AI Assistant Examples

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

• 

  Are You Dead?

  Are You Dead? (Chinese: 死了么; pinyin: Sǐleme), also known by its English name Demumu, is a Chinese application designed for young people living alone. It requires setting up one emergency contact and sends automatic notifications if the user has not checked in via the app for consecutive days. The app was released on the App Store on 10 June 2025. In early January 2026, the application gained popularity due to its name and the issue of safety for people living alone, and ranked high on the list of paid applications in the Chinese region of the Apple App Store before being removed. The app's rise in popularity sparked discussions about taboos about death in China. == History == Are You Dead? was founded and operated independently by three people born in the 1990s, and developed in a way that involved remote collaboration in their spare time. According to the New Yellow River report, Guo, the product manager, said that the application was designed for young people and that the inspiration came from the discussion of netizens on social platforms about "an app that everyone must have and will definitely download" that he observed two or three years ago. The name was also "not their original creation". After realizing its potential demand and social significance, the team successfully registered the name and completed the product development in about a month. Regarding the development entity, the New Yellow River cited information from the Apple App Store that the application was developed by Yuejing (Zhengzhou) Technology Service Co., Ltd. According to Tianyancha information, the company was established in March 2025 with a registered capital of 100,000 yuan. === Rise in popularity === The app has been generating buzz on social media since 9 January 2026, due to its name and the topic of safety for people living alone. Around 10 January, it topped the Apple paid app chart. As of 10:00 a.m. on January 11, it ranked first in the App Store paid app chart. It also ranked highly in the utility app chart; it ranked first or second in the paid utility app charts in the United States, Singapore and Hong Kong, and first or fourth in Australia and Spain. The app was subsequently removed from the Apple App Store in China. In terms of functionality and usage, First Financial praised the product for its "simple interface and single function," but pointed out that the interface lacks a display of consecutive check-in days, and there is also the possibility that users may forget to check in, leading to the mistaken issuance of reminders. In addition, since the application mainly relies on email reminders and lacks SMS or telephone notifications, it does not conform to Chinese social habits; the untimely notifications also make the application more like a "death notification" tool, losing its early warning significance for emergency rescue. Hu Xijin, former editor-in-chief of the Global Times, commented on the application on Weibo that it is "really good and can help many lonely elderly people." The Beijing News Quick Review pointed out that the role of technical tools is limited and needs to be connected with real support such as community patrols and liaison mechanisms. Due to the price increase, there have also been questions about the motivation for the price increase. The app's rise in popularity sparked discussions about taboos about death in China. Regarding the popularity of the application, both Southern Metropolis Daily and The Beijing News commented that it reflects the public issue of the risks of living alone and reflects the general anxiety of the living alone group about dying alone. Shangguan News further pointed out that although such technology products provide a certain "low-cost sense of security", their "cold notifications" may not only cause false alarms, but also highlight the embarrassing reality that "there is no one to fill in the emergency contact". It also emphasized that algorithms or applications cannot bring true happiness and called on society to reconstruct a support network full of humanistic care while relying on technology. The name of the application has also sparked controversy. Most netizens believe that the name "Are You Dead?" is unlucky and makes it awkward to share the application. They suggest changing it to a milder name such as "Are You Alive?". Hu Xijin also said that the name change could "give the elderly who use it more psychological comfort" and "believe that the application will become more popular after the name change". Some people also believe that this straightforward name just points out the real dilemma faced by people living alone and has a special meaning. BBC News commented that the name "Are You Dead" is playing a word game with Ele.me (Chinese: 饿了么; pinyin: Èleme) and the pronunciation is also similar. Legal professionals believe that its name is highly similar to Ele.me and may cause confusion. They also raised the possibility of trademark infringement and unfair competition. However, the developers said that the application is developed for young people and death is not a sensitive topic. They will "consider launching a new application that is more suitable for middle-aged and elderly people". They have not yet received any name change requests from relevant departments. On the evening of 13 January 2026, the Are You Dead? team announced that it would change its name to the English brand name Demumu in the upcoming new version. On 11 January, the development team also issued a statement through its official Weibo account, stating that it would study the renaming suggestion and plan to enrich the SMS reminder function, consider adding the message function and explore the direction of age-friendly products; it also stated that it would launch an 8 yuan paid plan to cover the costs of SMS, servers, etc., and welcomed investors to discuss cooperation. In terms of financing and valuation, it plans to sell 10% of the company's shares for 1 million yuan and proposed a valuation of 10 million yuan. On the evening of January 15, the application was removed from the app store in mainland China. == Functions == The application does not require users to enter phone numbers or other information to register. After filling in their name and setting an emergency contact, users can click the sign-in button every day. If they fail to sign in for two consecutive days, the system will send an email reminder to the emergency contact the next day. In addition, users can also bind a smart bracelet to monitor physiological signs, pre-designate a hearse driver and funeral music, and trigger the "one-click body collection" function when no pulse is detected. The application was initially available for free download, but a one yuan paid download option was introduced at the end of 2025. In January 2026, the application team issued a statement saying that an 8 yuan paid option would be launched based on the costs of SMS, servers, etc.

  Read more →
• 

  Neuroph

  Neuroph is an object-oriented artificial neural network framework written in Java. It can be used to create and train neural networks in Java programs. Neuroph provides Java class library as well as GUI tool easyNeurons for creating and training neural networks. It is an open-source project hosted at SourceForge under the Apache License. Versions before 2.4 were licensed under LGPL 3, from this version the license is Apache 2.0 License. == Features == Neuroph's core classes correspond to basic neural network concepts like artificial neuron, neuron layer, neuron connections, weight, transfer function, input function, learning rule etc. Neuroph supports common neural network architectures such as Multilayer perceptron with Backpropagation, Kohonen and Hopfield networks. All these classes can be extended and customized to create custom neural networks and learning rules. Neuroph has built-in support for image recognition.

  Read more →
• 

  Tomáš Mikolov

  Tomáš Mikolov is a Czech computer scientist working in the field of machine learning. In March 2020, Mikolov became a senior research scientist at the Czech Institute of Informatics, Robotics and Cybernetics. == Career == Mikolov obtained his PhD in Computer Science from Brno University of Technology for his work on recurrent neural network-based language models. He is the lead author of the 2013 paper that introduced the Word2vec technique in natural language processing and is an author on the FastText architecture. Mikolov came up with the idea to generate text from neural language models in 2007 and his RNNLM toolkit was the first to demonstrate the capability to train language models on large corpora, resulting in large improvements over the state of the art. Prior to joining Facebook in 2014, Mikolov worked as a visiting researcher at Johns Hopkins University, Université de Montréal, Microsoft and Google. He left Facebook at some time in 2019/2020 to join the Czech Institute of Informatics, Robotics and Cybernetics. Mikolov has argued that humanity might be at a greater existential risk if an artificial general intelligence is not developed.

  Read more →
• 

  Transfer-based machine translation

  Transfer-based machine translation is a type of machine translation (MT). It is currently one of the most widely used methods of machine translation. In contrast to the simpler direct model of MT, transfer MT breaks translation into three steps: analysis of the source language text to determine its grammatical structure, transfer of the resulting structure to a structure suitable for generating text in the target language, and finally generation of this text. Transfer-based MT systems are thus capable of using knowledge of the source and target languages. == Design == Both transfer-based and interlingua-based machine translation have the same idea: to make a translation it is necessary to have an intermediate representation that captures the "meaning" of the original sentence in order to generate the correct translation. In interlingua-based MT this intermediate representation must be independent of the languages in question, whereas in transfer-based MT, it has some dependence on the language pair involved. The way in which transfer-based machine translation systems work varies substantially, but in general they follow the same pattern: they apply sets of linguistic rules which are defined as correspondences between the structure of the source language and that of the target language. The first stage involves analysing the input text for morphology and syntax (and sometimes semantics) to create an internal representation. The translation is generated from this representation using both bilingual dictionaries and grammatical rules. It is possible with this translation strategy to obtain fairly high quality translations, with accuracy in the region of 90% (although this is highly dependent on the language pair in question, for example the distance between the two). == Operation == In a rule-based machine translation system the original text is first analysed morphologically and syntactically in order to obtain a syntactic representation. This representation can then be refined to a more abstract level putting emphasis on the parts relevant for translation and ignoring other types of information. The transfer process then converts this final representation (still in the original language) to a representation of the same level of abstraction in the target language. These two representations are referred to as "intermediate" representations. From the target language representation, the stages are then applied in reverse. == Analysis and transformation == Various methods of analysis and transformation can be used before obtaining the final result. Along with these statistical approaches may be augmented generating hybrid systems. The methods which are chosen and the emphasis depends largely on the design of the system, however, most systems include at least the following stages: Morphological analysis. Surface forms of the input text are classified as to part-of-speech (e.g. noun, verb, etc.) and sub-category (number, gender, tense, etc.). All of the possible "analyses" for each surface form are typically made output at this stage, along with the lemma of the word. Lexical categorisation. In any given text some of the words may have more than one meaning, causing ambiguity in analysis. Lexical categorisation looks at the context of a word to try to determine the correct meaning in the context of the input. This can involve part-of-speech tagging and word sense disambiguation. Lexical transfer. This is basically dictionary translation; the source language lemma (perhaps with sense information) is looked up in a bilingual dictionary and the translation is chosen. Structural transfer. While the previous stages deal with words, this stage deals with larger constituents, for example phrases and chunks. Typical features of this stage include concordance of gender and number, and re-ordering of words or phrases. Morphological generation. From the output of the structural transfer stage, the target language surface forms are generated. == Transfer types == One of the main features of transfer-based machine translation systems is a phase that "transfers" an intermediate representation of the text in the original language to an intermediate representation of text in the target language. This can work at one of two levels of linguistic analysis, or somewhere in between. The levels are: Superficial transfer (or syntactic). This level is characterised by transferring "syntactic structures" between the source and target languages. It is suitable for languages in the same family or of the same type, for example in the Romance languages between Spanish, Catalan, French, Italian, etc. Deep transfer (or semantic). This level constructs a semantic representation that is dependent on the source language. This representation can consist of a series of structures which represent the meaning. In these transfer systems predicates are typically produced. The translation also typically requires structural transfer. This level is used to translate between more distantly related languages (e.g. Spanish-English or Spanish-Basque, etc.)

  Read more →
• 

  KeyBase

  KeyBase is a database and web application for managing and deploying interactive taxonomic keys for plants and animals developed by the Royal Botanic Gardens Victoria. KeyBase provides a medium where pathway keys which were traditionally developed for print and other classical types of media, can be used more effectively in the internet environment. The platform uses a concept called "keys" which can be easily linked together, joined with other keys, or merged into larger other seamless keys groups, with each still available to be browsed independently. Keys in the KeyBase database can be filtered and displayed in a variety of ways, filters, and formats.

  Read more →
• 

  The Best Free AI Video Generator for Beginners

  Trying to pick the best AI video generator? An AI video generator is software that uses machine learning to help you get more done — it scales effortlessly from a single task to thousands. The best picks balance beginner-friendly simplicity with the depth power users need, and they ship updates often. Whether you are a beginner or a pro, the right AI video generator slots into your workflow and pays for itself fast. Read on for hands-on impressions, pricing tiers, and the standout features that matter.

  Read more →
• 

  Top 10 AI Essay Writers Compared (2026)

  Curious about the best AI essay writer? An AI essay writer is software that uses machine learning to help you get more done — it combines speed, accuracy, and an interface that just works. Hands-on testing shows real-world results vary, so a short free trial is the smartest way to decide. Whether you are a beginner or a pro, the right AI essay writer slots into your workflow and pays for itself fast. This guide breaks down the top picks, their pros and cons, and who each one is best for.

  Read more →
• 

  Comparison of machine translation applications

  Machine translation is an algorithm which attempts to translate text or speech from one natural language to another. == General information == Basic general information for popular machine translation applications. == Languages features comparison == The following table compares the number of languages which the following machine translation programs can translate between. (Moses and Moses for Mere Mortals allow you to train translation models for any language pair, though collections of translated texts (parallel corpus) need to be provided by the user. The Moses site provides links to training corpora.) This is not an all-encompassing list. Some applications have many more language pairs than those listed below. This is a general comparison of key languages only. A full and accurate list of language pairs supported by each product should be found on each of the product's websites. === Multi-pair translations === === Paired translations ===

  Read more →
• 

  Box blur

  A box blur (also known as a box linear filter) is a spatial domain linear filter in which each pixel in the resulting image has a value equal to the average value of its neighboring pixels in the input image. It is a form of low-pass ("blurring") filter. A 3 by 3 box blur ("radius 1") can be written as matrix 1 9 [ 1 1 1 1 1 1 1 1 1 ] . {\displaystyle {\frac {1}{9}}{\begin{bmatrix}1&1&1\\1&1&1\\1&1&1\end{bmatrix}}.} Due to its property of using equal weights, it can be implemented using a much simpler accumulation algorithm, which is significantly faster than using a sliding-window algorithm. Box blurs are frequently used to approximate a Gaussian blur. By the central limit theorem, repeated application of a box blur will approximate a Gaussian blur. In the frequency domain, a box blur has zeros and negative components. That is, a sine wave with a period equal to the size of the box will be blurred away entirely, and wavelengths shorter than the size of the box may be phase-reversed, as seen when two bokeh circles touch to form a bright spot where there would be a dark spot between two bright spots in the original image. == Extensions == Gwosdek, et al. has extended Box blur to take a fractional radius: the edges of the 1-D filter are expanded with a fraction. It makes slightly better gaussian approximation possible due to the elimination of integer-rounding error. Mario Klingemann has a "stack blur" that tries to better emulate gaussian's look in one pass by stacking weights: 1 9 [ 1 2 3 2 1 ] {\displaystyle {\frac {1}{9}}{\begin{bmatrix}1&2&3&2&1\end{bmatrix}}} The triangular impulse response it forms decomposes to two rounds of box blur. Stacked Integral Image by Bhatia et al. takes the weighted average of a few box blurs to fit the gaussian response curve. == Implementation == The following pseudocode implements a 3x3 box blur. The example does not handle the edges of the image, which would not fit inside the kernel, so that these areas remain unblurred. In practice, the issue is better handled by: Introducing an alpha channel to represent the absence of colors; Extending the boundary by filling in values, ranked by quality: Fill in a mirrored image at the border Fill in a constant color extending from the last pixel Pad in a fixed color A number of optimizations can be applied when implementing the box blur of a radius r and N pixels: The box blur is a separable filter, so that only two 1D passes of averaging 2 r + 1 pixels will be needed, one horizontal and one vertical, for each pixel. This lowers the complexity from O(Nr2) to O(Nr). In digital signal processing terminology, each pass is a moving-average filter. Accumulation. Instead of discarding the sum for each pixel, the algorithm re-uses the previous sum, and updates it by subtracting away the old pixel and adding the new pixel in the blurring range. A summed-area table can be used similarly. This lowers the complexity from O(Nr) to O(N). When being used in multiple passes to approximate a Gaussian blur, the cascaded integrator–comb filter construction allows for doing the equivalent operation in a single pass.

  Read more →
• 

  Is an AI Resume Builder Worth It in 2026?

  Looking for the best AI resume builder? An AI resume builder is software that uses machine learning to help you get more done — it can save you hours every week by automating repetitive work. Most options offer a generous free tier, with paid plans unlocking higher limits, faster processing, and team features. Whether you are a beginner or a pro, the right AI resume builder slots into your workflow and pays for itself fast. This guide breaks down the top picks, their pros and cons, and who each one is best for.

  Read more →
• 

  Markov chain central limit theorem

  In the mathematical theory of random processes, the Markov chain central limit theorem has a conclusion somewhat similar in form to that of the classic central limit theorem (CLT) of probability theory, but the quantity in the role taken by the variance in the classic CLT has a more complicated definition. See also the general form of Bienaymé's identity. == Statement == Suppose that: the sequence X 1 , X 2 , X 3 , … {\textstyle X_{1},X_{2},X_{3},\ldots } of random elements of some set is a Markov chain that has a stationary probability distribution; and the initial distribution of the process, i.e. the distribution of X 1 {\textstyle X_{1}} , is the stationary distribution, so that X 1 , X 2 , X 3 , … {\textstyle X_{1},X_{2},X_{3},\ldots } are identically distributed. In the classic central limit theorem these random variables would be assumed to be independent, but here we have only the weaker assumption that the process has the Markov property; and g {\textstyle g} is some (measurable) real-valued function for which var ⁡ ( g ( X 1 ) ) < + ∞ . {\textstyle \operatorname {var} (g(X_{1}))<+\infty .} Now let μ = E ⁡ ( g ( X 1 ) ) , μ ^ n = 1 n ∑ k = 1 n g ( X k ) σ 2 := lim n → ∞ var ⁡ ( n μ ^ n ) = lim n → ∞ n var ⁡ ( μ ^ n ) = var ⁡ ( g ( X 1 ) ) + 2 ∑ k = 1 ∞ cov ⁡ ( g ( X 1 ) , g ( X 1 + k ) ) . {\displaystyle {\begin{aligned}\mu &=\operatorname {E} (g(X_{1})),\\{\widehat {\mu }}_{n}&={\frac {1}{n}}\sum _{k=1}^{n}g(X_{k})\\\sigma ^{2}&:=\lim _{n\to \infty }\operatorname {var} ({\sqrt {n}}{\widehat {\mu }}_{n})=\lim _{n\to \infty }n\operatorname {var} ({\widehat {\mu }}_{n})=\operatorname {var} (g(X_{1}))+2\sum _{k=1}^{\infty }\operatorname {cov} (g(X_{1}),g(X_{1+k})).\end{aligned}}} Then as n → ∞ , {\textstyle n\to \infty ,} we have n ( μ ^ n − μ ) → D Normal ( 0 , σ 2 ) , {\displaystyle {\sqrt {n}}({\hat {\mu }}_{n}-\mu )\ {\xrightarrow {\mathcal {D}}}\ {\text{Normal}}(0,\sigma ^{2}),} where the decorated arrow indicates convergence in distribution. == Monte Carlo Setting == The Markov chain central limit theorem can be guaranteed for functionals of general state space Markov chains under certain conditions. In particular, this can be done with a focus on Monte Carlo settings. An example of the application in a MCMC (Markov Chain Monte Carlo) setting is the following: Consider a simple hard spheres model on a grid. Suppose X = { 1 , … , n 1 } × { 1 , … , n 2 } ⊆ Z 2 {\displaystyle X=\{1,\ldots ,n_{1}\}\times \{1,\ldots ,n_{2}\}\subseteq Z^{2}} . A proper configuration on X {\displaystyle X} consists of coloring each point either black or white in such a way that no two adjacent points are white. Let χ {\displaystyle \chi } denote the set of all proper configurations on X {\displaystyle X} , N χ ( n 1 , n 2 ) {\displaystyle N_{\chi }(n_{1},n_{2})} be the total number of proper configurations and π be the uniform distribution on χ {\displaystyle \chi } so that each proper configuration is equally likely. Suppose our goal is to calculate the typical number of white points in a proper configuration; that is, if W ( x ) {\displaystyle W(x)} is the number of white points in x ∈ χ {\displaystyle x\in \chi } then we want the value of E π W = ∑ x ∈ χ W ( x ) N χ ( n 1 , n 2 ) {\displaystyle E_{\pi }W=\sum _{x\in \chi }{\frac {W(x)}{N_{\chi }{\bigl (}n_{1},n_{2}{\bigr )}}}} If n 1 {\displaystyle n_{1}} and n 2 {\displaystyle n_{2}} are even moderately large then we will have to resort to an approximation to E π W {\displaystyle E_{\pi }W} . Consider the following Markov chain on χ {\displaystyle \chi } . Fix p ∈ ( 0 , 1 ) {\displaystyle p\in (0,1)} and set X 1 = x 1 {\displaystyle X_{1}=x_{1}} where x 1 ∈ χ {\displaystyle x_{1}\in \chi } is an arbitrary proper configuration. Randomly choose a point ( x , y ) ∈ X {\displaystyle (x,y)\in X} and independently draw U ∼ U n i f o r m ( 0 , 1 ) {\displaystyle U\sim \mathrm {Uniform} (0,1)} . If u ≤ p {\displaystyle u\leq p} and all of the adjacent points are black then color ( x , y ) {\displaystyle (x,y)} white leaving all other points alone. Otherwise, color ( x , y ) {\displaystyle (x,y)} black and leave all other points alone. Call the resulting configuration X 1 {\displaystyle X_{1}} . Continuing in this fashion yields a Harris ergodic Markov chain { X 1 , X 2 , X 3 , … } {\displaystyle \{X_{1},X_{2},X_{3},\ldots \}} having π {\displaystyle \pi } as its invariant distribution. It is now a simple matter to estimate E π W {\displaystyle E_{\pi }W} with w n ¯ = ∑ i = 1 n W ( X i ) / n {\displaystyle {\overline {w_{n}}}=\sum _{i=1}^{n}W(X_{i})/n} . Also, since χ {\displaystyle \chi } is finite (albeit potentially large) it is well known that X {\displaystyle X} will converge exponentially fast to π {\displaystyle \pi } which implies that a CLT holds for w n ¯ {\displaystyle {\overline {w_{n}}}} . == Implications == Not taking into account the additional terms in the variance which stem from correlations (e.g. serial correlations in markov chain monte carlo simulations) can result in the problem of pseudoreplication when computing e.g. the confidence intervals for the sample mean.

  Read more →
• 

  MRF optimization via dual decomposition

  In dual decomposition a problem is broken into smaller subproblems and a solution to the relaxed problem is found. This method can be employed for MRF optimization. Dual decomposition is applied to markov logic programs as an inference technique. == Background == Discrete MRF Optimization (inference) is very important in Machine Learning and Computer vision, which is realized on CUDA graphical processing units. Consider a graph G = ( V , E ) {\displaystyle G=(V,E)} with nodes V {\displaystyle V} and Edges E {\displaystyle E} . The goal is to assign a label l p {\displaystyle l_{p}} to each p ∈ V {\displaystyle p\in V} so that the MRF Energy is minimized: (1) min Σ p ∈ V θ p ( l p ) + Σ p q ∈ ε θ p q ( l p ) ( l q ) {\displaystyle \min \Sigma _{p\in V}\theta _{p}(l_{p})+\Sigma _{pq\in \varepsilon }\theta _{pq}(l_{p})(l_{q})} Major MRF Optimization methods are based on Graph cuts or Message passing. They rely on the following integer linear programming formulation (2) min x E ( θ , x ) = θ . x = ∑ p ∈ V θ p . x p + ∑ p q ∈ ε θ p q . x p q {\displaystyle \min _{x}E(\theta ,x)=\theta .x=\sum _{p\in V}\theta _{p}.x_{p}+\sum _{pq\in \varepsilon }\theta _{pq}.x_{pq}} In many applications, the MRF-variables are {0,1}-variables that satisfy: x p ( l ) = 1 {\displaystyle x_{p}(l)=1} ⇔ {\displaystyle \Leftrightarrow } label l {\displaystyle l} is assigned to p {\displaystyle p} , while x p q ( l , l ′ ) = 1 {\displaystyle x_{pq}(l,l^{\prime })=1} , labels l , l ′ {\displaystyle l,l^{\prime }} are assigned to p , q {\displaystyle p,q} . == Dual Decomposition == The main idea behind decomposition is surprisingly simple: decompose your original complex problem into smaller solvable subproblems, extract a solution by cleverly combining the solutions from these subproblems. A sample problem to decompose: min x Σ i f i ( x ) {\displaystyle \min _{x}\Sigma _{i}f^{i}(x)} where x ∈ C {\displaystyle x\in C} In this problem, separately minimizing every single f i ( x ) {\displaystyle f^{i}(x)} over x {\displaystyle x} is easy; but minimizing their sum is a complex problem. So the problem needs to get decomposed using auxiliary variables { x i } {\displaystyle \{x^{i}\}} and the problem will be as follows: min { x i } , x Σ i f i ( x i ) {\displaystyle \min _{\{x^{i}\},x}\Sigma _{i}f^{i}(x^{i})} where x i ∈ C , x i = x {\displaystyle x^{i}\in C,x^{i}=x} Now we can relax the constraints by multipliers { λ i } {\displaystyle \{\lambda ^{i}\}} which gives us the following Lagrangian dual function: g ( { λ i } ) = min { x i ∈ C } , x Σ i f i ( x i ) + Σ i λ i . ( x i − x ) = min { x i ∈ C } , x Σ i [ f i ( x i ) + λ i . x i ] − ( Σ i λ i ) x {\displaystyle g(\{\lambda ^{i}\})=\min _{\{x^{i}\in C\},x}\Sigma _{i}f^{i}(x^{i})+\Sigma _{i}\lambda ^{i}.(x^{i}-x)=\min _{\{x^{i}\in C\},x}\Sigma _{i}[f^{i}(x^{i})+\lambda ^{i}.x^{i}]-(\Sigma _{i}\lambda ^{i})x} Now we eliminate x {\displaystyle x} from the dual function by minimizing over x {\displaystyle x} and dual function becomes: g ( { λ i } ) = min { x i ∈ C } Σ i [ f i ( x i ) + λ i . x i ] {\displaystyle g(\{\lambda ^{i}\})=\min _{\{x^{i}\in C\}}\Sigma _{i}[f^{i}(x^{i})+\lambda ^{i}.x^{i}]} We can set up a Lagrangian dual problem: (3) max { λ i } ∈ Λ g ( λ i ) = Σ i g i ( x i ) , {\displaystyle \max _{\{\lambda ^{i}\}\in \Lambda }g({\lambda ^{i}})=\Sigma _{i}g^{i}(x^{i}),} The Master problem (4) g i ( x i ) = m i n x i f i ( x i ) + λ i . x i {\displaystyle g^{i}(x^{i})=min_{x^{i}}f^{i}(x^{i})+\lambda ^{i}.x^{i}} where x i ∈ C {\displaystyle x^{i}\in C} The Slave problems == MRF optimization via Dual Decomposition == The original MRF optimization problem is NP-hard and we need to transform it into something easier. τ {\displaystyle \tau } is a set of sub-trees of graph G {\displaystyle G} where its trees cover all nodes and edges of the main graph. And MRFs defined for every tree T {\displaystyle T} in τ {\displaystyle \tau } will be smaller. The vector of MRF parameters is θ T {\displaystyle \theta ^{T}} and the vector of MRF variables is x T {\displaystyle x^{T}} , these two are just smaller in comparison with original MRF vectors θ , x {\displaystyle \theta ,x} . For all vectors θ T {\displaystyle \theta ^{T}} we'll have the following: (5) ∑ T ∈ τ ( p ) θ p T = θ p , ∑ T ∈ τ ( p q ) θ p q T = θ p q . {\displaystyle \sum _{T\in \tau (p)}\theta _{p}^{T}=\theta _{p},\sum _{T\in \tau (pq)}\theta _{pq}^{T}=\theta _{pq}.} Where τ ( p ) {\displaystyle \tau (p)} and τ ( p q ) {\displaystyle \tau (pq)} denote all trees of τ {\displaystyle \tau } than contain node p {\displaystyle p} and edge p q {\displaystyle pq} respectively. We simply can write: (6) E ( θ , x ) = ∑ T ∈ τ E ( θ T , x T ) {\displaystyle E(\theta ,x)=\sum _{T\in \tau }E(\theta ^{T},x^{T})} And our constraints will be: (7) x T ∈ χ T , x T = x | T , ∀ T ∈ τ {\displaystyle x^{T}\in \chi ^{T},x^{T}=x_{|T},\forall T\in \tau } Our original MRF problem will become: (8) min { x T } , x Σ T ∈ τ E ( θ T , x T ) {\displaystyle \min _{\{x^{T}\},x}\Sigma _{T\in \tau }E(\theta ^{T},x^{T})} where x T ∈ χ T , ∀ T ∈ τ {\displaystyle x^{T}\in \chi ^{T},\forall T\in \tau } and x T ∈ x | T , ∀ T ∈ τ {\displaystyle x^{T}\in x_{|T},\forall T\in \tau } And we'll have the dual problem we were seeking: (9) max { λ T } ∈ Λ g ( { λ T } ) = ∑ T ∈ τ g T ( λ T ) , {\displaystyle \max _{\{\lambda ^{T}\}\in \Lambda }g(\{\lambda ^{T}\})=\sum _{T\in \tau }g^{T}(\lambda ^{T}),} The Master problem where each function g T ( . ) {\displaystyle g^{T}(.)} is defined as: (10) g T ( λ T ) = min x T E ( θ T + λ T , x T ) {\displaystyle g^{T}(\lambda ^{T})=\min _{x^{T}}E(\theta ^{T}+\lambda ^{T},x^{T})} where x T ∈ χ T {\displaystyle x^{T}\in \chi ^{T}} The Slave problems == Theoretical Properties == Theorem 1. Lagrangian relaxation (9) is equivalent to the LP relaxation of (2). min { x T } , x { E ( x , θ ) | x p T = s p , x T ∈ CONVEXHULL ( χ T ) } {\displaystyle \min _{\{x^{T}\},x}\{E(x,\theta )|x_{p}^{T}=s_{p},x^{T}\in {\text{CONVEXHULL}}(\chi ^{T})\}} Theorem 2. If the sequence of multipliers { α t } {\displaystyle \{\alpha _{t}\}} satisfies α t ≥ 0 , lim t → ∞ α t = 0 , ∑ t = 0 ∞ α t = ∞ {\displaystyle \alpha _{t}\geq 0,\lim _{t\to \infty }\alpha _{t}=0,\sum _{t=0}^{\infty }\alpha _{t}=\infty } then the algorithm converges to the optimal solution of (9). Theorem 3. The distance of the current solution { θ T } {\displaystyle \{\theta ^{T}\}} to the optimal solution { θ ¯ T } {\displaystyle \{{\bar {\theta }}^{T}\}} , which decreases at every iteration. Theorem 4. Any solution obtained by the method satisfies the WTA (weak tree agreement) condition. Theorem 5. For binary MRFs with sub-modular energies, the method computes a globally optimal solution.

  Read more →
• 

  Neural processing unit

  A neural processing unit (NPU), also known as an AI accelerator or deep learning processor, is a class of specialized hardware accelerator or computer system designed to accelerate artificial intelligence and machine learning applications, including artificial neural networks and computer vision. == Use == Their purpose is either to efficiently execute already trained AI models (inference) or to train AI models. NPUs can be more efficient in terms of speed or power consumption. NPU applications include algorithms for robotics, Internet of things, and data-intensive or sensor-driven tasks. They are often manycore or spatial designs and focus on low-precision arithmetic, novel dataflow architectures, or in-memory computing capability. As of 2024, a widely used datacenter-grade AI integrated circuit chip, the Nvidia H100 GPU, contains tens of billions of MOSFETs. === Consumer devices === AI accelerators are used in Apple silicon, Qualcomm, Samsung, Huawei, and Google Tensor smartphone processors. Vision processing units are accelerators specialized for machine vision algorithms such as CNN (convolutional neural networks) and SIFT (scale-invariant feature transform). They are used in devices that need to keep track of objects visually such as AR headsets and drones. It is more recently (circa 2017) added to processors from Apple and (circa 2022) to processors from Intel and AMD. All models of Intel Meteor Lake processors have a built-in versatile processor unit (VPU) for accelerating inference for computer vision and deep learning. On consumer devices, the NPU is intended to be small, power-efficient, but reasonably fast when used to run small models. To do this they are designed to support low-bitwidth operations using data types such as INT4, INT8, FP8, and FP16. A common metric is trillions of operations per second (TOPS). Although TOPS does not explicitly specify the kind of operations, it is typically INT8 additions and multiplications. === Datacenters === Accelerators are used in cloud computing servers: e.g., tensor processing units (TPU) for Google Cloud Platform, and Trainium and Inferentia chips for Amazon Web Services. Many vendor-specific terms exist for devices in this category, and it is an emerging technology without a dominant design. Since the late 2010s, graphics processing units designed by companies such as Nvidia and AMD often include AI-specific hardware in the form of dedicated functional units for low-precision matrix-multiplication operations. These GPUs are commonly used as AI accelerators, both for training and inference. === Scientific computation === Although NPUs are tailored for low-precision (e.g., FP16, INT8) matrix multiplication operations, they can be used to emulate higher-precision matrix multiplications in scientific computing. As modern GPUs place much focus on making the NPU part fast, using emulated FP64 (Ozaki scheme) on NPUs can potentially outperform native FP64. This has been demonstrated using FP16-emulated FP64 on NVIDIA TITAN RTX and using INT8-emulated FP64 on NVIDIA consumer GPUs and the A100 GPU. Consumer GPUs especially benefited as they have limited FP64 hardware capacity, showing a 6× speedup. Since CUDA Toolkit 13.0 Update 2, cuBLAS automatically uses INT8-emulated FP64 matrix multiplication of the equivalent precision if it is faster than native. This is in addition to the FP16-emulated FP32 feature introduced in version 12.9. == Programming == An operating system or a higher-level library may provide application programming interfaces such as TensorFlow with LiteRT Next (Android), CoreML (iOS, macOS) or DirectML (Windows). Formats such as ONNX are used to represent trained neural networks. Consumer CPU-integrated NPUs are accessible through vendor-specific APIs. AMD (Ryzen AI), Intel (OpenVINO), Apple silicon (CoreML), and Qualcomm (SNPE) each have their own APIs, which can be built upon by a higher-level library. GPUs generally use existing GPGPU pipelines such as CUDA and OpenCL adapted for lower precisions and specialized matrix-multiplication operations. Vulkan is also being used. Custom-built systems such as the Google TPU use private interfaces. There are a large number of separate underlying acceleration APIs and compilers/runtimes in use in the AI field, causing a great increase in software development effort due to the many combinations involved. As of 2025, the open standard organization Khronos Group is pursuing standardization of AI-related interfaces to reduce the amount of work needed. Khronos is working on three separate fronts: expansion of data types and intrinsic operations in OpenCL and Vulkan, inclusion of compute graphs in SPIR-V, and a NNEF/SkriptND file format for describing a neural network.

  Read more →
• 

  Interlingual machine translation

  Interlingual machine translation is one of the classic approaches to machine translation. In this approach, the source language, i.e. the text to be translated is transformed into an interlingua, i.e., an abstract language-independent representation. The target language is then generated from the interlingua. Within the rule-based machine translation paradigm, the interlingual approach is an alternative to the direct approach and the transfer approach. In the direct approach, words are translated directly without passing through an additional representation. In the transfer approach the source language is transformed into an abstract, less language-specific representation. Linguistic rules which are specific to the language pair then transform the source language representation into an abstract target language representation and from this the target sentence is generated. The interlingual approach to machine translation has advantages and disadvantages. The advantages are that it requires fewer components in order to relate each source language to each target language, it takes fewer components to add a new language, it supports paraphrases of the input in the original language, it allows both the analysers and generators to be written by monolingual system developers, and it handles languages that are very different from each other (e.g. English and Arabic). The obvious disadvantage is that the definition of an interlingua is difficult and maybe even impossible for a wider domain. The ideal context for interlingual machine translation is thus multilingual machine translation in a very specific domain. For example, Interlingua has been used as a pivot language in international conferences and has been proposed as a pivot language for the European Union. == History == The first ideas about interlingual machine translation appeared in the 17th century with Descartes and Leibniz, who came up with theories of how to create dictionaries using universal numerical codes, not unlike numerical tokens used by large language models nowadays. Others, such as Cave Beck, Athanasius Kircher and Johann Joachim Becher worked on developing an unambiguous universal language based on the principles of logic and iconographs. In 1668, John Wilkins described his interlingua in his "Essay towards a Real Character and a Philosophical Language". In the 18th and 19th centuries many proposals for "universal" international languages were developed, the most well known being Esperanto. That said, applying the idea of a universal language to machine translation did not appear in any of the first significant approaches. Instead, work started on pairs of languages. However, during the 1950s and 60s, researchers in Cambridge headed by Margaret Masterman, in Leningrad headed by Nikolai Andreev and in Milan by Silvio Ceccato started work in this area. The idea was discussed extensively by the Israeli philosopher Yehoshua Bar-Hillel in 1969. During the 1970s, noteworthy research was done in Grenoble by researchers attempting to translate physics and mathematical texts from Russian to French, and in Texas a similar project (METAL) was ongoing for Russian to English. Early interlingual MT systems were also built at Stanford in the 1970s by Roger Schank and Yorick Wilks; the former became the basis of a commercial system for the transfer of funds, and the latter's code is preserved at The Computer Museum at Boston as the first interlingual machine translation system. In the 1980s, renewed relevance was given to interlingua-based, and knowledge-based approaches to machine translation in general, with much research going on in the field. The uniting factor in this research was that high-quality translation required abandoning the idea of requiring total comprehension of the text. Instead, the translation should be based on linguistic knowledge and the specific domain in which the system would be used. The most important research of this era was done in distributed language translation (DLT) in Utrecht, which worked with a modified version of Esperanto, and the Fujitsu system in Japan. In 2016, Google Neural Machine Translation achieved "zero-shot translation", that is it directly translates one language into another. For example, it might be trained just for Japanese-English and Korean-English translation, but can perform Japanese-Korean translation. The system appears to have learned to produce a language-independent intermediate representation of language (an "interlingua"), which allows it to perform zero-shot translation by converting from and to the interlingua. == Outline == In this method of translation, the interlingua can be thought of as a way of describing the analysis of a text written in a source language such that it is possible to convert its morphological, syntactic, semantic (and even pragmatic) characteristics, that is "meaning" into a target language. This interlingua is able to describe all of the characteristics of all of the languages which are to be translated, instead of simply translating from one language to another. Sometimes two interlinguas are used in translation. It is possible that one of the two covers more of the characteristics of the source language, and the other possess more of the characteristics of the target language. The translation then proceeds by converting sentences from the first language into sentences closer to the target language through two stages. The system may also be set up such that the second interlingua uses a more specific vocabulary that is closer, or more aligned with the target language, and this could improve the translation quality. The above-mentioned system is based on the idea of using linguistic proximity to improve the translation quality from a text in one original language to many other structurally similar languages from only one original analysis. This principle is also used in pivot machine translation, where a natural language is used as a "bridge" between two more distant languages. For example, in the case of translating to English from Ukrainian using Russian as an intermediate language. == Translation process == In interlingual machine translation systems, there are two monolingual components: the analysis of the source language and the interlingual, and the generation of the interlingua and the target language. It is however necessary to distinguish between interlingual systems using only syntactic methods (for example the systems developed in the 1970s at the universities of Grenoble and Texas) and those based on artificial intelligence (from 1987 in Japan and the research at the universities of Southern California and Carnegie Mellon). The first type of system corresponds to that outlined in Figure 1. while the other types would be approximated by the diagram in Figure 4. The following resources are necessary to an interlingual machine translation system: Dictionaries (or lexicons) for analysis and generation (specific to the domain and the languages involved). A conceptual lexicon (specific to the domain), which is the knowledge base about events and entities known in the domain. A set of projection rules (specific to the domain and the languages). Grammars for the analysis and generation of the languages involved. One of the problems of knowledge-based machine translation systems is that it becomes impossible to create databases for domains larger than very specific areas. Another is that processing these databases is very computationally expensive. == Efficacy == One of the main advantages of this strategy is that it provides an economical way to make multilingual translation systems. With an interlingua it becomes unnecessary to make a translation pair between each pair of languages in the system. So instead of creating n ( n − 1 ) {\displaystyle n(n-1)} language pairs, where n {\displaystyle n} is the number of languages in the system, it is only necessary to make 2 n {\displaystyle 2n} pairs between the n {\displaystyle n} languages and the interlingua. The main disadvantage of this strategy is the difficulty of creating an adequate interlingua. It should be both abstract and independent of the source and target languages. The more languages added to the translation system, and the more different they are, the more potent the interlingua must be to express all possible translation directions. Another problem is that it is difficult to extract meaning from texts in the original languages to create the intermediate representation. == Existing interlingual machine translation systems == Calliope-Aero Carabao Linguistic Virtual Machine Grammatical Framework Number Translator Google Translate use English internally as a pivot language for some language pairs such as Chinese and Japanese, and more generally those with "higher quality" neural-network translators with English but not between each other.

  Read more →
• 

  Restricted Boltzmann machine

  A restricted Boltzmann machine (RBM) (also called a restricted Sherrington–Kirkpatrick model with external field or restricted stochastic Ising–Lenz–Little model) is a generative stochastic artificial neural network that can learn a probability distribution over its set of inputs. RBMs were initially proposed under the name Harmonium by Paul Smolensky in 1986, and rose to prominence after Geoffrey Hinton and collaborators used fast learning algorithms for them in the mid-2000s. RBMs have found applications in dimensionality reduction, classification, collaborative filtering, feature learning, topic modelling, immunology, and even many‑body quantum mechanics. They can be trained in either supervised or unsupervised ways, depending on the task. As their name implies, RBMs are a variant of Boltzmann machines, with the restriction that their neurons must form a bipartite graph: a pair of nodes from each of the two groups of units (commonly referred to as the "visible" and "hidden" units respectively) may have a symmetric connection between them; and there are no connections between nodes within a group. By contrast, "unrestricted" Boltzmann machines may have connections between hidden units. This restriction allows for more efficient training algorithms than are available for the general class of Boltzmann machines, in particular the gradient-based contrastive divergence algorithm. Restricted Boltzmann machines can also be used in deep learning networks. In particular, deep belief networks can be formed by "stacking" RBMs and optionally fine-tuning the resulting deep network with gradient descent and backpropagation. == Structure == The standard type of RBM has binary-valued (Boolean) hidden and visible units, and consists of a matrix of weights W {\displaystyle W} of size m × n {\displaystyle m\times n} . Each weight element ( w i , j ) {\displaystyle (w_{i,j})} of the matrix is associated with the connection between the visible (input) unit v i {\displaystyle v_{i}} and the hidden unit h j {\displaystyle h_{j}} . In addition, there are bias weights (offsets) a i {\displaystyle a_{i}} for v i {\displaystyle v_{i}} and b j {\displaystyle b_{j}} for h j {\displaystyle h_{j}} . Given the weights and biases, the energy of a configuration (pair of Boolean vectors) (v,h) is defined as E ( v , h ) = − ∑ i a i v i − ∑ j b j h j − ∑ i ∑ j v i w i , j h j {\displaystyle E(v,h)=-\sum _{i}a_{i}v_{i}-\sum _{j}b_{j}h_{j}-\sum _{i}\sum _{j}v_{i}w_{i,j}h_{j}} or, in matrix notation, E ( v , h ) = − a T v − b T h − v T W h . {\displaystyle E(v,h)=-a^{\mathrm {T} }v-b^{\mathrm {T} }h-v^{\mathrm {T} }Wh.} This energy function is analogous to that of a Hopfield network. As with general Boltzmann machines, the joint probability distribution for the visible and hidden vectors is defined in terms of the energy function as follows, P ( v , h ) = 1 Z e − E ( v , h ) {\displaystyle P(v,h)={\frac {1}{Z}}e^{-E(v,h)}} where Z {\displaystyle Z} is a partition function defined as the sum of e − E ( v , h ) {\displaystyle e^{-E(v,h)}} over all possible configurations, which can be interpreted as a normalizing constant to ensure that the probabilities sum to 1. The marginal probability of a visible vector is the sum of P ( v , h ) {\displaystyle P(v,h)} over all possible hidden layer configurations, P ( v ) = 1 Z ∑ { h } e − E ( v , h ) {\displaystyle P(v)={\frac {1}{Z}}\sum _{\{h\}}e^{-E(v,h)}} , and vice versa. Since the underlying graph structure of the RBM is bipartite (meaning there are no intra-layer connections), the hidden unit activations are mutually independent given the visible unit activations. Conversely, the visible unit activations are mutually independent given the hidden unit activations. That is, for m visible units and n hidden units, the conditional probability of a configuration of the visible units v, given a configuration of the hidden units h, is P ( v | h ) = ∏ i = 1 m P ( v i | h ) {\displaystyle P(v|h)=\prod _{i=1}^{m}P(v_{i}|h)} . Conversely, the conditional probability of h given v is P ( h | v ) = ∏ j = 1 n P ( h j | v ) {\displaystyle P(h|v)=\prod _{j=1}^{n}P(h_{j}|v)} . The individual activation probabilities are given by P ( h j = 1 | v ) = σ ( b j + ∑ i = 1 m w i , j v i ) {\displaystyle P(h_{j}=1|v)=\sigma \left(b_{j}+\sum _{i=1}^{m}w_{i,j}v_{i}\right)} and P ( v i = 1 | h ) = σ ( a i + ∑ j = 1 n w i , j h j ) {\displaystyle \,P(v_{i}=1|h)=\sigma \left(a_{i}+\sum _{j=1}^{n}w_{i,j}h_{j}\right)} where σ {\displaystyle \sigma } denotes the logistic sigmoid. The visible units of Restricted Boltzmann Machine can be multinomial, although the hidden units are Bernoulli. In this case, the logistic function for visible units is replaced by the softmax function P ( v i k = 1 | h ) = exp ⁡ ( a i k + Σ j W i j k h j ) Σ k ′ = 1 K exp ⁡ ( a i k ′ + Σ j W i j k ′ h j ) {\displaystyle P(v_{i}^{k}=1|h)={\frac {\exp(a_{i}^{k}+\Sigma _{j}W_{ij}^{k}h_{j})}{\Sigma _{k'=1}^{K}\exp(a_{i}^{k'}+\Sigma _{j}W_{ij}^{k'}h_{j})}}} where K is the number of discrete values that the visible values have. They are applied in topic modeling, and recommender systems. === Relation to other models === Restricted Boltzmann machines are a special case of Boltzmann machines and Markov random fields. The graphical model of RBMs corresponds to that of factor analysis. == Training algorithm == Restricted Boltzmann machines are trained to maximize the product of probabilities assigned to some training set V {\displaystyle V} (a matrix, each row of which is treated as a visible vector v {\displaystyle v} ), arg ⁡ max W ∏ v ∈ V P ( v ) {\displaystyle \arg \max _{W}\prod _{v\in V}P(v)} or equivalently, to maximize the expected log probability of a training sample v {\displaystyle v} selected randomly from V {\displaystyle V} : arg ⁡ max W E [ log ⁡ P ( v ) ] {\displaystyle \arg \max _{W}\mathbb {E} \left[\log P(v)\right]} The algorithm most often used to train RBMs, that is, to optimize the weight matrix W {\displaystyle W} , is the contrastive divergence (CD) algorithm due to Hinton, originally developed to train PoE (product of experts) models. The algorithm performs Gibbs sampling and is used inside a gradient descent procedure (similar to the way backpropagation is used inside such a procedure when training feedforward neural nets) to compute weight update. The basic, single-step contrastive divergence (CD-1) procedure for a single sample can be summarized as follows: Take a training sample v, compute the probabilities of the hidden units and sample a hidden activation vector h from this probability distribution. Compute the outer product of v and h and call this the positive gradient. From h, sample a reconstruction v' of the visible units, then resample the hidden activations h' from this. (Gibbs sampling step) Compute the outer product of v' and h' and call this the negative gradient. Let the update to the weight matrix W {\displaystyle W} be the positive gradient minus the negative gradient, times some learning rate: Δ W = ϵ ( v h T − v ′ h ′ T ) {\displaystyle \Delta W=\epsilon (vh^{\mathsf {T}}-v'h'^{\mathsf {T}})} . Update the biases a and b analogously: Δ a = ϵ ( v − v ′ ) {\displaystyle \Delta a=\epsilon (v-v')} , Δ b = ϵ ( h − h ′ ) {\displaystyle \Delta b=\epsilon (h-h')} . A Practical Guide to Training RBMs written by Hinton can be found on his homepage. == Stacked Restricted Boltzmann Machine == The difference between the Stacked Restricted Boltzmann Machines and RBM is that RBM has lateral connections within a layer that are prohibited to make analysis tractable. On the other hand, the Stacked Boltzmann consists of a combination of an unsupervised three-layer network with symmetric weights and a supervised fine-tuned top layer for recognizing three classes. The usage of Stacked Boltzmann is to understand Natural languages, retrieve documents, image generation, and classification. These functions are trained with unsupervised pre-training and/or supervised fine-tuning. Unlike the undirected symmetric top layer, with a two-way unsymmetric layer for connection for RBM. The restricted Boltzmann's connection is three-layers with asymmetric weights, and two networks are combined into one. Stacked Boltzmann does share similarities with RBM, the neuron for Stacked Boltzmann is a stochastic binary Hopfield neuron, which is the same as the Restricted Boltzmann Machine. The energy from both Restricted Boltzmann and RBM is given by Gibb's probability measure: E = − 1 2 ∑ i , j w i j s i s j + ∑ i θ i s i {\displaystyle E=-{\frac {1}{2}}\sum _{i,j}{w_{ij}{s_{i}}{s_{j}}}+\sum _{i}{\theta _{i}}{s_{i}}} . The training process of Restricted Boltzmann is similar to RBM. Restricted Boltzmann train one layer at a time and approximate equilibrium state with a 3-segment pass, not performing back propagation. Restricted Boltzmann uses both supervised and unsupervised on different RBM for pre-training for classification and recognition. The training uses contrastive divergence with

  Read more →