The Semi-Automated Ground Environment (SAGE) was a system of large computers and associated networking equipment that coordinated data from many radar sites and processed it to produce a single unified image of the airspace over a wide area. SAGE directed and controlled the NORAD response to a possible Soviet air attack, operating in this role from the late 1950s into the 1980s. The processing power behind SAGE was supplied by the largest discrete component-based computer ever built, the AN/FSQ-7, manufactured by IBM. Each SAGE Direction Center (DC) housed an FSQ-7 which occupied an entire floor, approximately 22,000 square feet (2,000 m2) not including supporting equipment. The FSQ-7 was actually two computers, "A" side and "B" side. Computer processing was switched from "A" side to "B" side on a regular basis, allowing maintenance on the unused side. Information was fed to the DCs from a network of radar stations as well as readiness information from various defense sites. The computers, based on the raw radar data, developed "tracks" for the reported targets, and automatically calculated which defenses were within range. Operators used light guns to select targets on-screen for further information, select one of the available defenses, and issue commands to attack. These commands would then be automatically sent to the defense site via teleprinter. Connecting the various sites was an enormous network of telephones, modems and teleprinters. Later additions to the system allowed SAGE's tracking data to be sent directly to CIM-10 Bomarc missiles and some of the US Air Force's interceptor aircraft in-flight, directly updating their autopilots to maintain an intercept course without operator intervention. Each DC also forwarded data to a Combat Center (CC) for "supervision of the several sectors within the division" ("each combat center [had] the capability to coordinate defense for the whole nation"). SAGE became operational in the late 1950s and early 1960s at an estimated total cost between 8 and 12 billion dollars, four times the cost of the Manhattan Project. Throughout its development, there were continual concerns about its real ability to deal with large attacks, and the Operation Sky Shield tests showed that only about one-fourth of enemy bombers would have been intercepted. Nevertheless, SAGE was the backbone of NORAD's air defense system into the 1980s, by which time the tube-based FSQ-7s were increasingly costly to maintain and completely outdated. Today the same command and control task is carried out by microcomputers, based on the same basic underlying data. == Background == === Earlier systems === Just prior to World War II, Royal Air Force (RAF) tests with the new Chain Home (CH) radars had demonstrated that relaying information to the fighter aircraft directly from the radar sites was not feasible. The radars determined the map coordinates of the enemy, but could generally not see the fighters at the same time. This meant the fighters had to be able to determine where to fly to perform an interception but were often unaware of their own exact location and unable to calculate an interception while also flying their aircraft. The solution was to send all of the radar information to a central control station where operators collated the reports into single tracks, and then reported these tracks to the airbases, or sectors. The sectors used additional systems to track their own aircraft, plotting both on a single large map. Operators viewing the map could then see what direction their fighters would have to fly to approach their targets and relay that simply by telling them to fly along a certain heading or vector. This Dowding system was the first ground-controlled interception (GCI) system of large scale, covering the entirety of the UK. It proved enormously successful during the Battle of Britain, and is credited as being a key part of the RAF's success. The system was slow, often providing information that was up to five minutes out of date. Against propeller driven bombers flying at perhaps 225 miles per hour (362 km/h) this was not a serious concern, but it was clear the system would be of little use against jet-powered bombers flying at perhaps 600 miles per hour (970 km/h). The system was extremely expensive in manpower terms, requiring hundreds of telephone operators, plotters and trackers in addition to the radar operators. This was a serious drain on manpower, making it difficult to expand the network. The idea of using a computer to handle the task of taking reports and developing tracks had been explored beginning late in the war. By 1944, analog computers had been installed at the CH stations to automatically convert radar readings into map locations, eliminating two people. Meanwhile, the Royal Navy began experimenting with the Comprehensive Display System (CDS), another analog computer that took X and Y locations from a map and automatically generated tracks from repeated inputs. Similar systems began development with the Royal Canadian Navy, DATAR, and the US Navy, the Naval Tactical Data System (NTDS). A similar system was also specified for the Nike SAM project, specifically referring to a US version of CDS, coordinating the defense over a battle area so that multiple batteries did not fire on a single target. All of these systems were relatively small in geographic scale, generally tracking within a city-sized area. === Valley Committee === When the Soviet Union tested its first atomic bomb in August 1949, the topic of air defense of the US became important for the first time. A study group, the "Air Defense Systems Engineering Committee", was set up under the direction of Dr. George Valley to consider the problem and is known to history as the "Valley Committee". Their December report noted a key problem in air defense using ground-based radars. A bomber approaching a radar station would detect the signals from the radar long before the reflection off the bomber was strong enough to be detected by the station. The committee suggested that when this occurred, the bomber would descend to low altitude, thereby greatly limiting the radar horizon, allowing the bomber to fly past the station undetected. Although flying at low altitude greatly increased fuel consumption, the team calculated that the bomber would only need to do this for about 10% of its flight, making the fuel penalty acceptable. The only solution to this problem was to build a huge number of stations with overlapping coverage. At that point the problem became one of managing the information. Manual plotting was ruled out as too slow, and a computerized solution was the only possibility. To handle this task, the computer would need to be fed information directly, eliminating any manual translation by phone operators, and it would have to be able to analyze that information and automatically develop tracks. A system tasked with defending cities against the predicted future Soviet bomber fleet would have to be dramatically more powerful than the models used in the NTDS or DATAR. The Committee then had to consider whether or not such a computer was possible. The Valley Committee was introduced to Jerome Wiesner, associate director of the Research Laboratory of Electronics at MIT. Wiesner noted that the Servomechanisms Laboratory had already begun development of a machine that might be fast enough. This was the Whirlwind I, originally developed for the Office of Naval Research as a general purpose flight simulator that could simulate any current or future aircraft by changing its software. Wiesner introduced the Valley Committee to Whirlwind's project lead, Jay Forrester, who convinced him that Whirlwind was sufficiently capable. In September 1950, an early microwave early-warning radar system at Hanscom Field was connected to Whirlwind using a custom interface developed by Forrester's team. An aircraft was flown past the site, and the system digitized the radar information and successfully sent it to Whirlwind. With this demonstration, the technical concept was proven. Forrester was invited to join the committee. === Project Charles === With this successful demonstration, Louis Ridenour, chief scientist of the Air Force, wrote a memo stating "It is now apparent that the experimental work necessary to develop, test, and evaluate the systems proposals made by ADSEC will require a substantial amount of laboratory and field effort." Ridenour approached MIT President James Killian with the aim of beginning a development lab similar to the war-era Radiation Laboratory that made enormous progress in radar technology. Killian was initially uninterested, desiring to return the school to its peacetime civilian charter. Ridenour eventually convinced Killian the idea was sound by describing the way the lab would lead to the development of a local electronics industry based on the needs of the lab and the students who would leave the lab to start their
Natural-language user interface
Natural-language user interface (LUI or NLUI) is a type of computer human interface where linguistic phenomena such as verbs, phrases and clauses act as UI controls for creating, selecting and modifying data in software applications. Chatbots are a common implementation of natural-language interfaces, enabling users to interact with software through conversational text or speech. In interface design, natural-language interfaces are sought after for their speed and ease of use, but most suffer the challenges to understanding wide varieties of ambiguous input. Natural-language interfaces are an active area of study in the field of natural-language processing and computational linguistics. An intuitive general natural-language interface is one of the active goals of the Semantic Web. Text interfaces are "natural" to varying degrees. Many formal (un-natural) programming languages incorporate idioms of natural human language. Likewise, a traditional keyword search engine could be described as a "shallow" natural-language user interface. == Overview == A natural-language search engine would in theory find targeted answers to user questions (as opposed to keyword search). For example, when confronted with a question of the form 'which U.S. state has the highest income tax?', conventional search engines ignore the question and instead search on the keywords 'state', 'income' and 'tax'. Natural-language search, on the other hand, attempts to use natural-language processing to understand the nature of the question and then to search and return a subset of the web that contains the answer to the question. If it works, results would have a higher relevance than results from a keyword search engine, due to the question being included. == History == Prototype Nl interfaces had already appeared in the late sixties and early seventies. SHRDLU, a natural-language interface that manipulates blocks in a virtual "blocks world" Lunar, a natural-language interface to a database containing chemical analyses of Apollo 11 Moon rocks by William A. Woods. Chat-80 transformed English questions into Prolog expressions, which were evaluated against the Prolog database. The code of Chat-80 was circulated widely, and formed the basis of several other experimental Nl interfaces. An online demo is available on the LPA website. ELIZA, written at MIT by Joseph Weizenbaum between 1964 and 1966, mimicked a psychotherapist and was operated by processing users' responses to scripts. Using almost no information about human thought or emotion, the DOCTOR script sometimes provided a startlingly human-like interaction. An online demo is available on the LPA website. Janus is also one of the few systems to support temporal questions. Intellect from Trinzic (formed by the merger of AICorp and Aion). BBN's Parlance built on experience from the development of the Rus and Irus systems. IBM Languageaccess Q&A from Symantec. Datatalker from Natural Language Inc. Loqui from BIM Systems. English Wizard from Linguistic Technology Corporation. == Challenges == Natural-language interfaces have in the past led users to anthropomorphize the computer, or at least to attribute more intelligence to machines than is warranted. On the part of the user, this has led to unrealistic expectations of the capabilities of the system. Such expectations will make it difficult to learn the restrictions of the system if users attribute too much capability to it, and will ultimately lead to disappointment when the system fails to perform as expected as was the case in the AI winter of the 1970s and 80s. A 1995 paper titled 'Natural Language Interfaces to Databases – An Introduction', describes some challenges: Modifier attachment The request "List all employees in the company with a driving licence" is ambiguous unless you know that companies can't have driving licences. Conjunction and disjunction "List all applicants who live in California and Arizona" is ambiguous unless you know that a person can't live in two places at once. Anaphora resolution resolve what a user means by 'he', 'she' or 'it', in a self-referential query. Other goals to consider more generally are the speed and efficiency of the interface, in all algorithms these two points are the main point that will determine if some methods are better than others and therefore have greater success in the market. In addition, localisation across multiple language sites requires extra consideration - this is based on differing sentence structure and language syntax variations between most languages. Finally, regarding the methods used, the main problem to be solved is creating a general algorithm that can recognize the entire spectrum of different voices, while disregarding nationality, gender or age. The significant differences between the extracted features - even from speakers who says the same word or phrase - must be successfully overcome. == Uses and applications == The natural-language interface gives rise to technology used for many different applications. Some of the main uses are: Dictation, is the most common use for automated speech recognition (ASR) systems today. This includes medical transcriptions, legal and business dictation, and general word processing. In some cases special vocabularies are used to increase the accuracy of the system. Command and control, ASR systems that are designed to perform functions and actions on the system are defined as command and control systems. Utterances like "Open Netscape" and "Start a new xterm" will do just that. Telephony, some PBX/Voice Mail systems allow callers to speak commands instead of pressing buttons to send specific tones. Wearables, because inputs are limited for wearable devices, speaking is a natural possibility. Medical, disabilities, many people have difficulty typing due to physical limitations such as repetitive strain injuries (RSI), muscular dystrophy, and many others. For example, people with difficulty hearing could use a system connected to their telephone to convert a caller's speech to text. Embedded applications, some new cellular phones include C&C speech recognition that allow utterances such as "call home". This may be a major factor in the future of automatic speech recognition and Linux. Below are named and defined some of the applications that use natural-language recognition, and so have integrated utilities listed above. === Ubiquity === Ubiquity, an add-on for Mozilla Firefox, is a collection of quick and easy natural-language-derived commands that act as mashups of web services, thus allowing users to get information and relate it to current and other webpages. === Wolfram Alpha === Wolfram Alpha is an online service that answers factual queries directly by computing the answer from structured data, rather than providing a list of documents or web pages that might contain the answer as a search engine would. It was announced in March 2009 by Stephen Wolfram, and was released to the public on May 15, 2009. === Siri === Siri is an intelligent personal assistant application integrated with operating system iOS. The application uses natural language processing to answer questions and make recommendations. Siri's marketing claims include that it adapts to a user's individual preferences over time and personalizes results, and performs tasks such as making dinner reservations while trying to catch a cab. === Others === Ask.com – The original idea behind Ask Jeeves (Ask.com) was traditional keyword searching with an ability to get answers to questions posed in everyday, natural language. The current Ask.com still supports this, with added support for math, dictionary, and conversion questions. Braina – Braina is a natural language interface for Windows OS that allows to type or speak English language sentences to perform a certain action or find information. GNOME Do – Allows for quick finding miscellaneous artifacts of GNOME environment (applications, Evolution and Pidgin contacts, Firefox bookmarks, Rhythmbox artists and albums, and so on) and execute the basic actions on them (launch, open, email, chat, play, etc.). hakia – hakia was an Internet search engine. The company invented an alternative new infrastructure to indexing that used SemanticRank algorithm, a solution mix from the disciplines of ontological semantics, fuzzy logic, computational linguistics, and mathematics. hakia closed in 2014. Lexxe – Lexxe was an Internet search engine that used natural-language processing for queries (semantic search). Searches could be made with keywords, phrases, and questions, such as "How old is Wikipedia?" Lexxe closed its search engine services in 2015. Pikimal – Pikimal used natural-language tied to user preference to make search recommendations by template. Pikimal closed in 2015. Powerset – 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, it was purchased by
PyTorch
PyTorch is an open-source deep learning library, originally developed by Meta Platforms and currently developed with support from the Linux Foundation. The successor to Torch, PyTorch provides a high-level API that builds upon optimised, low-level implementations of deep learning algorithms and architectures, such as the Transformer, or SGD. Notably, this API simplifies model training and inference to a few lines of code. PyTorch allows for automatic parallelization of training and, internally, implements CUDA bindings that speed training further by leveraging GPU resources. PyTorch utilises the tensor as a fundamental data type, similarly to NumPy. Training is facilitated by a reversed automatic differentiation system, Autograd, that constructs a directed acyclic graph of the operations (and their arguments) executed by a model during its forward pass. With a loss, backpropagation is then undertaken. As of 2025, PyTorch remains one of the most popular deep learning libraries, alongside others such as TensorFlow and Keras. It can be installed using Anaconda package managers. A number of commercial deep learning architectures are built on top of PyTorch, including ChatGPT, Tesla Autopilot, Uber's Pyro, and Hugging Face's Transformers. == History == In 2001, Torch was written and released under a GPL. It was a machine-learning library written in C++ and CUDA, supporting methods including neural networks, support vector machines (SVM), hidden Markov models, etc. Around 2010, it was rewritten by Ronan Collobert, Clement Farabet and Koray Kavuckuoglu. This was known as Torch7 or LuaTorch. This was written so that the backend was in C and the frontend was in Lua. In mid-2016, some developers refactored it to decouple the frontend and the backend, with strong influence from torch-autograd and Chainer. In turn, torch-autograd was influenced by HIPS/autograd. Development on Torch7 ceased in 2018 and was subsumed by the PyTorch project. Meta (formerly known as Facebook) operates both PyTorch and Convolutional Architecture for Fast Feature Embedding (Caffe2), but models defined by the two frameworks were mutually incompatible. The Open Neural Network Exchange (ONNX) project was created by Meta and Microsoft in September 2017 to decouple deep learning frameworks from hardware-specific runtimes, allowing models to be converted between frameworks and optimized for execution providers like NVIDIA’s TensorRT. Caffe2 was merged into PyTorch at the end of March 2018. In September 2022, Meta announced that PyTorch would be governed by the independent PyTorch Foundation, a newly created subsidiary of the Linux Foundation. PyTorch 2.0 was released on 15 March 2023, introducing TorchDynamo, a Python-level compiler that makes code run up to two times faster, along with significant improvements in training and inference performance across major cloud platforms. == PyTorch tensors == PyTorch defines a class called Tensor (torch.Tensor) to store and operate on homogeneous multidimensional rectangular arrays of numbers. PyTorch supports various sub-types of multi-dimensional arrays, or Tensors. PyTorch Tensors are similar to NumPy Arrays, but can also be operated on by a CUDA-capable NVIDIA GPU. PyTorch has also been developing support for other GPU platforms, for example, AMD's ROCm and Apple's Metal Framework. == PyTorch neural networks == PyTorch defines a module called nn (torch.nn) to describe neural networks and to support training. This module offers a comprehensive collection of building blocks for neural networks, including various layers and activation functions, enabling the construction of complex models. Networks are built by inheriting from the torch.nn module and defining the sequence of operations in the forward() function. == PyTorch Serialized File Format == Pytorch can save and load models using its own file format, which is a ZIP64 archive containing the model weights in a Python pickle file, and other information such as the byte order. The file extensions .pt and .pth are commonly used for these files. == Example == The following program shows the low-level functionality of the library with a simple example. The following code block defines a neural network with linear layers using the nn module.
Stanford Research Institute Problem Solver
The Stanford Research Institute Problem Solver, known by its acronym STRIPS, is an automated planner developed by Richard Fikes and Nils Nilsson in 1971 at SRI International. The same name was later used to refer to the formal language of the inputs to this planner. This language is the base for most of the languages for expressing automated planning problem instances in use today; such languages are commonly known as action languages. This article only describes the language, not the planner. == Definition == A STRIPS instance is composed of: An initial state; The specification of the goal states – situations that the planner is trying to reach; A set of actions. For each action, the following are included: preconditions (what must be established before the action is performed); postconditions (what is established after the action is performed). Mathematically, a STRIPS instance is a quadruple ⟨ P , O , I , G ⟩ {\displaystyle \langle P,O,I,G\rangle } , in which each component has the following meaning: P {\displaystyle P} is a set of conditions (i.e., propositional variables); O {\displaystyle O} is a set of operators (i.e., actions); each operator is itself a quadruple ⟨ α , β , γ , δ ⟩ {\displaystyle \langle \alpha ,\beta ,\gamma ,\delta \rangle } , each element being a set of conditions. These four sets specify, in order, which conditions must be true for the action to be executable, which ones must be false, which ones are made true by the action and which ones are made false; I {\displaystyle I} is the initial state, given as the set of conditions that are initially true (all others are assumed false); G {\displaystyle G} is the specification of the goal state; this is given as a pair ⟨ N , M ⟩ {\displaystyle \langle N,M\rangle } , which specify which conditions are true and false, respectively, in order for a state to be considered a goal state. A plan for such a planning instance is a sequence of operators that can be executed from the initial state and that leads to a goal state. Formally, a state is a set of conditions: a state is represented by the set of conditions that are true in it. Transitions between states are modeled by a transition function, which is a function mapping states into new states that result from the execution of actions. Since states are represented by sets of conditions, the transition function relative to the STRIPS instance ⟨ P , O , I , G ⟩ {\displaystyle \langle P,O,I,G\rangle } is a function succ : 2 P × O → 2 P , {\displaystyle \operatorname {succ} :2^{P}\times O\rightarrow 2^{P},} where 2 P {\displaystyle 2^{P}} is the set of all subsets of P {\displaystyle P} , and is therefore the set of all possible states. The transition function succ {\displaystyle \operatorname {succ} } for a state C ⊆ P {\displaystyle C\subseteq P} , can be defined as follows, using the simplifying assumption that actions can always be executed but have no effect if their preconditions are not met: The function succ {\displaystyle \operatorname {succ} } can be extended to sequences of actions by the following recursive equations: succ ( C , [ ] ) = C {\displaystyle \operatorname {succ} (C,[\ ])=C} succ ( C , [ a 1 , a 2 , … , a n ] ) = succ ( succ ( C , a 1 ) , [ a 2 , … , a n ] ) {\displaystyle \operatorname {succ} (C,[a_{1},a_{2},\ldots ,a_{n}])=\operatorname {succ} (\operatorname {succ} (C,a_{1}),[a_{2},\ldots ,a_{n}])} A plan for a STRIPS instance is a sequence of actions such that the state that results from executing the actions in order from the initial state satisfies the goal conditions. Formally, [ a 1 , a 2 , … , a n ] {\displaystyle [a_{1},a_{2},\ldots ,a_{n}]} is a plan for G = ⟨ N , M ⟩ {\displaystyle G=\langle N,M\rangle } if F = succ ( I , [ a 1 , a 2 , … , a n ] ) {\displaystyle F=\operatorname {succ} (I,[a_{1},a_{2},\ldots ,a_{n}])} satisfies the following two conditions: N ⊆ F {\displaystyle N\subseteq F} M ∩ F = ∅ {\displaystyle M\cap F=\varnothing } == Extensions == The above language is actually the propositional version of STRIPS; in practice, conditions are often about objects: for example, that the position of a robot can be modeled by a predicate A t {\displaystyle At} , and A t ( r o o m 1 ) {\displaystyle At(room1)} means that the robot is in Room1. In this case, actions can have free variables, which are implicitly existentially quantified. In other words, an action represents all possible propositional actions that can be obtained by replacing each free variable with a value. The initial state is considered fully known in the language described above: conditions that are not in I {\displaystyle I} are all assumed false. This is often a limiting assumption, as there are natural examples of planning problems in which the initial state is not fully known. Extensions of STRIPS have been developed to deal with partially known initial states. == A sample STRIPS problem == A monkey is at location A in a lab. There is a box in location C. The monkey wants the bananas that are hanging from the ceiling in location B, but it needs to move the box and climb onto it in order to reach them. Initial state: At(A), Level(low), BoxAt(C), BananasAt(B) Goal state: Have(bananas) Actions: // move from X to Y _Move(X, Y)_ Preconditions: At(X), Level(low) Postconditions: not At(X), At(Y) // climb up on the box _ClimbUp(Location)_ Preconditions: At(Location), BoxAt(Location), Level(low) Postconditions: Level(high), not Level(low) // climb down from the box _ClimbDown(Location)_ Preconditions: At(Location), BoxAt(Location), Level(high) Postconditions: Level(low), not Level(high) // move monkey and box from X to Y _MoveBox(X, Y)_ Preconditions: At(X), BoxAt(X), Level(low) Postconditions: BoxAt(Y), not BoxAt(X), At(Y), not At(X) // take the bananas _TakeBananas(Location)_ Preconditions: At(Location), BananasAt(Location), Level(high) Postconditions: Have(bananas) == Complexity == Deciding whether any plan exists for a propositional STRIPS instance is PSPACE-complete. Various restrictions can be enforced in order to decide if a plan exists in polynomial time or at least make it an NP-complete problem. == Macro operator == In the monkey and banana problem, the robot monkey has to execute a sequence of actions to reach the banana at the ceiling. A single action provides a small change in the game. To simplify the planning process, it make sense to invent an abstract action, which isn't available in the normal rule description. The super-action consists of low level actions and can reach high-level goals. The advantage is that the computational complexity is lower, and longer tasks can be planned by the solver. Identifying new macro operators for a domain can be realized with genetic programming. The idea is, not to plan the domain itself, but in the pre-step, a heuristics is created that allows the domain to be solved much faster. In the context of reinforcement learning, a macro-operator is called an option. Similar to the definition within AI planning, the idea is, to provide a temporal abstraction (span over a longer period) and to modify the game state directly on a higher layer.
Horovod (machine learning)
Horovod is a free and open-source distributed deep learning training framework for TensorFlow, Keras, PyTorch and Apache MXNet. It is designed to scale existing single-GPU training scripts to efficiently run on multiple GPUs and computer nodes with minimal code changes, using synchronous data-parallel training based on the ring-allreduce communication pattern. Horovod was initially developed at Uber and released as an open-source project in 2017, and is now hosted by the LF AI & Data Foundation, a project of the Linux Foundation. == History == Horovod was created at Uber as part of the company's internal machine learning platform Michelangelo to simplify scaling TensorFlow models across many GPUs. The first public release of the library, version 0.9.0, was tagged on GitHub in August 2017 under the Apache 2.0 licence. In October 2017, Uber Engineering publicly introduced Horovod as an open-source component of its deep learning toolkit. In February 2018 Alexander Sergeev and Mike Del Balso published a technical paper describing Horovod's design and benchmarking its performance on up to 512 GPUs, showing near-linear scaling for several image-classification models when compared with single-GPU baselines. In December 2018 Uber contributed Horovod to the LF Deep Learning Foundation (later LF AI & Data), making it a Linux Foundation project. Horovod entered incubation under LF AI & Data and graduated as a full foundation project in 2020. Since its initial release the project has expanded beyond TensorFlow to provide APIs for PyTorch, Keras and Apache MXNet, as well as integrations with frameworks such as Apache Spark and Ray, support for elastic training, and tooling for automated performance tuning and profiling. == Design and features == Horovod core principles are based on the MPI concepts size, rank, local rank, allreduce, allgather, broadcast, and alltoall. Horovod implements synchronous data-parallel training, in which each worker process maintains a replica of the model and computes gradients on different mini-batches of data. The gradients are aggregated across workers using the ring-allreduce communication pattern rather than a central parameter server, which reduces communication bottlenecks and can improve scaling on multi-GPU clusters. Communication is built on top of collective-communication libraries such as MPI, NCCL, Gloo and Intel oneCCL, and supports both GPU and CPU training. In the benchmark experiments reported in the original paper, Horovod achieved around 90% scaling efficiency on 512 GPUs for the ResNet-101 and Inception v3 convolutional neural networks, and around 68% scaling efficiency for the VGG-16 model. Horovod can be deployed on-premises or in cloud environments and is distributed as a Python package with optional GPU support via CUDA. The official documentation provides guides for running Horovod with Docker, Kubernetes (including via Kubeflow and the MPI Operator), commercial platforms such as Databricks, and cluster schedulers such as LSF. == Adoption and use cases == Within Uber, Horovod has been used for applications including autonomous driving research, fraud detection and trip forecasting. Major cloud providers have integrated Horovod into their managed machine learning offerings. Amazon Web Services supports distributed training with Horovod in services such as Amazon SageMaker and AWS Deep Learning Containers, while Microsoft Azure documents Horovod-based training workflows for Azure Synapse Analytics. Technical guides from academic and research computing centres, including Purdue University and the NASA Advanced Supercomputing programme, describe Horovod-based workflows for multi-GPU training on supercomputers and clusters. Horovod is also used in conjunction with Apache Spark and dedicated storage systems as part of end-to-end data processing and model-training pipelines. Industry blogs and technical tutorials describe deployments of Horovod on Kubernetes, on-premises clusters and cloud-managed Kubernetes services such as Amazon EKS.
Textual case-based reasoning
Textual case-based reasoning (TCBR) is a subtopic of case-based reasoning, in short CBR, a popular area in artificial intelligence. CBR suggests the ways to use past experiences to solve future similar problems, requiring that past experiences be structured in a form similar to attribute-value pairs. This leads to the investigation of textual descriptions for knowledge exploration whose output will be, in turn, used to solve similar problems. == Subareas == Textual case-base reasoning research has focused on: measuring similarity between textual cases mapping texts into structured case representations adapting textual cases for reuse automatically generating representations.
Pax Silica
Pax Silica is a United States-led international initiative focused on strengthening and coordinating "trusted" supply chains for advanced technologies—especially semiconductors, artificial intelligence (AI) infrastructure, critical minerals, advanced manufacturing, logistics, and associated energy and data infrastructure. The initiative is coordinated by the US Department of State and was launched in December 2025 alongside the signing of the non-binding Pax Silica Declaration by an initial group of partner countries. The initiative describes itself as a "positive-sum" partnership intended to reduce "coercive dependencies" and improve resilience across the full technology stack, from mineral extraction and processing through chip manufacturing and computing infrastructure. US officials described Pax Silica as a framework for coordinating flagship projects and policy alignment across partner countries, including supply-chain mapping, investment and co-investment initiatives, and protection of critical infrastructure and sensitive technologies. Reuters reported discussions of projects linked to trade and logistics routes and an industrial park initiative in Israel. Gulf countries, such as the UAE and Qatar, are betting on attracting AI companies with cheap energy. Moreover, the UAE's potential to invest in Pax Silica's activities has been noted as a fundamental asset for the initiative. In early 2026, the U.S. announced plans to contribute $250M toward an investmest consortium that's intended to strengthen energy and critical mineral supply chains. == Launch and background == During the 2020s, governments increasingly treated supply-chain resilience in semiconductors, critical minerals, and AI-related computing infrastructure as a national-security priority, amid export controls, industrial policy measures, and geopolitical competition over the technologies underpinning advanced manufacturing and AI. Pax Silica was presented by US officials as an economic-security framework aimed at aligning policies and investment among "trusted partners" that host major technology firms and key industrial capacity. Pacific Forum's analyst Akhil Ramesh, writing for the National Interest magazine, described the initiative as understanding that: "economic security today is inseparable from control over energy, critical minerals, high-end manufacturing, and advanced models." On December 11, 2025, the US Department of State announced the inaugural Pax Silica Summit and a planned signing of the Pax Silica Declaration, describing Pax Silica as the Department's flagship effort on AI and supply-chain security. The initial summit was held in Washington, D.C. on December 12, 2025. The State Department fact sheet described cooperation areas including connectivity and data infrastructure, compute and semiconductors, advanced manufacturing, logistics, mineral refining and processing, and energy. == Membership == Pax Silica participation has been discussed in terms of (1) countries that have signed the declaration and (2) countries invited to summit discussions or publicly reported as prospective signatories but which had not (as of mid-January 2026) signed the declaration. === Countries that signed the Pax Silica Declaration === Seven countries signed the declaration at the December 12, 2025, summit in Washington, D.C.: Australia Israel Japan South Korea Singapore United Kingdom United States Some countries who attended the initial conversations did not immediately sign, while additional countries were invited to join after the discussions concluded. The following are the later signatory countries on the declaration: Greece Netherlands (joined December 17, 2025; "non-signing partner") Qatar (joined January 13, 2026) United Arab Emirates (joined January 14, 2026) India (joined February 20, 2026) Sweden (signed March 17, 2026) Finland (signed April 16, 2026) Philippines (signed April 17, 2026) Norway (signed May 6, 2026) === Countries invited / participating, but not yet signed === At launch, US materials and contemporaneous reporting described additional invited participants and observers, including: Canada – observer/participant in related discussions, per US briefing materials; not listed among signatories. Taiwan – participated in summit sessions according to a State Department briefing; not listed among signatories. The Organisation for Economic Co-operation and Development (OECD) and European Union were also noted by US officials as present in an observer capacity, but are not countries.