Hangzhou, They are active in artificial intelligence, robotics, gaming, and brain-computer interface technology. Hangzhou is referred to as the China’s “e-commerce capital” (电商之都). The nickname "Six Little Dragons" originated from the Chinese internet. == Background == === Chinese government investments (2002 — 2010s) === From 2002 to 2007, under Xi Jinping's leadership as party secretary of Zhejiang, provincial spending on technology research grew over four times to 28 billion RMB. The province launched "Digital Zhejiang" (数字浙江) to advance modernization and the "Eight Eight Strategy" (八八战略), focusing on eight advantages and actions to boost industrial development, including specialized industries. In 2010, Hangzhou's government started "Project Eagle" (雏鹰计划) to aid science and technology startups. The project works with incubators and accelerators to find promising tech companies and offers public funding and other help, especially for startups by graduates and returning students. Unitree received support in the initial phase, along with government subsidies from Binjiang District. === AI-startups and further investments (2025 — present) === In January 2025, the Chinese government created the "Hangzhou AI Industry Chain High-Quality Development Action Plan" which focuses on computing power, LLM technologies, and AI applications. The plan was made to certify over 2,000 new high-tech enterprises, initiate over 300 major tech projects, and invest more than 300 billion RMB (US$40 billion) annually. The Chinese government also renewed "Project Eagle" and to allocate 15% of industrial policy funds for future industries. Hangzhou aimed to become a center for tech startups, highlighting the "six little dragons of Hangzhou," a nickname popularized in early 2025. This group includes DeepSeek, Game Science, Unitree Robotics, Manycore Tech, BrainCo, and DEEP Robotics, companies in gaming, robotics, and software development. Earlier in 2025, DeepSeek, one of the six dragons, launched an AI system at a much lower cost than those from Silicon Valley. Since then, DeepSeek and Alibaba have produced top-performing open source AI models. Game Science launched the successful video game Black Myth: Wukong in 2024, while Unitree gained attention for their dancing robots in the 2025 annual spring gala broadcast by Chinese state media. The group was acknowledged by Chinese authorities in Hangzhou in a New Years message for local businesses in January 2025. Hangzhou’s universities were given credit for the development of Chinese technological industry. Zhejiang University alumni founded three of the "Six Little Dragons". By September 2024, the university produced 102 executives in Chinese AI start-ups, ranking third among China's top institutions. On February 20, 2025, Alibaba's Eddie Wu stated that the company would focus on artificial generative intelligence and plans significant investment in AI. The company also sought to boost foreign investment to China's "Six Little Dragons" following Alibaba's founder Jack Ma attended General Secretary of the Chinese Communist Party Xi Jinping's business symposium with corporate leaders and entrepreneurs that same month. == Challenges == China's net foreign direct investment (FDI) fell by US$168 billion in 2024, marking the largest capital flight since 1990. Foreign investment peaked at US$344 billion in 2021 but has since declined according to the State Administration of Foreign Exchange. In 2024, foreign investors put in only US$4.5 billion while Chinese firms invested US$173 billion abroad. According to interviews conducted by The New York Times, some start-up company founders believe that Chinese government's support for Hangzhou's technological sector has deterred foreign investors. Tensions with the United States led many international companies to adopt a China Plus One strategy, while Chinese firms build factories overseas to avoid potential Trump tariffs. China also faced US restrictions on its access of advanced chips, forcing Chinese tech companies to stockpile Nvidia chips while Chinese producers like Huawei and Semiconductor Manufacturing International Corporation (SMIC) were competing to produce their own.