Barney Pell

Barney Pell (born March 18, 1968) is an American entrepreneur, angel investor and computer scientist. He was co-founder and CEO of Powerset, a pioneering natural language search startup, search strategist and architect for Microsoft's Bing search engine, a pioneer in the field of general game playing in artificial intelligence, and the architect of the first intelligent agent to fly onboard and control a spacecraft. He was co-founder, Vice Chairman and Chief Strategy Officer of Moon Express; co-founder and chairman of LocoMobi; and Associate Founder of Singularity University. == Career == === Education === Pell received his Bachelor of Science degree in symbolic systems from Stanford University in 1989, where he graduated Phi Beta Kappa and was a National Merit Scholar. Pell earned a PhD in computer science from Cambridge University in 1993, supervised by Stephen Pulman, where he was a Marshall Scholar. === Research === Pell's research is focused on basic problems in the study of intelligence, computer game playing, machine learning, natural language processing, autonomous robotics, and web search. Barney Pell has published over 30 technical papers on topics related to information retrieval, knowledge management, machine learning, artificial intelligence, and scheduling systems. In computer game playing and machine learning, he was a pioneer in the field of General Game Playing, and created programs to generate the rules of chess-like games and programs to play individual games directly from the rules without human assistance. He also did early work on machine learning in the game of Go and on an architecture for pragmatic reasoning for bidding in the game of Bridge. In natural language processing, he was a scientist in the Artificial Intelligence Center at SRI International, where he worked on the Core Language Engine. Barney Pell was the Technical Area Manager of the Collaborative and Assistant Systems area within the Computational Sciences Division (now the Intelligent Systems Division) at NASA Ames Research Center, where he oversaw a staff of 80 scientists working on information retrieval, search, knowledge management, machine learning, semantic technology, human centered systems, collaboration technology, adaptive user interfaces, human robot interaction, and other areas of artificial intelligence. From 1993 to 1998, Barney Pell worked as a Principal Investigator and Senior Computer Scientist at NASA Ames, where he conducted advanced research and development of autonomous control software for NASA's deep space missions. He was the Architect for the Deep Space One Remote Agent Experiment and the Project Lead for the Executive component of the Remote Agent Experiment, the first intelligent agent to fly onboard and control a spacecraft. === Business === Pell is an entrepreneur who has founded or co-founded several business ventures, including Powerset, Moon Express, and LocoMobi. He was the founder and CEO of Powerset, a San Francisco startup company that built a search engine based on natural language processing technology originally developed at XEROX PARC. On May 11, 2008, the company unveiled a tool for searching a fixed subset of Wikipedia using conversational phrases rather than keywords. On July 1, 2008, Microsoft signed an agreement to acquire Powerset for an estimated $100 million. Powerset became a part of Microsoft's search engine, Bing. From 2008 until August 2011, Pell served as Partner, Search Strategist, and Evangelist for Microsoft's search engine, Bing and as Head of Bing's Local and Mobile Search teams. Prior to joining Powerset, Pell was an Entrepreneur-in-Residence at Mayfield Fund, a venture capital firm in Silicon Valley. Pell is also a founder of Moon Express, Inc., a U.S. company awarded a $10M commercial lunar contract by NASA and a competitor in the Google Lunar X PRIZE. Pell was also co-founder and chairman of LocoMobi, Inc., a U.S. company developing mobile, software and hardware technology solutions for the parking industry. LocoMobi was winner of the Tie50 Award in 2014. Pell is also an associate founder of Singularity University and a Machine Learning Fellow at the Creative Destruction Lab at the Rotman School of Management From 1998 to 2000, Pell served as chief strategist and vice president of business development at StockMaster.com (acquired by Red Herring in March, 2000). From 2000 to 2002, Pell was Chief Strategist and Vice President of Business Development for Whizbang Labs. Pell has been an angel investor and advisor to numerous startup companies, including Pulse.io (acquired by Google), Aardvark (acquired by Google), Appjet (acquired by Google), Jibe Mobile (acquired by Google), Movity (acquired by Trulia), QuestBridge, BrandYourself, CrowdFlower (acquired by Appen), and LinkedIn. === Views and predictions === Pell has expressed views and predictions regarding technological advancements in coming years. He believes that humans will soon have "brain-machine interfaces that will let people interact with each other as if they had 'hangouts' in their mind." Pell predicts these interfaces to become available within 20 to 30 years. Pell also predicts advancements in bodily augmentation, such as "even-better-than-human prosthetics and high-quality tissue engineering within 10 years." Pell believes that with advancements in space exploration technology the moon will soon be a commercially viable resource for material such as platinum and water. == Awards and recognition == In 1986, Pell was awarded a National Merit Scholarship. In 1989, Pell was awarded a Marshall Scholarship. In 1989, Pell was elected Phi Beta Kappa. In 1997, Pell was part of the team award a NASA Software of the Year Award for the Deep Space 1 Remote Agent.

Transparency in the software supply chain

Transparency in the software supply chain is a condition in which participants involved in the development, procurement, operation, auditing, or regulation of software can determine which components, dependencies, build stages, identifiers, and relationships within the supply chain make up the delivered product. The disclosure of information about software components, their interrelationships, origins, and development methods—for the purposes of risk management, vulnerability detection, and compliance—takes place throughout the software lifecycle. Transparency is one of the key security attributes of the software supply chain, as a deeper understanding of the chain enables participants to identify vulnerabilities and mitigate threats. Problems in the software supply chain can cause billions in losses and create operational challenges for government and commercial entities, as demonstrated by incidents involving SolarWinds, Bybit, 3CX, Jaguar Land Rover, GitHub, and NotPetya. Modern software is often assembled from third-party libraries and open-source components. According to research by the Linux Foundation and Synopsys, 96% of the commercial codebases analyzed contained open-source software, and 70–90% of a typical codebase may consist of open-source components. Without transparency, any software component can become a threat. As a result, companies may spend billions of dollars building robust external defenses, but this will not protect against vulnerabilities in legitimate software inside the perimeter. At the same time, supply chain attacks also erode trust between customers and their IT providers, as malicious code is often embedded in official updates with certificates and digital signatures. One of the primary ways to ensure transparency is through a software bill of materials, which documents the components used to create the software and the relationships within the supply chain. == Concept == The software supply chain is the collection of systems, devices, people, artifacts, and processes involved in the creation of the final software product. Attacks on the software supply chain differ from conventional attacks in that they follow a four-stage pattern: compromise, modification, distribution, and subsequent exploitation of the compromised or modified component. A defining feature of a supply chain attack is the introduction or manipulation of a change at an upstream stage, which is subsequently exploited at a downstream stage. Transparency refers to the availability of knowledge about the chain, while validity concerns the integrity of operations and artifacts and the authentication of participants, and separation involves reducing unnecessary trust relationships and the radius of impact through compartmentalization. In this framework, transparency primarily helps during the pre-compromise and detection phases, as a clearer understanding of participants, operations, and artifacts makes it easier to identify weak links before attackers exploit them. Current major attack vectors include dependencies and containers, build infrastructure, and human participants, such as maintainers or developers. == History == Software supply-chain transparency developed from earlier efforts to document software components, long before the term came into widespread use in the cybersecurity field. Early component-documentation formats included SPDX, first published in 2011, and CycloneDX, first published in 2017. Initially, these formats were created to support license compliance, package identification, and tool compatibility. Their development helped shape a broader concept of software supply chain transparency, encompassing component documentation, disclosure practices, risk management, security analysis, and regulatory compliance. In 2018, the U.S. National Telecommunications and Information Administration launched a multistakeholder process on promoting software component transparency. This process helped move work on SBOMs from a specialized technical practice into the realm of policy and procurement to identify components used in software products. The 2020 compromise of the SolarWinds Orion platform made software supply chain security a central issue in government cybersecurity policy. An analysis of the “Sunburst” campaign prepared by the Atlantic Council noted that the vulnerability of the software supply chain had become a realized risk for national-security agencies. In May 2021, U.S. President Joe Biden issued Executive Order 14028, which directed federal agencies to improve cybersecurity and increase transparency in the software supply chain, including requirements related to SBOMs. Reuters reported that the executive order required software developers selling their products to the federal government to provide greater visibility into their software and make security data available. In July 2021, the NTIA published the document “The Minimum Elements for a Software Bill of Materials (SBOM)”, defining the basic data fields and practices for creating SBOMs. Between 2021 and 2025, the U.S. Cybersecurity and Infrastructure Security Agency updated its guidance on “Framing Software Component Transparency”, expanding the set of SBOM attributes, metadata requirements, and operational recommendations for the creation, exchange, and use of SBOMs. Major incidents that occurred following the SolarWinds attack have underscored the importance of transparency in vulnerability management and supply chain security. The Log4Shell vulnerability in the Log4j library, disclosed in December 2021, demonstrated how difficult it can be for organizations to identify a vulnerable component deeply embedded within applications and services. In 2024, an attempt to plant a backdoor in XZ Utils showed how attackers could exploit trust in open-source maintenance processes to introduce malicious code into widely used infrastructure software. By the mid-2020s, software supply chain transparency had become part of international cybersecurity coordination and regulation. On September 3, 2025, Japan's Ministry of Economy, Trade and Industry and the National Cybersecurity Office, in collaboration with cybersecurity agencies from 15 countries, released the document “A Shared Vision of Software Bill of Materials (SBOM) for Cybersecurity.” In the European Union, the Cyber Resilience Act required manufacturers of products with digital elements to create, maintain, and retain SBOMs as part of the technical documentation for software placed on the EU market. == Transparency mechanisms == The primary mechanism for ensuring transparency is the software bill of materials (SBOM). An SBOM is a structured list of components, libraries, and tools used to build and distribute a software product, and it records dependencies in a way that helps organizations understand and assess their software supply chains. It can also be described as a formal record of components and their interdependencies, which gives users insight into their actual exposure to risks and threats. Five key areas of SBOM application in software supply chain security have been identified: vulnerability management, ensuring transparency, component evaluation, risk assessment, and ensuring supply chain integrity. In software supply chains, an SBOM documents all components, both open-source and proprietary. Under Executive Order 14028, U.S. federal agencies require software suppliers to provide SBOMs for government-procured software. The list of minimum required SBOM elements defined by NTIA includes three main categories: required data fields for describing each component (name, version, identifiers), automation support (machine-readable format, generation tools), and recommendations for creating SBOMs during development and purchasing. The post-2021 push for SBOMs was intended to provide visibility into the components used within software and to expose parts of an application that would otherwise remain hidden. This information can be used to prioritize patches, manage vulnerabilities, and support compliance work. Transparency also supports software traceability, which is becoming a standard feature of developer platforms. Traceability has become important because organizations are increasingly required to demonstrate how software was created, rather than simply listing its components. Higher levels of assurance require signed, tamper-proof traceability and more isolated, verifiable build environments. A related mechanism is build reproducibility. Reproducible builds are defined as build processes that make the compilation process deterministic, ensuring that the same source code always produces the same binary file. These builds are considered a foundational element for distributed verification, transparency-log maintenance, supply-chain workflow integration, and the creation of keyless signatures based on verifiable logs. Although reproducibility does not replace inventory or attestation, it gives external par

Feature hashing

In machine learning, feature hashing, also known as the hashing trick (by analogy to the kernel trick), is a fast and space-efficient way of vectorizing features, i.e. turning arbitrary features into indices in a vector or matrix. It works by applying a hash function to the features and using their hash values as indices directly (after a modulo operation), rather than looking the indices up in an associative array. In addition to its use for encoding non-numeric values, feature hashing can also be used for dimensionality reduction. This trick is often attributed to Weinberger et al. (2009), but there exists a much earlier description of this method published by John Moody in 1989. == Motivation == === Motivating example === In a typical document classification task, the input to the machine learning algorithm (both during learning and classification) is free text. From this, a bag of words (BOW) representation is constructed: the individual tokens are extracted and counted, and each distinct token in the training set defines a feature (independent variable) of each of the documents in both the training and test sets. Machine learning algorithms, however, are typically defined in terms of numerical vectors. Therefore, the bags of words for a set of documents is regarded as a term-document matrix where each row is a single document, and each column is a single feature/word; the entry i, j in such a matrix captures the frequency (or weight) of the j'th term of the vocabulary in document i. (An alternative convention swaps the rows and columns of the matrix, but this difference is immaterial.) Typically, these vectors are extremely sparse—according to Zipf's law. The common approach is to construct, at learning time or prior to that, a dictionary representation of the vocabulary of the training set, and use that to map words to indices. Hash tables and tries are common candidates for dictionary implementation. E.g., the three documents John likes to watch movies. Mary likes movies too. John also likes football. can be converted, using the dictionary to the term-document matrix ( John likes to watch movies Mary too also football 1 1 1 1 1 0 0 0 0 0 1 0 0 1 1 1 0 0 1 1 0 0 0 0 0 1 1 ) {\displaystyle {\begin{pmatrix}{\textrm {John}}&{\textrm {likes}}&{\textrm {to}}&{\textrm {watch}}&{\textrm {movies}}&{\textrm {Mary}}&{\textrm {too}}&{\textrm {also}}&{\textrm {football}}\\1&1&1&1&1&0&0&0&0\\0&1&0&0&1&1&1&0&0\\1&1&0&0&0&0&0&1&1\end{pmatrix}}} (Punctuation was removed, as is usual in document classification and clustering.) The problem with this process is that such dictionaries take up a large amount of storage space and grow in size as the training set grows. On the contrary, if the vocabulary is kept fixed and not increased with a growing training set, an adversary may try to invent new words or misspellings that are not in the stored vocabulary so as to circumvent a machine learned filter. To address this challenge, Yahoo! Research attempted to use feature hashing for their spam filters. Note that the hashing trick isn't limited to text classification and similar tasks at the document level, but can be applied to any problem that involves large (perhaps unbounded) numbers of features. === Mathematical motivation === Mathematically, a token is an element t {\displaystyle t} in a finite (or countably infinite) set T {\displaystyle T} . Suppose we only need to process a finite corpus, then we can put all tokens appearing in the corpus into T {\displaystyle T} , meaning that T {\displaystyle T} is finite. However, suppose we want to process all possible words made of the English letters, then T {\displaystyle T} is countably infinite. Most neural networks can only operate on real vector inputs, so we must construct a "dictionary" function ϕ : T → R n {\displaystyle \phi :T\to \mathbb {R} ^{n}} . When T {\displaystyle T} is finite, of size | T | = m ≤ n {\displaystyle |T|=m\leq n} , then we can use one-hot encoding to map it into R n {\displaystyle \mathbb {R} ^{n}} . First, arbitrarily enumerate T = { t 1 , t 2 , . . , t m } {\displaystyle T=\{t_{1},t_{2},..,t_{m}\}} , then define ϕ ( t i ) = e i {\displaystyle \phi (t_{i})=e_{i}} . In other words, we assign a unique index i {\displaystyle i} to each token, then map the token with index i {\displaystyle i} to the unit basis vector e i {\displaystyle e_{i}} . One-hot encoding is easy to interpret, but it requires one to maintain the arbitrary enumeration of T {\displaystyle T} . Given a token t ∈ T {\displaystyle t\in T} , to compute ϕ ( t ) {\displaystyle \phi (t)} , we must find out the index i {\displaystyle i} of the token t {\displaystyle t} . Thus, to implement ϕ {\displaystyle \phi } efficiently, we need a fast-to-compute bijection h : T → { 1 , . . . , m } {\displaystyle h:T\to \{1,...,m\}} , then we have ϕ ( t ) = e h ( t ) {\displaystyle \phi (t)=e_{h(t)}} . In fact, we can relax the requirement slightly: It suffices to have a fast-to-compute injection h : T → { 1 , . . . , n } {\displaystyle h:T\to \{1,...,n\}} , then use ϕ ( t ) = e h ( t ) {\displaystyle \phi (t)=e_{h(t)}} . In practice, there is no simple way to construct an efficient injection h : T → { 1 , . . . , n } {\displaystyle h:T\to \{1,...,n\}} . However, we do not need a strict injection, but only an approximate injection. That is, when t ≠ t ′ {\displaystyle t\neq t'} , we should probably have h ( t ) ≠ h ( t ′ ) {\displaystyle h(t)\neq h(t')} , so that probably ϕ ( t ) ≠ ϕ ( t ′ ) {\displaystyle \phi (t)\neq \phi (t')} . At this point, we have just specified that h {\displaystyle h} should be a hashing function. Thus we reach the idea of feature hashing. == Algorithms == === Feature hashing (Weinberger et al. 2009) === The basic feature hashing algorithm presented in (Weinberger et al. 2009) is defined as follows. First, one specifies two hash functions: the kernel hash h : T → { 1 , 2 , . . . , n } {\displaystyle h:T\to \{1,2,...,n\}} , and the sign hash ζ : T → { − 1 , + 1 } {\displaystyle \zeta :T\to \{-1,+1\}} . Next, one defines the feature hashing function: ϕ : T → R n , ϕ ( t ) = ζ ( t ) e h ( t ) {\displaystyle \phi :T\to \mathbb {R} ^{n},\quad \phi (t)=\zeta (t)e_{h(t)}} Finally, extend this feature hashing function to strings of tokens by ϕ : T ∗ → R n , ϕ ( t 1 , . . . , t k ) = ∑ j = 1 k ϕ ( t j ) {\displaystyle \phi :T^{}\to \mathbb {R} ^{n},\quad \phi (t_{1},...,t_{k})=\sum _{j=1}^{k}\phi (t_{j})} where T ∗ {\displaystyle T^{}} is the set of all finite strings consisting of tokens in T {\displaystyle T} . Equivalently, ϕ ( t 1 , . . . , t k ) = ∑ j = 1 k ζ ( t j ) e h ( t j ) = ∑ i = 1 n ( ∑ j : h ( t j ) = i ζ ( t j ) ) e i {\displaystyle \phi (t_{1},...,t_{k})=\sum _{j=1}^{k}\zeta (t_{j})e_{h(t_{j})}=\sum _{i=1}^{n}\left(\sum _{j:h(t_{j})=i}\zeta (t_{j})\right)e_{i}} ==== Geometric properties ==== We want to say something about the geometric property of ϕ {\displaystyle \phi } , but T {\displaystyle T} , by itself, is just a set of tokens, we cannot impose a geometric structure on it except the discrete topology, which is generated by the discrete metric. To make it nicer, we lift it to T → R T {\displaystyle T\to \mathbb {R} ^{T}} , and lift ϕ {\displaystyle \phi } from ϕ : T → R n {\displaystyle \phi :T\to \mathbb {R} ^{n}} to ϕ : R T → R n {\displaystyle \phi :\mathbb {R} ^{T}\to \mathbb {R} ^{n}} by linear extension: ϕ ( ( x t ) t ∈ T ) = ∑ t ∈ T x t ζ ( t ) e h ( t ) = ∑ i = 1 n ( ∑ t : h ( t ) = i x t ζ ( t ) ) e i {\displaystyle \phi ((x_{t})_{t\in T})=\sum _{t\in T}x_{t}\zeta (t)e_{h(t)}=\sum _{i=1}^{n}\left(\sum _{t:h(t)=i}x_{t}\zeta (t)\right)e_{i}} There is an infinite sum there, which must be handled at once. There are essentially only two ways to handle infinities. One may impose a metric, then take its completion, to allow well-behaved infinite sums, or one may demand that nothing is actually infinite, only potentially so. Here, we go for the potential-infinity way, by restricting R T {\displaystyle \mathbb {R} ^{T}} to contain only vectors with finite support: ∀ ( x t ) t ∈ T ∈ R T {\displaystyle \forall (x_{t})_{t\in T}\in \mathbb {R} ^{T}} , only finitely many entries of ( x t ) t ∈ T {\displaystyle (x_{t})_{t\in T}} are nonzero. Define an inner product on R T {\displaystyle \mathbb {R} ^{T}} in the obvious way: ⟨ e t , e t ′ ⟩ = { 1 , if t = t ′ , 0 , else. ⟨ x , x ′ ⟩ = ∑ t , t ′ ∈ T x t x t ′ ⟨ e t , e t ′ ⟩ {\displaystyle \langle e_{t},e_{t'}\rangle ={\begin{cases}1,{\text{ if }}t=t',\\0,{\text{ else.}}\end{cases}}\quad \langle x,x'\rangle =\sum _{t,t'\in T}x_{t}x_{t'}\langle e_{t},e_{t'}\rangle } As a side note, if T {\displaystyle T} is infinite, then the inner product space R T {\displaystyle \mathbb {R} ^{T}} is not complete. Taking its completion would get us to a Hilbert space, which allows well-behaved infinite sums. Now we have an inner product space, with enough structure to describe the geometry of the feature hashing function ϕ : R T → R n {\displaystyle \phi :\ma

Learning curve (machine learning)

In machine learning (ML), a learning curve (or training curve) is a graphical representation that shows how a model's performance on a training set (and usually a validation set) changes with the number of training iterations (epochs) or the amount of training data. Typically, the number of training epochs or training set size is plotted on the x-axis, and the value of the loss function (and possibly some other metric such as the cross-validation score) on the y-axis. Synonyms include error curve, experience curve, improvement curve and generalization curve. More abstractly, learning curves plot the difference between learning effort and predictive performance, where "learning effort" usually means the number of training samples, and "predictive performance" means accuracy on testing samples. Learning curves have many useful purposes in ML, including: choosing model parameters during design, adjusting optimization to improve convergence, and diagnosing problems such as overfitting (or underfitting). Learning curves can also be tools for determining how much a model benefits from adding more training data, and whether the model suffers more from a variance error or a bias error. If both the validation score and the training score converge to a certain value, then the model will no longer significantly benefit from more training data. == Formal definition == When creating a function to approximate the distribution of some data, it is necessary to define a loss function L ( f θ ( X ) , Y ) {\displaystyle L(f_{\theta }(X),Y)} to measure how good the model output is (e.g., accuracy for classification tasks or mean squared error for regression). We then define an optimization process which finds model parameters θ {\displaystyle \theta } such that L ( f θ ( X ) , Y ) {\displaystyle L(f_{\theta }(X),Y)} is minimized, referred to as θ ∗ {\displaystyle \theta ^{}} . === Training curve for amount of data === If the training data is { x 1 , x 2 , … , x n } , { y 1 , y 2 , … y n } {\displaystyle \{x_{1},x_{2},\dots ,x_{n}\},\{y_{1},y_{2},\dots y_{n}\}} and the validation data is { x 1 ′ , x 2 ′ , … x m ′ } , { y 1 ′ , y 2 ′ , … y m ′ } {\displaystyle \{x_{1}',x_{2}',\dots x_{m}'\},\{y_{1}',y_{2}',\dots y_{m}'\}} , a learning curve is the plot of the two curves i ↦ L ( f θ ∗ ( X i , Y i ) ( X i ) , Y i ) {\displaystyle i\mapsto L(f_{\theta ^{}(X_{i},Y_{i})}(X_{i}),Y_{i})} i ↦ L ( f θ ∗ ( X i , Y i ) ( X i ′ ) , Y i ′ ) {\displaystyle i\mapsto L(f_{\theta ^{}(X_{i},Y_{i})}(X_{i}'),Y_{i}')} where X i = { x 1 , x 2 , … x i } {\displaystyle X_{i}=\{x_{1},x_{2},\dots x_{i}\}} === Training curve for number of iterations === Many optimization algorithms are iterative, repeating the same step (such as backpropagation) until the process converges to an optimal value. Gradient descent is one such algorithm. If θ i ∗ {\displaystyle \theta _{i}^{}} is the approximation of the optimal θ {\displaystyle \theta } after i {\displaystyle i} steps, a learning curve is the plot of i ↦ L ( f θ i ∗ ( X , Y ) ( X ) , Y ) {\displaystyle i\mapsto L(f_{\theta _{i}^{}(X,Y)}(X),Y)} i ↦ L ( f θ i ∗ ( X , Y ) ( X ′ ) , Y ′ ) {\displaystyle i\mapsto L(f_{\theta _{i}^{}(X,Y)}(X'),Y')}

Exploration–exploitation dilemma

The exploration–exploitation dilemma, also known as the explore–exploit tradeoff, is a fundamental concept in decision-making that arises in many domains. It is depicted as the balancing act between two opposing strategies. Exploitation involves choosing the best option based on current knowledge of the system (which may be incomplete or misleading), while exploration involves trying out new options that may lead to better outcomes in the future at the expense of an exploitation opportunity. Finding the optimal balance between these two strategies is a crucial challenge in many decision-making problems whose goal is to maximize long-term benefits. == Application in machine learning == In the context of machine learning, the exploration–exploitation tradeoff is fundamental in reinforcement learning (RL), a type of machine learning that involves training agents to make decisions based on feedback from the environment. Crucially, this feedback may be incomplete or delayed. The agent must decide whether to exploit the current best-known policy or explore new policies to improve its performance. === Multi-armed bandit methods === The multi-armed bandit (MAB) problem was a classic example of the tradeoff, and many methods were developed for it, such as epsilon-greedy, Thompson sampling, and the upper confidence bound (UCB). See the page on MAB for details. In more complex RL situations than the MAB problem, the agent can treat each choice as a MAB, where the payoff is the expected future reward. For example, if the agent performs an epsilon-greedy method, then the agent will often "pull the best lever" by picking the action that had the best predicted expected reward (exploit). However, it would pick a random action with probability epsilon (explore). Monte Carlo tree search, for example, uses a variant of the UCB method. === Exploration problems === There are some problems that make exploration difficult. Sparse reward. If rewards occur only once a long while, then the agent might not persist in exploring. Furthermore, if the space of actions is large, then the sparse reward would mean the agent would not be guided by the reward to find a good direction for deeper exploration. A standard example is Montezuma's Revenge. Deceptive reward. If some early actions give immediate small reward, but other actions give later large reward, then the agent might be lured away from exploring the other actions. Noisy TV problem. If certain observations are irreducibly noisy (such as a television showing random images), then the agent might be trapped exploring those observations (watching the television). === Exploration reward === This section based on. The exploration reward (also called exploration bonus) methods convert the exploration-exploitation dilemma into a balance of exploitations. That is, instead of trying to get the agent to balance exploration and exploitation, exploration is simply treated as another form of exploitation, and the agent simply attempts to maximize the sum of rewards from exploration and exploitation. The exploration reward can be treated as a form of intrinsic reward. We write these as r t i , r t e {\displaystyle r_{t}^{i},r_{t}^{e}} , meaning the intrinsic and extrinsic rewards at time step t {\displaystyle t} . However, exploration reward is different from exploitation in two regards: The reward of exploitation is not freely chosen, but given by the environment, but the reward of exploration may be picked freely. Indeed, there are many different ways to design r t i {\displaystyle r_{t}^{i}} described below. The reward of exploitation is usually stationary (i.e. the same action in the same state gives the same reward), but the reward of exploration is non-stationary (i.e. the same action in the same state should give less and less reward). Count-based exploration uses N n ( s ) {\displaystyle N_{n}(s)} , the number of visits to a state s {\displaystyle s} during the time-steps 1 : n {\displaystyle 1:n} , to calculate the exploration reward. This is only possible in small and discrete state space. Density-based exploration extends count-based exploration by using a density model ρ n ( s ) {\displaystyle \rho _{n}(s)} . The idea is that, if a state has been visited, then nearby states are also partly-visited. In maximum entropy exploration, the entropy of the agent's policy π {\displaystyle \pi } is included as a term in the intrinsic reward. That is, r t i = − ∑ a π ( a | s t ) ln ⁡ π ( a | s t ) + ⋯ {\displaystyle r_{t}^{i}=-\sum _{a}\pi (a|s_{t})\ln \pi (a|s_{t})+\cdots } . === Prediction-based === This section based on. The forward dynamics model is a function for predicting the next state based on the current state and the current action: f : ( s t , a t ) ↦ s t + 1 {\displaystyle f:(s_{t},a_{t})\mapsto s_{t+1}} . The forward dynamics model is trained as the agent plays. The model becomes better at predicting state transition for state-action pairs that had been done many times. A forward dynamics model can define an exploration reward by r t i = ‖ f ( s t , a t ) − s t + 1 ‖ 2 2 {\displaystyle r_{t}^{i}=\|f(s_{t},a_{t})-s_{t+1}\|_{2}^{2}} . That is, the reward is the squared-error of the prediction compared to reality. This rewards the agent to perform state-action pairs that had not been done many times. This is however susceptible to the noisy TV problem. Dynamics model can be run in latent space. That is, r t i = ‖ f ( s t , a t ) − ϕ ( s t + 1 ) ‖ 2 2 {\displaystyle r_{t}^{i}=\|f(s_{t},a_{t})-\phi (s_{t+1})\|_{2}^{2}} for some featurizer ϕ {\displaystyle \phi } . The featurizer can be the identity function (i.e. ϕ ( x ) = x {\displaystyle \phi (x)=x} ), randomly generated, the encoder-half of a variational autoencoder, etc. A good featurizer improves forward dynamics exploration. The Intrinsic Curiosity Module (ICM) method trains simultaneously a forward dynamics model and a featurizer. The featurizer is trained by an inverse dynamics model, which is a function for predicting the current action based on the features of the current and the next state: g : ( ϕ ( s t ) , ϕ ( s t + 1 ) ) ↦ a t {\displaystyle g:(\phi (s_{t}),\phi (s_{t+1}))\mapsto a_{t}} . By optimizing the inverse dynamics, both the inverse dynamics model and the featurizer are improved. Then, the improved featurizer improves the forward dynamics model, which improves the exploration of the agent. Random Network Distillation (RND) method attempts to solve this problem by teacher–student distillation. Instead of a forward dynamics model, it has two models f , f ′ {\displaystyle f,f'} . The f ′ {\displaystyle f'} teacher model is fixed, and the f {\displaystyle f} student model is trained to minimize ‖ f ( s ) − f ′ ( s ) ‖ 2 2 {\displaystyle \|f(s)-f'(s)\|_{2}^{2}} on states s {\displaystyle s} . As a state is visited more and more, the student network becomes better at predicting the teacher. Meanwhile, the prediction error is also an exploration reward for the agent, and so the agent learns to perform actions that result in higher prediction error. Thus, we have a student network attempting to minimize the prediction error, while the agent attempting to maximize it, resulting in exploration. The states are normalized by subtracting a running average and dividing a running variance, which is necessary since the teacher model is frozen. The rewards are normalized by dividing with a running variance. Exploration by disagreement trains an ensemble of forward dynamics models, each on a random subset of all ( s t , a t , s t + 1 ) {\displaystyle (s_{t},a_{t},s_{t+1})} tuples. The exploration reward is the variance of the models' predictions. === Noise === For neural network–based agents, the NoisyNet method changes some of its neural network modules by noisy versions. That is, some network parameters are random variables from a probability distribution. The parameters of the distribution are themselves learnable. For example, in a linear layer y = W x + b {\displaystyle y=Wx+b} , both W , b {\displaystyle W,b} are sampled from Gaussian distributions N ( μ W , Σ W ) , N ( μ b , Σ b ) {\displaystyle {\mathcal {N}}(\mu _{W},\Sigma _{W}),{\mathcal {N}}(\mu _{b},\Sigma _{b})} at every step, and the parameters μ W , Σ W , μ b , Σ b {\displaystyle \mu _{W},\Sigma _{W},\mu _{b},\Sigma _{b}} are learned via the reparameterization trick.

Outline of machine learning

The following outline is provided as an overview of, and topical guide to, machine learning: Machine learning (ML) is a subfield of artificial intelligence within computer science that evolved from the study of pattern recognition and computational learning theory. In 1959, Arthur Samuel defined machine learning as a "field of study that gives computers the ability to learn without being explicitly programmed". ML involves the study and construction of algorithms that can learn from and make predictions on data. These algorithms operate by building a model from a training set of example observations to make data-driven predictions or decisions expressed as outputs, rather than following strictly static program instructions. == How can machine learning be categorized? == An academic discipline A branch of science An applied science A subfield of computer science A branch of artificial intelligence A subfield of soft computing Application of statistics === Paradigms of machine learning === Supervised learning, where the model is trained on labeled data Unsupervised learning, where the model tries to identify patterns in unlabeled data Reinforcement learning, where the model learns to make decisions by receiving rewards or penalties. == Applications of machine learning == Applications of machine learning Bioinformatics Biomedical informatics Computer vision Customer relationship management Data mining Earth sciences Email filtering Inverted pendulum (balance and equilibrium system) Natural language processing Named Entity Recognition Automatic summarization Automatic taxonomy construction Dialog system Grammar checker Language recognition Handwriting recognition Optical character recognition Speech recognition Text to Speech Synthesis Speech Emotion Recognition Machine translation Question answering Speech synthesis Text mining Term frequency–inverse document frequency Text simplification Pattern recognition Facial recognition system Handwriting recognition Image recognition Optical character recognition Speech recognition Recommendation system Collaborative filtering Content-based filtering Hybrid recommender systems Search engine Search engine optimization Social engineering == Machine learning hardware == Graphics processing unit Tensor processing unit Vision processing unit == Machine learning tools == Comparison of machine learning software Comparison of deep learning software === Machine learning frameworks === ==== Proprietary machine learning frameworks ==== Amazon Machine Learning Microsoft Azure Machine Learning Studio DistBelief (replaced by TensorFlow) ==== Open source machine learning frameworks ==== Apache Singa Apache MXNet Caffe PyTorch mlpack TensorFlow Torch CNTK Accord.Net Jax MLJ.jl – A machine learning framework for Julia === Machine learning libraries === Deeplearning4j Theano scikit-learn Keras === Machine learning algorithms === == Machine learning methods == === Instance-based algorithm === K-nearest neighbors algorithm (KNN) Learning vector quantization (LVQ) Self-organizing map (SOM) === Regression analysis === Logistic regression Ordinary least squares regression (OLSR) Linear regression Stepwise regression Multivariate adaptive regression splines (MARS) Regularization algorithm Ridge regression Least Absolute Shrinkage and Selection Operator (LASSO) Elastic net Least-angle regression (LARS) Classifiers Probabilistic classifier Naive Bayes classifier Binary classifier Linear classifier Hierarchical classifier === Dimensionality reduction === Dimensionality reduction Canonical correlation analysis (CCA) Factor analysis Feature extraction Feature selection Independent component analysis (ICA) Linear discriminant analysis (LDA) Multidimensional scaling (MDS) Non-negative matrix factorization (NMF) Partial least squares regression (PLSR) Principal component analysis (PCA) Principal component regression (PCR) Projection pursuit Sammon mapping t-distributed stochastic neighbor embedding (t-SNE) === Ensemble learning === Ensemble learning AdaBoost Boosting Bootstrap aggregating (also "bagging" or "bootstrapping") Ensemble averaging Gradient boosted decision tree (GBDT) Gradient boosting Random Forest Stacked Generalization === Meta-learning === Meta-learning Inductive bias Metadata === Reinforcement learning === Reinforcement learning Q-learning State–action–reward–state–action (SARSA) Temporal difference learning (TD) Learning Automata === Supervised learning === Supervised learning Averaged one-dependence estimators (AODE) Artificial neural network Case-based reasoning Gaussian process regression Gene expression programming Group method of data handling (GMDH) Inductive logic programming Instance-based learning Lazy learning Learning Automata Learning Vector Quantization Logistic Model Tree Minimum message length (decision trees, decision graphs, etc.) Nearest Neighbor Algorithm Analogical modeling Probably approximately correct learning (PAC) learning Ripple down rules, a knowledge acquisition methodology Symbolic machine learning algorithms Support vector machines Random Forests Ensembles of classifiers Bootstrap aggregating (bagging) Boosting (meta-algorithm) Ordinal classification Conditional Random Field ANOVA Quadratic classifiers k-nearest neighbor Boosting SPRINT Bayesian networks Naive Bayes Hidden Markov models Hierarchical hidden Markov model ==== Bayesian ==== Bayesian statistics Bayesian knowledge base Naive Bayes Gaussian Naive Bayes Multinomial Naive Bayes Averaged One-Dependence Estimators (AODE) Bayesian Belief Network (BBN) Bayesian Network (BN) ==== Decision tree algorithms ==== Decision tree algorithm Decision tree Classification and regression tree (CART) Iterative Dichotomiser 3 (ID3) C4.5 algorithm C5.0 algorithm Chi-squared Automatic Interaction Detection (CHAID) Decision stump Conditional decision tree ID3 algorithm Random forest SLIQ ==== Linear classifier ==== Linear classifier Fisher's linear discriminant Linear regression Logistic regression Multinomial logistic regression Naive Bayes classifier Perceptron Support vector machine === Unsupervised learning === Unsupervised learning Expectation-maximization algorithm Vector Quantization Generative topographic map Information bottleneck method Association rule learning algorithms Apriori algorithm Eclat algorithm ==== Artificial neural networks ==== Artificial neural network Feedforward neural network Extreme learning machine Convolutional neural network Recurrent neural network Long short-term memory (LSTM) Logic learning machine Self-organizing map ==== Association rule learning ==== Association rule learning Apriori algorithm Eclat algorithm FP-growth algorithm ==== Hierarchical clustering ==== Hierarchical clustering Single-linkage clustering Conceptual clustering ==== Cluster analysis ==== Cluster analysis BIRCH DBSCAN Expectation–maximization (EM) Fuzzy clustering Hierarchical clustering k-means clustering k-medians Mean-shift OPTICS algorithm ==== Anomaly detection ==== Anomaly detection k-nearest neighbors algorithm (k-NN) Local outlier factor === Semi-supervised learning === Semi-supervised learning Active learning Generative models Low-density separation Graph-based methods Co-training Transduction === Deep learning === Deep learning Deep belief networks Deep Boltzmann machines Deep Convolutional neural networks Deep Recurrent neural networks Hierarchical temporal memory Generative Adversarial Network Style transfer Transformer Stacked Auto-Encoders === Other machine learning methods and problems === Anomaly detection Association rules Bias-variance dilemma Classification Multi-label classification Clustering Data Pre-processing Empirical risk minimization Feature engineering Feature learning Learning to rank Occam learning Online machine learning PAC learning Regression Reinforcement Learning Semi-supervised learning Statistical learning Structured prediction Graphical models Bayesian network Conditional random field (CRF) Hidden Markov model (HMM) Unsupervised learning VC theory == Machine learning research == List of artificial intelligence projects List of datasets for machine learning research == History of machine learning == History of machine learning Timeline of machine learning == Machine learning projects == Machine learning projects: DeepMind Google Brain OpenAI Meta AI Hugging Face == Machine learning organizations == === Machine learning conferences and workshops === Artificial Intelligence and Security (AISec) (co-located workshop with CCS) Conference on Neural Information Processing Systems (NIPS) ECML PKDD International Conference on Machine Learning (ICML) ML4ALL (Machine Learning For All) == Machine learning publications == === Books on machine learning === Mathematics for Machine Learning Hands-On Machine Learning Scikit-Learn, Keras, and TensorFlow The Hundred-Page Machine Learning Book === Machine learning journals === Machine Learning Journal of Machine Learning Research (JMLR) Neural Computation == Pe

Spreading activation

Spreading activation is a method for searching associative networks, biological and artificial neural networks, or semantic networks. The search process is initiated by labeling a set of source nodes (e.g. concepts in a semantic network) with weights or "activation" and then iteratively propagating or "spreading" that activation out to other nodes linked to the source nodes. Most often these "weights" are real values that decay as activation propagates through the network. When the weights are discrete this process is often referred to as marker passing. Activation may originate from alternate paths, identified by distinct markers, and terminate when two alternate paths reach the same node. However brain studies show that several different brain areas play an important role in semantic processing. Spreading activation in semantic networks as a model were invented in cognitive psychology to model the fan out effect. Spreading activation can also be applied in information retrieval, by means of a network of nodes representing documents and terms contained in those documents. == Cognitive psychology == As it relates to cognitive psychology, spreading activation is the theory of how the brain iterates through a network of associated ideas to retrieve specific information. The spreading activation theory presents the array of concepts within our memory as cognitive units, each consisting of a node and its associated elements or characteristics, all connected together by edges. A spreading activation network can be represented schematically, in a sort of web diagram with shorter lines between two nodes meaning the ideas are more closely related and will typically be associated more quickly to the original concept. In memory psychology, the spreading activation model holds that people organize their knowledge of the world based on their personal experiences, which in turn form the network of ideas that is the person's knowledge of the world. When a word (the target) is preceded by an associated word (the prime) in word recognition tasks, participants seem to perform better in the amount of time that it takes them to respond. For instance, subjects respond faster to the word "doctor" when it is preceded by "nurse" than when it is preceded by an unrelated word like "carrot". This semantic priming effect with words that are close in meaning within the cognitive network has been seen in a wide range of tasks given by experimenters, ranging from sentence verification to lexical decision and naming. As another example, if the original concept is "red" and the concept "vehicles" is primed, they are much more likely to say "fire engine" instead of something unrelated to vehicles, such as "cherries". If instead "fruits" was primed, they would likely name "cherries" and continue on from there. The activation of pathways in the network has everything to do with how closely linked two concepts are by meaning, as well as how a subject is primed. == Algorithm == A directed graph is populated by Nodes[ 1...N ] each having an associated activation value A [ i ] which is a real number in the range [0.0 ... 1.0]. A Link[ i, j ] connects source node[ i ] with target node[ j ]. Each edge has an associated weight W [ i, j ] usually a real number in the range [0.0 ... 1.0]. Parameters: Firing threshold F, a real number in the range [0.0 ... 1.0] Decay factor D, a real number in the range [0.0 ... 1.0] Steps: Initialize the graph setting all activation values A [ i ] to zero. Set one or more origin nodes to an initial activation value greater than the firing threshold F. A typical initial value is 1.0. For each unfired node [ i ] in the graph having an activation value A [ i ] greater than the node firing threshold F: For each Link [ i, j ] connecting the source node [ i ] with target node [ j ], adjust A [ j ] = A [ j ] + (A [ i ] W [ i, j ] D) where D is the decay factor. If a target node receives an adjustment to its activation value so that it would exceed 1.0, then set its new activation value to 1.0. Likewise maintain 0.0 as a lower bound on the target node's activation value should it receive an adjustment to below 0.0. Once a node has fired it may not fire again, although variations of the basic algorithm permit repeated firings and loops through the graph. Nodes receiving a new activation value that exceeds the firing threshold F are marked for firing on the next spreading activation cycle. If activation originates from more than one node, a variation of the algorithm permits marker passing to distinguish the paths by which activation is spread over the graph The procedure terminates when either there are no more nodes to fire or in the case of marker passing from multiple origins, when a node is reached from more than one path. Variations of the algorithm that permit repeated node firings and activation loops in the graph, terminate after a steady activation state, with respect to some delta, is reached, or when a maximum number of iterations is exceeded. == Examples ==