In the context of IBM mainframe computers in the IBM System/360 line and its successors, a data set (IBM preferred) or dataset is a computer file having a record organization. Use of this term began with, e.g., DOS/360 and OS/360, and is still used by their successors, including the current VSE and z/OS. Documentation for these systems historically preferred this term rather than file. A data set is typically stored on a direct access storage device (DASD) or magnetic tape, however unit record devices, such as punch card readers, card punches, line printers and page printers can provide input/output (I/O) for a data set (file). Data sets are not unstructured streams of bytes, but rather are organized in various logical record and block structures determined by the DSORG (data set organization), RECFM (record format), and other parameters. These parameters are specified at the time of the data set allocation (creation), for example with Job Control Language DD statements. Within a running program they are stored in the Data Control Block (DCB) or Access Control Block (ACB), which are data structures used to access data sets using access methods. Records in a data set may be fixed, variable, or “undefined” length. == Data set organization == For OS/360, the DCB's DSORG parameter specifies how the data set is organized. It may be CQ Queued Telecommunications Access Method (QTAM) in Message Control Program (MCP) CX Communications line group DA Basic Direct Access Method (BDAM) GS Graphics device for Graphics Access Method(GAM) IS Indexed Sequential Access Method (ISAM) MQ QTAM message queue in application PO Partitioned Organization PS Physical Sequential among others. Data sets on tape may only be DSORG=PS. The choice of organization depends on how the data is to be accessed, and in particular, how it is to be updated. Programmers utilize various access methods (such as QSAM or VSAM) in programs for reading and writing data sets. Access method depends on the given data set organization. == Record format (RECFM) == Regardless of organization, the physical structure of each record is essentially the same, and is uniform throughout the data set. This is specified in the DCB RECFM parameter. RECFM=F means that the records are of fixed length, specified via the LRECL parameter. RECFM=V specifies a variable-length record. V records when stored on media are prefixed by a Record Descriptor Word (RDW) containing the integer length of the record in bytes and flag bits. With RECFM=FB and RECFM=VB, multiple logical records are grouped together into a single physical block on tape or DASD. FB and VB are fixed-blocked, and variable-blocked, respectively. RECFM=U (undefined) is also variable length, but the length of the record is determined by the length of the block rather than by a control field. The BLKSIZE parameter specifies the maximum length of the block. RECFM=FBS could be also specified, meaning fixed-blocked standard, meaning all the blocks except the last one were required to be in full BLKSIZE length. RECFM=VBS, or variable-blocked spanned, means a logical record could be spanned across two or more blocks, with flags in the RDW indicating whether a record segment is continued into the next block and/or was continued from the previous one. This mechanism eliminates the need for using any "delimiter" byte value to separate records. Thus data can be of any type, including binary integers, floating-point, or characters, without introducing a false end-of-record condition. The data set is an abstraction of a collection of records, in contrast to files as unstructured streams of bytes. == Partitioned data set == A partitioned data set (PDS) is a data set containing multiple members, each of which holds a separate sub-data set, similar to a directory in other types of file systems. This type of data set is often used to hold load modules (old format bound executable programs), source program libraries (especially Assembler macro definitions), ISPF screen definitions, and Job Control Language. A PDS may be compared to a Zip file or COM Structured Storage. A Partitioned Data Set can only be allocated on a single volume and have a maximum size of 65,535 tracks. Besides members, a PDS contains also a directory. Each member can be accessed indirectly via the directory structure. Once a member is located, the data stored in that member are handled in the same manner as a PS (sequential) data set. Whenever a member is deleted, the space it occupied is unusable for storing other data. Likewise, if a member is re-written, it is stored in a new spot at the back of the PDS and leaves wasted “dead” space in the middle. The only way to recover “dead” space is to perform file compression. Compression, which is done using the IEBCOPY utility, moves all members to the front of the data space and leaves free usable space at the back. (Note that in modern parlance, this kind of operation might be called defragmentation or garbage collection; data compression nowadays refers to a different, more complicated concept.) PDS files can only reside on DASD, not on magnetic tape, in order to use the directory structure to access individual members. Partitioned data sets are most often used for storing multiple job control language files, utility control statements, and executable modules. An improvement of this scheme is a Partitioned Data Set Extended (PDSE or PDS/E, sometimes just libraries) introduced with DFSMSdfp for MVS/XA and MVS/ESA systems. A PDS/E library can store program objects or other types of members, but not both. BPAM cannot process a PDS/E containing program objects. PDS/E structure is similar to PDS and is used to store the same types of data. However, PDS/E files have a better directory structure which does not require pre-allocation of directory blocks when the PDS/E is defined (and therefore does not run out of directory blocks if not enough were specified). Also, PDS/E automatically stores members in such a way that compression operation is not needed to reclaim "dead" space. PDS/E files can only reside on DASD in order to use the directory structure to access individual members. == Generation Data Group == A Generation Data Group (GDG) is a group of non-VSAM data sets that are successive generations of historically-related data stored on an IBM mainframe (running OS/360 and its successors or DOS/360 and its successors). A GDG is usually cataloged. An individual member of the GDG collection is called a "Generation Data Set." The latter may be identified by an absolute number, ACCTG.OURGDG(1234), or a relative number: (-1) for the previous generation, (0) for the current one, and (+1) the next generation. A GDG specifies how many generations of a data set are to be kept and at what age a generation will be deleted. Whenever a new generation is created, the system checks whether one or more obsolete generations are to be deleted. The purpose of GDGs is to automate archival, using the command language JCL, the data set name given is generic. When DSN appears, the GDG data set appears along with the history number, where (0) is the most recent version (-1), (-2), ... are previous generations (+1) a new generation (see DD) Another use of GDGs is to be able to address all generations simultaneously within a JCL script without having to know the number of currently available generations. To do this, you have to omit the parentheses and the generation number in the JCL when specifying the dataset. === GDG JCL & features === Generation Data Groups are defined using either the BLDG statement of the IEHPROGM utility or the DEFINE GENERATIONGROUP statement of the newer IDCAMS utility, which allows setting various parameters. LIMIT(10) would limit the number of generations limit to 10. SCRATCH FOR (91) would retain each member, up to the limited#generations, at least 91 days. IDCAMS can also delete (and optionally uncatalog) a GDG. ==== Example ==== Creation of a standard GDG for five safety scopes, each at least 35 days old: Delete a standard GDG:
Software design
Software design is the process of conceptualizing how a software system will work before it is implemented or modified. Software design also refers to the direct result of the design process – the concepts of how the software will work which may be formally documented or may be maintained less formally, including via oral tradition. The design process enables a designer to model aspects of a software system before it exists with the intent of making the effort of writing the code more efficiently. Creativity, past experience, a sense of what makes "good" software, and a commitment to quality are success factors for a competent design. A software design can be compared to an architected plan for a house. High-level plans represent the totality of the house (e.g., a three-dimensional rendering of the house). Lower-level plans provide guidance for constructing each detail (e.g., the plumbing lay). Similarly, the software design model provides a variety of views of the proposed software solution. == Part of the overall process == In terms of the waterfall development process, software design is the activity that occurs after requirements analysis and before coding. Requirements analysis determines what the system needs to do without determining how it will do it, and thus, multiple designs can be imagined that satisfy the requirements. The design can be created while coding, without a plan or requirements analysis, but for more complex projects this is less feasible. Completing a design prior to coding allows for multidisciplinary designers and subject-matter experts to collaborate with programmers to produce software that is useful and technically sound. Sometimes, a simulation or prototype is created to model the system in an effort to determine a valid and good design. == Code as design == A common point of confusion with the term design in software is that the process applies at multiple levels of abstraction such as a high-level software architecture and lower-level components, functions and algorithms. A relatively formal process may occur at high levels of abstraction but at lower levels, the design process is almost always less formal where the only artifact of design may be the code itself. To the extent that this is true, software design refers to the design of the design. Edsger W. Dijkstra referred to this layering of semantic levels as the "radical novelty" of computer programming, and Donald Knuth used his experience writing TeX to describe the futility of attempting to design a program prior to implementing it: TEX would have been a complete failure if I had merely specified it and not participated fully in its initial implementation. The process of implementation constantly led me to unanticipated questions and to new insights about how the original specifications could be improved. == Artifacts == A design process may include the production of art Software design documentation such as flow chart, use case, Pseudocode, Unified Modeling Language model and other Fundamental modeling concepts. For user centered software, design may involve user experience design yielding a storyboard to help determine those specifications. Documentation may be reviewed to allow constraints, specifications and even requirements to be adjusted prior to coding. == Iterative design == Software systems inherently deal with uncertainties, and the size of software components can significantly influence a system's outcomes, both positively and negatively. Neal Ford and Mark Richards propose an iterative approach to address the challenge of identifying and right-sizing components. This method emphasizes continuous refinement as teams develop a more nuanced understanding of system behavior and requirements. The approach typically involves a cycle with several stages: A high-level partitioning strategy is established, often categorized as technical or domain-based. Guidelines for the smallest meaningful deployable unit, referred to as "quanta," are defined. While these foundational decisions are made early, they may be revisited later in the cycle if necessary. Initial components are identified based on the established strategy. Requirements are assigned to the identified components. The roles and responsibilities of each component are analyzed to ensure clarity and minimize overlap. Architectural characteristics, such as scalability, fault tolerance, and maintainability, are evaluated. Components may be restructured based on feedback from development teams. This cycle serves as a general framework and can be adapted to different domains. == Design principles == Design principles enable a software engineer to navigate the design process. Davis suggested principles which have been refined over time as: The design process should not suffer from "tunnel vision" A good designer should consider alternative approaches, judging each based on the requirements of the problem, the resources available to do the job. The design should be traceable to the analysis model Because a single element of the design model can often be traced back to multiple requirements, it is necessary to have a means for tracking how requirements have been satisfied by the design model. The design should not reinvent the wheel Systems are constructed using a set of design patterns, many of which have likely been encountered before. These patterns should always be chosen as an alternative to reinvention. Time is short and resources are limited; design time should be invested in representing (truly new) ideas by integrating patterns that already exist (when applicable). The design should "minimize the intellectual distance" between the software and the problem as it exists in the real world That is, the structure of the software design should, whenever possible, mimic the structure of the problem domain. The design should exhibit uniformity and integration A design is uniform if it appears fully coherent. In order to achieve this outcome, rules of style and format should be defined for a design team before design work begins. A design is integrated if care is taken in defining interfaces between design components. The design should be structured to accommodate change The design concepts discussed in the next section enable a design to achieve this principle. The design should be structured to degrade gently, even when aberrant data, events, or operating conditions are encountered Well-designed software should never "bomb"; it should be designed to accommodate unusual circumstances, and if it must terminate processing, it should do so in a graceful manner. Design is not coding, coding is not design Even when detailed procedural designs are created for program components, the level of abstraction of the design model is higher than the source code. The only design decisions made at the coding level should address the small implementation details that enable the procedural design to be coded. The design should be assessed for quality as it is being created, not after the fact A variety of design concepts and design measures are available to assist the designer in assessing quality throughout the development process. The design should be reviewed to minimize conceptual (semantic) errors There is sometimes a tendency to focus on minutiae when the design is reviewed, missing the forest for the trees. A design team should ensure that major conceptual elements of the design (omissions, ambiguity, inconsistency) have been addressed before worrying about the syntax of the design model. == Design concepts == Design concepts provide a designer with a foundation from which more sophisticated methods can be applied. Design concepts include: Abstraction Reducing the information content of a concept or an observable phenomenon, typically to retain only information that is relevant for a particular purpose. It is an act of Representing essential features without including the background details or explanations. Architecture The overall structure of the software and the ways in which that structure provides conceptual integrity for a system. Good software architecture will yield a good return on investment with respect to the desired outcome of the project, e.g. in terms of performance, quality, schedule and cost. Control hierarchy A program structure that represents the organization of a program component and implies a hierarchy of control. Data structure Representing the logical relationship between elements of data. Design pattern A designer may identify a design aspect of the system that has solved in the past. The reuse of such patterns can increase software development velocity. Information hiding Modules should be specified and designed so that information contained within a module is inaccessible to other modules that have no need for such information. Modularity Dividing the solution into parts (modules). Refinement The process of elaboration. A hierarchy is developed by decomposing a macrosco
Automation in construction
Automation in construction is the combination of methods, processes, and systems that allow for greater machine autonomy in construction activities. Construction automation may have multiple goals, including but not limited to, reducing jobsite injuries, decreasing activity completion times, and assisting with quality control and quality assurance. Some systems may be fielded as a direct response to increasing skilled labor shortages in some countries. Opponents claim that increased automation may lead to less construction jobs and that software leaves heavy equipment vulnerable to hackers. Research insights on this subject are today published in several journals such as Automation in Construction by Elsevier. == Uses of automation in construction == Equipment control and management: Automation can be used to control and monitor construction equipment, such as cranes, excavators, and bulldozers. Material handling: Automated systems can be used to handle, transport, and place materials such as concrete, bricks, and stones. Surveying: Automated survey equipment and drones can be used to collect and analyze data on construction sites. Quality control: Automated systems can be used to monitor and control the quality of materials and construction processes. Safety management: Automated systems can be used to monitor and control safety conditions on construction sites. Scheduling and planning: Automated systems can be used to manage schedules, resources, and costs. Waste management: Automated systems can be used to manage and dispose of waste materials generated during construction. 3D printing: Automated 3D printing can be used to create prototypes, models, and even full-scale building components. == Autonomous heavy equipment == Advances in sensors, machine learning, and autonomous vehicle technology have led to the development of self-operating construction equipment and retrofit systems designed to automate excavators, bulldozers, tracked loaders, skid steer loaders, and haul trucks, allowing them to perform tasks with limited human supervision. Since 2017, tech companies have developed autonomous or semi-autonomous retrofit kits that can be installed on existing construction machinery. Examples include Bedrock Robotics, Built Robotics, and SafeAI, which develop sensor and software systems that enable excavators and other earthmoving machines to operate with varying degrees of autonomy. Major equipment manufacturers have also introduced autonomous capabilities: Caterpillar and John Deere have developed autonomous or semi-autonomous systems for construction and mining equipment, including haul trucks and earthmoving machines. == Transportation сonstruction == Kratos Defense & Security Solutions fielded the world’s first Autonomous Truck-Mounted Attenuator (ATMA) in 2017, in conjunction with Royal Truck & Equipment. == Benefits of automation in construction == The use of automation in construction has become increasingly prevalent in recent years due to its numerous benefits. Automation in construction refers to the use of machinery, software, and other technologies to perform tasks that were previously done manually by workers. One of the most significant benefits of automation in construction is increased productivity. Automation can help speed up construction processes, reduce project completion times, and improve overall efficiency. For example, using automated machinery for tasks such as concrete pouring, bricklaying, and welding can significantly increase the speed and accuracy of these tasks, allowing for more work to be completed in a shorter amount of time. Another benefit of automation in construction is improved safety. By automating tasks that are hazardous to workers, such as demolition or working at height, companies can reduce the risk of accidents and injuries on site. Automation can also help to reduce worker fatigue, which can be a significant factor in accidents and mistakes. Overall, the use of automation in construction can improve productivity, reduce costs, increase safety, and improve the quality of construction projects. As technology continues to advance, the use of automation is likely to become even more prevalent in the construction industry.
Something Big Is Happening
"Something Big Is Happening" is an essay by Matt Shumer, an AI entrepreneur, about the impact of artificial intelligence, published in February 2026, that has since been reportedly viewed more than 80 million times and widely discussed. Shumer noted that the technology has crossed an important threshold, where AI has become capable of creating self-improving systems. Referring to one the most recent AI models, he wrote: "It was making intelligent decisions. It had something that felt, for the first time, like judgment. Like taste." Speaking to CNBC's Power Lunch, Shumer said that his "core message" is "people in the workforce should start to use and experiment with AI tools so they can understand what’s coming". Even as the essay was widely shared and discussed, the essay also elicited criticism. Paulo Carvao, in an essay published by the Forbes Magazine stated that some of his advice is sound, but added: "It reads at times like a sales pitch. He urges readers to subscribe to the most advanced AI tools. He implies that those with access to premium models will outpace those without. He frames paid AI subscriptions as a form of insurance against obsolescence." Writing in The Guardian, Dan Milmo and Aisha Down mentioned Shumer as having a history of AI hype and stated, "He previously excited the internet by announcing the release of the world's "top open-source model", which it was not". Many workers in the technology sector criticized the article in blog posts shared on Hacker News; Edward Zitron commented that "while coding LLMs can test products, or scan/fix some bugs, this suggests they A) do this autonomously without human input, B) they do this correctly every time (or ever!)." In an article alluding to Shumer's original post, Ari Colaprete wrote "the LLM is fundamentally a writing machine, it does everything via text, and if you make it produce writing that exists purely to serve some sort of mechanical function, and you train it to succeed in that task, then it will tend to do so, even with vast intricacy."
Environmental impact of AI
The environmental impact of the design, training, deployment and use of artificial intelligence includes the greenhouse gas emissions from generating electricity for data centres and computing hardware, operational and upstream water use, and material impacts from hardware manufacturing, mining and electronic waste. Estimating AI's environmental effects can be difficult because results depend on how impacts are measured, including whether accounting includes only model computation or also data-centre overhead, idle capacity, hardware manufacture, and local electricity supply. As these issues have received greater attention, governments and regulators have increasingly considered data-centre reporting requirements, energy-efficiency standards, and broader transparency measures for AI-related resource use. == Carbon footprint and energy use == AI-related energy use arises at multiple stages, including model training, fine-tuning, inference, storage, networking, and supporting infrastructure such as cooling and power conversion. === Individual level === Published estimates of energy use per AI request vary widely across models, tasks and measurement methods. A benchmark study presented at the 2024 ACM Conference on Fairness, Accountability, and Transparency found substantial differences between task types, with lower energy use for some text tasks and much higher energy use for image generation in the study's test conditions. In that benchmark, simple classification tasks consumed about 0.002–0.007 Wh per prompt on average (about 9% of a smartphone charge for 1,000 prompts), while text generation and text summarisation each used about 0.05 Wh per prompt; image generation averaged 2.91 Wh per prompt, and the least efficient image model in the study used 11.49 Wh per image (roughly equivalent to half a smartphone charge). First-party measurements in production environments have also been published. A 2025 Google study on Gemini assistant serving reported median per-prompt energy, emissions, and water-use estimates under the authors' accounting framework, while noting that different system boundaries can produce substantially different results. The study reported a median text-prompt estimate of about 0.24 Wh, which is roughly as much energy as watching nine seconds of television. The study also stated that software and infrastructure improvements reduced energy use by a factor of 33 and carbon emissions by a factor of 44 for a typical prompt over one year within the authors' framework. Researchers at the University of Michigan measured the energy consumption of various Meta Llama 3.1 models released in 2024 and found that smaller language models (8 billion parameters) use about 114 joules (0.03167 Wh) per response, while larger models (405 billion parameters) require up to 6,700 joules (1.861 Wh) per response. This corresponds to the energy needed to run a microwave oven for roughly one-tenth of a second and eight seconds, respectively. Comparisons between AI systems and human labour for specific tasks have produced mixed results and remain sensitive to assumptions about output quality, workload and system boundaries. A 2024 study in Scientific Reports reported 130 to 2900 times lower estimated carbon emissions for selected AI systems than for human writers and illustrators under its assumptions. A later Scientific Reports paper reported a counterexample for programming tasks under its assumptions, finding 5 to 19 times higher estimated emissions for the evaluated AI system than for human programmers on the benchmark used in that study. === System level === ==== Energy use and efficiency ==== AI electricity intensity depends not only on model architecture but also on hardware and facility efficiency. Data-centre operators commonly report Power usage effectiveness (PUE), which measures the ratio of total facility energy to IT equipment energy; a lower PUE indicates less overhead energy for cooling and other supporting infrastructure. Operators may also publish metrics and case studies on hardware efficiency, cooling systems and power sourcing. In its 2024 environmental report, Google stated that its 2023 total greenhouse gas emissions increased 13% year over year, primarily because of increased data-centre energy consumption and supply-chain emissions, while also reporting lower PUE than industry averages for its own facilities. The International Energy Agency has also reported that data centres remain a relatively small share of global electricity use overall, but that their local effects can be much more pronounced because demand is geographically concentrated. ==== Carbon footprint ==== At system level, AI contributes to rising electricity demand in data centres and related infrastructure. The International Energy Agency estimated that data centres used about 415 TWh of electricity in 2024, or around 1.5% of global electricity consumption, and projected that data-centre electricity use could rise to about 945 TWh by 2030, with AI identified as the main driver of that growth alongside other digital services. The carbon footprint of AI systems depends strongly on electricity sources, hardware efficiency, utilisation rates, and what stages are included in the accounting. Training large models can require substantial electricity, while total lifecycle impacts also depend on deployment scale and the amount of inference performed after training. Early analyses of frontier-model development reported rapid historical growth in training compute for selected systems, although later trends have depended on changes in model design, hardware and efficiency gains. Accounting methods that include upstream or embodied impacts, such as hardware manufacture and facilities construction, can materially affect estimates of AI-related emissions. === Decisions and strategies by individual companies === Large technology companies have reported that the expansion of AI and cloud infrastructure affects their sustainability targets, electricity demand, and resource use. Google, for example, attributed part of its emissions growth in 2023 to increased data-centre energy consumption and supply-chain emissions in its 2024 environmental report. Cloud and AI companies have also announced measures intended to reduce environmental impacts, including investment in more efficient hardware, low-carbon electricity procurement, alternative cooling systems, and water stewardship programmes. The extent, comparability, and third-party verification of such disclosures vary between firms and jurisdictions. == Water usage == Data centres can use water directly for cooling and indirectly through the water used in electricity generation, depending on the local energy mix. Public reporting on data-centre water use has often been inconsistent, making comparisons between operators and regions difficult. To standardise operational reporting, The Green Grid proposed the metric water usage effectiveness (WUE), defined as annual site water use divided by IT equipment energy use. WUE does not by itself measure local water stress, source sustainability, or all upstream water impacts. Studies of AI water use also distinguish between water withdrawal and water consumption. Research on AI-specific water use has argued that the water footprint of AI systems can be difficult to observe and may vary substantially by location, cooling design, and electricity source. A 2025 Communications of the ACM article summarised methods for estimating AI water footprints and emphasised the distinction between water withdrawal and water consumption. Li and colleagues estimated that global AI water withdrawal could reach 4.2–6.6 billion cubic metres in 2027 under the scenarios examined in their article. Using GPT-3, released by OpenAI in 2020, as an example, they estimated that training the model in Microsoft's U.S. data centres could consume about 700,000 litres of onsite water and about 5.4 million litres in total when offsite electricity-related water use was included; they also estimated that 10–50 medium-length GPT-3 responses could consume about 500 mL of water, depending on when and where the model was deployed. Published prompt-level estimates have also varied by system and accounting framework: the 2025 Google study on Gemini assistant serving reported a median text-prompt estimate of about 0.26 mL under its framework. Location can materially affect the significance of data-centre water use. Research on U.S. data centres found that one-fifth of servers' direct water footprint came from moderately to highly water-stressed watersheds, while nearly half of servers were fully or partially powered by plants located in water-stressed regions. A 2025 Reuters report, citing data from Verisk Maplecroft and NatureFinance, said that an average mid-sized data centre uses about 1.4 million litres of water per day for cooling and that Phoenix would experience a 32% increase in annual water stress if currently pl
System integrity
In telecommunications, the term system integrity has the following meanings: That condition of a system wherein its mandated operational and technical parameters are within the prescribed limits. The quality of an AIS when it performs its intended function in an unimpaired manner, free from deliberate or inadvertent unauthorized manipulation of the system. The state that exists when there is complete assurance that under all conditions an IT system is based on the logical correctness and reliability of the operating system, the logical completeness of the hardware and software that implement the protection mechanisms, and data integrity.
Active learning (machine learning)
Active learning is a special case of machine learning in which a learning algorithm can interactively query a human user (or some other information source) to label new data points with the desired outputs. The human user must possess expertise in the problem domain, including the ability to consult authoritative sources when necessary. In statistics literature, it is sometimes also called optimal experimental design. The information source is also called teacher or oracle. There are situations in which unlabeled data is abundant but manual labeling is expensive. In such a scenario, learning algorithms can actively query the teacher for labels. Since the learner chooses the examples, the number of examples to learn a concept can often be much lower than the number required in normal supervised learning. However, there is a risk that the algorithm is overwhelmed by uninformative examples. Recent developments are dedicated to multi-label active learning, hybrid active learning and active learning in a single-pass (on-line) context, combining concepts from the field of machine learning (e.g. conflict and ignorance) with adaptive, incremental learning policies in the field of online machine learning. Using active learning allows for faster development of a machine learning algorithm, when comparative updates would require a quantum or super computer. Large-scale active learning projects may benefit from crowdsourcing frameworks such as Amazon Mechanical Turk that include many humans in the active learning loop. == Definitions == Let T be the total set of all data under consideration. For example, in a protein engineering problem, T would include all proteins that are known to have a certain interesting activity and all additional proteins that one might want to test for that activity. During each iteration, i, T is broken up into three subsets T K , i {\displaystyle \mathbf {T} _{K,i}} : Data points where the label is known. T U , i {\displaystyle \mathbf {T} _{U,i}} : Data points where the label is unknown. T C , i {\displaystyle \mathbf {T} _{C,i}} : A subset of TU,i that is chosen to be labeled. Most of the current research in active learning involves the best method to choose the data points for TC,i. == Scenarios == Pool-based sampling: In this approach, which is the most well known scenario, the learning algorithm attempts to evaluate the entire dataset before selecting data points (instances) for labeling. It is often initially trained on a fully labeled subset of the data using a machine-learning method such as logistic regression or SVM that yields class-membership probabilities for individual data instances. The candidate instances are those for which the prediction is most ambiguous. Instances are drawn from the entire data pool and assigned a confidence score, a measurement of how well the learner "understands" the data. The system then selects the instances for which it is the least confident and queries the teacher for the labels. The theoretical drawback of pool-based sampling is that it is memory-intensive and is therefore limited in its capacity to handle enormous datasets, but in practice, the rate-limiting factor is that the teacher is typically a (fatiguable) human expert who must be paid for their effort, rather than computer memory. Stream-based selective sampling: Here, each consecutive unlabeled instance is examined one at a time with the machine evaluating the informativeness of each item against its query parameters. The learner decides for itself whether to assign a label or query the teacher for each datapoint. As contrasted with Pool-based sampling, the obvious drawback of stream-based methods is that the learning algorithm does not have sufficient information, early in the process, to make a sound assign-label-vs ask-teacher decision, and it does not capitalize as efficiently on the presence of already labeled data. Therefore, the teacher is likely to spend more effort in supplying labels than with the pool-based approach. Membership query synthesis: This is where the learner generates synthetic data from an underlying natural distribution. For example, if the dataset are pictures of humans and animals, the learner could send a clipped image of a leg to the teacher and query if this appendage belongs to an animal or human. This is particularly useful if the dataset is small. The challenge here, as with all synthetic-data-generation efforts, is in ensuring that the synthetic data is consistent in terms of meeting the constraints on real data. As the number of variables/features in the input data increase, and strong dependencies between variables exist, it becomes increasingly difficult to generate synthetic data with sufficient fidelity. For example, to create a synthetic data set for human laboratory-test values, the sum of the various white blood cell (WBC) components in a white blood cell differential must equal 100, since the component numbers are really percentages. Similarly, the enzymes alanine transaminase (ALT) and aspartate transaminase (AST) measure liver function (though AST is also produced by other tissues, e.g., lung, pancreas) A synthetic data point with AST at the lower limit of normal range (8–33 units/L) with an ALT several times above normal range (4–35 units/L) in a simulated chronically ill patient would be physiologically impossible. == Query strategies == Algorithms for determining which data points should be labeled can be organized into a number of different categories, based upon their purpose: Balance exploration and exploitation: the choice of examples to label is seen as a dilemma between the exploration and the exploitation over the data space representation. This strategy manages this compromise by modelling the active learning problem as a contextual bandit problem. For example, Bouneffouf et al. propose a sequential algorithm named Active Thompson Sampling (ATS), which, in each round, assigns a sampling distribution on the pool, samples one point from this distribution, and queries the oracle for this sample point label. Expected model change: label those points that would most change the current model. Expected error reduction: label those points that would most reduce the model's generalization error. Exponentiated Gradient Exploration for Active Learning: In this paper, the author proposes a sequential algorithm named exponentiated gradient (EG)-active that can improve any active learning algorithm by an optimal random exploration. Uncertainty sampling: label those points for which the current model is least certain as to what the correct output should be. Query by committee: a variety of models are trained on the current labeled data, and vote on the output for unlabeled data; label those points for which the "committee" disagrees the most Querying from diverse subspaces or partitions: When the underlying model is a forest of trees, the leaf nodes might represent (overlapping) partitions of the original feature space. This offers the possibility of selecting instances from non-overlapping or minimally overlapping partitions for labeling. Variance reduction: label those points that would minimize output variance, which is one of the components of error. Conformal prediction: predicts that a new data point will have a label similar to old data points in some specified way and degree of the similarity within the old examples is used to estimate the confidence in the prediction. Mismatch-first farthest-traversal: The primary selection criterion is the prediction mismatch between the current model and nearest-neighbour prediction. It targets on wrongly predicted data points. The second selection criterion is the distance to previously selected data, the farthest first. It aims at optimizing the diversity of selected data. User-centered labeling strategies: Learning is accomplished by applying dimensionality reduction to graphs and figures like scatter plots. Then the user is asked to label the compiled data (categorical, numerical, relevance scores, relation between two instances). A wide variety of algorithms have been studied that fall into these categories. While the traditional AL strategies can achieve remarkable performance, it is often challenging to predict in advance which strategy is the most suitable in a particular situation. In recent years, meta-learning algorithms have been gaining in popularity. Some of them have been proposed to tackle the problem of learning AL strategies instead of relying on manually designed strategies. A benchmark which compares 'meta-learning approaches to active learning' to 'traditional heuristic-based Active Learning' may give intuitions if 'Learning active learning' is at the crossroads == Minimum marginal hyperplane == Some active learning algorithms are built upon support-vector machines (SVMs) and exploit the structure of the SVM to determine which data points to label. Such methods usually calculate the margin, W, of each u