AI & Machine Learning Glossary
A
Abductive Logic Programming
A framework in AI/ML for solving problems declaratively using abductive reasoning, allowing for incompletely defined predicates.
Abductive Reasoning
In AI/ML, abductive reasoning is inferring the most likely explanation from incomplete or ambiguous data.
Abstract Data Type (ADT)
A mathematical model defining a data type by its behavior, operations, and values from a user's perspective.
Abstraction
Simplifying complex systems by focusing on essential features, ignoring irrelevant detail in AI/ML contexts.
Accelerating Change
In AI/ML, it refers to the rapid advancement of technology, leading to faster, more impactful innovations.
Action Language
A formalism in AI for modeling state transitions resulting from actions within dynamic systems.
Action Model Learning
Learning how actions affect an environment, informing software agents' decision-making in AI/ML.
Action Selection
Deciding the next action for intelligent systems based on current state, goals, and available actions.
Activation Function
Determines a neural network node's output based on its input, influencing the network's ability to learn complex patterns.
Active Learning
A strategy where the model identifies uncertain data points for labeling to optimize the annotation process.
Adaptive Algorithm
Dynamically adjusts its strategy during runtime based on feedback to optimize performance in AI/ML tasks.
Adaptive Neuro Fuzzy Inference System (ANFIS)
Integrates neural networks and fuzzy logic to approximate nonlinear functions with learning capabilities.
Admissible Heuristic
A heuristic that never overestimates the cost to reach a goal in pathfinding and search algorithms.
Adversarial Examples
Inputs to machine learning models intentionally designed to cause the model to make a mistake.
Affective Computing
Computing that recognizes, interprets, and simulates human emotions using interdisciplinary approaches.
Agent Architecture
Design framework for intelligent agents, outlining component arrangement for decision-making and behavior execution.
AI Accelerator
Hardware designed to speed up AI tasks, particularly in neural networks, machine vision, and machine learning.
AI-complete
Problems as complex as achieving general human-level intelligence, unsolvable by simple algorithms.
Algorithm
A step-by-step procedure for calculations, data processing, and automated reasoning in computing and AI/ML.
Algorithmic Efficiency
Measures an algorithm's resource usage, including time and space, crucial for optimizing AI/ML performance.
Algorithmic Probability
A method from algorithmic information theory for assigning prior probabilities to observations, developed by Ray Solomonoff.
AlphaGo
AI program by Google DeepMind that plays Go, first to beat a professional human player on a full-sized board.
Ambient Intelligence
Electronic environments that are sensitive and responsive to the presence of people.
Analysis of Algorithms
The study of algorithm performance in terms of time and space complexity.
Analytics
The discovery, interpretation, and communication of meaningful patterns in data.
Annotation
The process of adding metadata to a dataset, essential for training AI/ML models.
Annotation as a Service (AaaS)
Outsourcing the data labeling process to specialized service providers, leveraging their expertise and resources.
Annotation Bias
Systematic errors introduced during the labeling process, affecting the fairness and performance of models.
Annotation Efficiency
Measures and techniques to maximize the output of the annotation process with minimal input, crucial for scaling AI projects.
Annotation Guidelines
A set of rules and standards for how data should be labeled, ensuring consistency and accuracy across annotators.
Annotation Project Management
The oversight of the entire annotation workflow, from task allocation to progress tracking to quality control.
Annotation Scalability
The ability to efficiently expand the data labeling process to accommodate growing datasets, a critical aspect for evolving AI projects.
Annotation Workflows
The sequence of steps and processes involved in annotating data, from initial setup to final review and approval.
Answer Set Programming
A declarative programming approach for solving NP-hard search problems using stable model semantics.
Anytime Algorithm
An algorithm that can provide a valid solution even if interrupted before completion.
Application Programming Interface (API)
A set of protocols and tools for building and integrating application software.
Approximate String Matching
Finding strings that closely match a pattern, rather than exactly.
Approximation Error
The difference between an exact value and its approximation.
Argumentation Framework
A structure for dealing with and reasoning about conflicting information.
Artificial General Intelligence (AGI)
AI with human-like cognitive abilities across a wide range of domains and tasks.
Artificial Immune System
Algorithms inspired by the human immune system's mechanisms for adaptive learning and memory.
Artificial Intelligence (AI)
Machine-based systems that emulate human cognitive functions such as learning, problem-solving, and decision-making.
Artificial Intelligence Markup Language (AIML)
An XML dialect for developing natural language conversational agents.
Artificial Neural Network (ANN)
Computing systems inspired by the biological neural networks that constitute animal brains.
Association for the Advancement of Artificial Intelligence (AAAI)
A society promoting research, education, and responsible use of artificial intelligence.
Asymptotic Computational Complexity
The behavior of an algorithm's resource usage as the input size approaches infinity, often expressed in big O notation.
Attention Mechanism
A technique that allows models to focus on relevant parts of the input dynamically.
Attention Mechanisms
Components of neural networks that weigh the importance of different inputs, crucial for tasks requiring focus on specific parts of the data.
Attributional Calculus
A logic system blending predicate logic, propositional calculus, and multi-valued logic for natural induction.
Augmented Reality (AR)
Enhancing real-world environments with computer-generated perceptual information across various sensory modalities.
Automata Theory
The study of abstract machines and the computational problems they can solve.
Automated Annotation
Using algorithms to automatically label data, often as a preliminary step before refinement through human annotation.
Automated Machine Learning (AutoML)
Automating the process of applying machine learning to real-world problems.
Automated Planning and Scheduling
AI techniques for devising action sequences and schedules for execution by intelligent systems.
Automated Reasoning
Using computers to emulate human reasoning processes for solving problems or proving statements.
AutoML (Automated Machine Learning)
Tools and methodologies that automate the process of applying machine learning to real-world problems.
Autonomic Computing
Computing systems capable of self-management and adaptation to changes, reducing complexity for users and operators.
Autonomous Car
A vehicle capable of sensing its environment and operating without human input.
Autonomous Robot
Robots that perform tasks with high autonomy, often integrating AI for decision-making and navigation.
B
Backpropagation
A method for updating neural network weights by propagating errors backward from output to input.
Backpropagation Through Time (BPTT)
An extension of backpropagation for training recurrent neural networks on sequence data.
Backward Chaining
An inference method that starts from the goal and works backward to deduce the required facts.
Bag-of-Words Model
A text representation model that disregards order and grammar, focusing on word frequency.
Bag-of-Words Model in Computer Vision
Treating image features as "words" for classification, using vectors of feature occurrence counts.
Batch Normalization
A method to normalize neural network inputs, improving stability and performance by adjusting and scaling activations.
Bayesian Programming
Specifying and solving problems using probabilistic models under conditions of uncertainty.
Bees Algorithm
An optimization algorithm inspired by the foraging behavior of honey bees.
Behavior Informatics
The study and analysis of behaviors through informatics to derive insights and intelligence.
Behavior Tree
A hierarchical model for structuring decision-making and task execution in AI, robotics, and game development.
Belief-Desire-Intention (BDI) Software Model
A model for programming intelligent agents based on their beliefs, desires, and intentions.
BERT (Bidirectional Encoder Representations from Transformers)
A technique for natural language processing pre-training, demonstrating the importance of contextual word meanings.
Bias
Systematic errors in data or models that can lead to unfair outcomes.
Bias Mitigation
Strategies and techniques to reduce or eliminate bias in data and models, ensuring fairness and equity in AI applications.
Bias–Variance Tradeoff
A fundamental tradeoff in machine learning between model simplicity (bias) and responsiveness to data (variance).
Big Data
Extremely large data sets that challenge traditional data processing and analysis methods.
Big O Notation
Mathematical notation expressing the upper bound of an algorithm's runtime or space complexity as input size grows.
Binary Tree
A tree data structure where each node has at most two children, known as the left and right child.
Blackboard System
An AI approach using a shared knowledge base for collaborative problem-solving by specialist components.
Boltzmann Machine
A stochastic recurrent neural network that serves as a generative model for learning probability distributions.
Boolean Satisfiability Problem (SAT)
Determining if a Boolean formula can be satisfied by some assignment of truth values to its variables.
Brain Technology
Technology leveraging neuroscience insights for self-learning systems like robots and knowledge management.
Branching Factor
The average number of child nodes in a tree or graph, indicating the breadth of each level.
Brute-Force Search
Systematically enumerating all possible solutions to find one that satisfies a problem's criteria.
C
Capsule Neural Network
An ANN variant designed to model hierarchical relationships more effectively, inspired by biological neural structures.
Case-Based Reasoning (CBR)
Solving new problems by adapting solutions from similar past cases.
Chatbot
A computer program that simulates human conversation through text or voice interactions.
Cloud Robotics
Integrating robotics with cloud computing and storage to enhance capabilities and efficiency.
Cluster Analysis
Grouping objects based on their similarities, widely used in data mining, machine learning, and statistical analysis.
Cobweb
An incremental algorithm for hierarchical conceptual clustering, organizing data into a probabilistic classification tree.
Cognitive Architecture
Structured design of the mind in natural or artificial systems, enabling intelligent behavior in complex environments.
Cognitive Computing
Computing that mimics human brain function to enhance decision-making processes.
Cognitive Science
Interdisciplinary study of mind and intelligence, encompassing psychology, AI, linguistics, philosophy, neuroscience, and anthropology.
Collaborative Annotation
Enabling multiple annotators to work together on labeling tasks, improving efficiency and consistency.
Combinatorial Optimization
Finding an optimal solution from a finite set of options in operations research and computer science.
Committee Machine
An artificial neural network model that combines responses from multiple networks to improve overall performance.
Commonsense Knowledge
Everyday world facts and understanding that humans are expected to know, crucial for AI contextual reasoning.
Commonsense Reasoning
AI's simulation of human-like assumptions and deductions about everyday situations and their inherent properties.
Computational Chemistry
Using computer simulations to solve chemical problems and predict chemical properties and reactions.
Computational Complexity Theory
The study of the inherent difficulty of computational problems and their classification based on resource requirements.
Computational Creativity
The study and building of AI systems that can perform tasks traditionally considered creative.
Computational Cybernetics
Fusion of cybernetics with computational intelligence to study and design self-regulating systems.
Computational Humor
AI and computational linguistics field focused on creating and understanding humor through computational methods.
Computational Intelligence
A computer's capability to learn tasks from data or empirical observations, often without explicit programming.
Computational Learning Theory
A subfield of AI focusing on the theoretical underpinnings of machine learning algorithm design and analysis.
Computational Linguistics
The study of language processing and analysis through computational methods and models.
Computational Mathematics
Mathematical research and application in scientific domains leveraging essential computational techniques.
Computational Neuroscience
Uses mathematical models and theories to understand brain functions and cognitive processes.
Computational Number Theory
The study and development of algorithms for solving number-theoretic problems.
Computational Problem
A question or task that a computer system can potentially solve or execute.
Computational Statistics
The intersection of statistics and computer science, focusing on data analysis through computational methods.
Computer Audition
The field focused on enabling machines to understand and interpret audio information.
Computer Science
The study of algorithms, data structures, and the principles of designing and using computers.
Computer Vision
Enabling computers to interpret and understand visual information from the world, akin to human vision.
Computer-Automated Design (CAutoD)
Advanced CAD integrating automation and machine learning for a broader range of design and engineering applications.
Concept Drift
The change in statistical properties of a target variable over time, affecting predictive model accuracy.
Confusion Matrix
A table used to describe the performance of a classification model, essential for understanding how a model's predictions compare to the true labels.
Connectionism
Explains mental phenomena through models based on artificial neural networks in cognitive science.
Consistent Heuristic
A heuristic in AI path-finding that never overestimates the cost to reach the goal from neighboring vertices.
Constrained Conditional Model
A machine learning framework combining conditional models with declarative constraints for enhanced learning and inference.
Constraint Logic Programming (CLP)
Extends logic programming with constraint satisfaction, allowing for declarative problem-solving.
Constraint Programming
Programming where solutions meet specified constraints without defining the steps to find the solution.
Constructed Language (Conlang)
A deliberately created language, unlike natural languages that evolve organically over time.
Control Theory
Mathematical study of controlling dynamical systems in engineering with stability, without delay or overshoot.
Convolutional Neural Network (CNN)
A deep learning model primarily used for visual imagery analysis, featuring minimal preprocessing.
Convolutional Neural Networks (CNNs)
A class of deep neural networks, most commonly applied to analyzing visual imagery, crucial for tasks like image classification and object detection.
Cross-Validation
A method for assessing how the results of a statistical analysis will generalize to an independent data set, vital for evaluating model performance.
Crossover
A genetic algorithm operator that mixes the genetic information of two parents to produce new offspring.
Crowd Labeling
Utilizing a large, often distributed group of human workers to label datasets, leveraging platforms like Amazon Mechanical Turk.
Crowdsourced Annotation Quality Control
Methods for ensuring the reliability and accuracy of data labels obtained from crowdsourced platforms, including redundancy, consensus mechanisms, and expert validation.
Crowdsourcing
A method for gathering annotations from a large group of people, often used to collect diverse labels for data.
Curriculum Learning
A training strategy that starts with easier examples and progressively increases the difficulty level, mimicking human learning processes.
D
Darkforest
A deep learning-based computer Go program developed by Facebook, enhanced by Monte Carlo tree search.
Dartmouth Workshop
The seminal 1956 event regarded as the birth of artificial intelligence as a distinct field.
Data Annotation Tools
Software specifically designed to facilitate the process of labeling data, including image annotation tools, text annotation platforms, etc.
Data Augmentation
Techniques to increase data quantity for machine learning, aiding in reducing overfitting.
Data Augmentation Strategies
Methods for artificially increasing the diversity of training data through transformations or synthetic data generation, enhancing model robustness.
Data Cleansing
The process of detecting and correcting (or removing) corrupt or inaccurate records from a dataset, ensuring high-quality data for training.
Data Curation
The activity of organizing, cleaning, enhancing, and otherwise preparing data for use in specific contexts, including the selection of data for labeling.
Data Fusion
Integrating multiple data sources to produce more accurate and useful information.
Data Imbalance
The occurrence when some classes have significantly more samples than others, a critical consideration in training balanced and fair models.
Data Integration
Combining data from different sources to provide a unified view.
Data Labeling Platforms
Comprehensive tools that facilitate the entire process of data annotation, management, and quality control.
Data Mining
Discovering patterns in large datasets using machine learning, statistics, and database systems.
Data Pipeline
The sequence of processes through which data is transformed and moved, from collection to storage to analysis, including steps for annotation and preprocessing.
Data Preprocessing
The techniques used to prepare data for further analysis or model training, including cleaning and transformation processes.
Data Privacy
Considerations and practices for protecting sensitive information in datasets, especially when annotating personal data.
Data Science
Interdisciplinary field using scientific methods to extract insights from structured and unstructured data.
Data Set
A collection of data, often structured in a table with rows for records and columns for variables.
Data Versioning
Keeping track of different versions of datasets and models, allowing for reproducibility and rollback to previous states.
Data Warehouse
Central repository for integrated data from multiple sources, used for reporting and analysis.
Datalog
A declarative logic programming language subset of Prolog, used for deductive databases and queries.
Decision Boundary
The threshold that separates classes in classification problems, defined by a model's parameters.
Decision Support System
An information system aiding organizational decision-making across management and operational levels.
Decision Theory
The study of an agent's choices, considering uncertainty and values.
Decision Tree Learning
A method using a tree-like model for decision making and predictions.
Declarative Programming
Programming that specifies what to do, not how to do it.
Deductive Classifier
An AI system that makes inferences based on predefined rules and domain knowledge.
Deep Blue
The process of using domain knowledge to create features that make machine learning algorithms work effectively.
Deep Learning
A subset of machine learning using neural networks with many layers to model complex patterns in data.
DeepMind Technologies
An AI company known for breakthroughs in deep learning and reinforcement learning applications.
Default Logic
A framework in AI for reasoning with assumptions that are typically true but not guaranteed.
Description Logic
Formal languages for structured knowledge representation with decidable reasoning tasks.
Developmental Robotics
Robotics focusing on lifelong, open-ended learning through interaction with the environment.
Diagnosis
Identifying system failures and faults through algorithmic analysis of behavior and observations.
Dialogue System
Computer systems designed to interact with humans using natural, coherent conversation across various communication modes.
Differential Privacy
Techniques that add randomness to data or models to prevent the disclosure of individual data points.
Diffusion Model
Generative models that learn data distribution through a process of adding and reversing noise.
Dimensionality Reduction
Process of reducing dataset complexity by decreasing the number of variables, preserving essential information.
Discrete System
A system characterized by distinct, separate states and countable variables, unlike continuous systems.
Distributed Artificial Intelligence (DAI)
AI research focused on developing distributed, collaborative solutions using multiple autonomous agents.
Domain Adaptation
Adjusting models trained on one domain to perform well on a different but related domain, crucial for applying models to new areas.
Dynamic Annotation
Adjusting the annotation process in response to model performance feedback, optimizing the utility of labeled data for model improvement.
Dynamic Epistemic Logic (DEL)
Logical framework for modeling and analyzing how knowledge changes due to events in multi-agent systems.
E
Eager Learning
Learning method constructing a generalized model from training data before receiving new queries.
Ebert Test
A test assessing if a synthesized voice can deliver a joke convincingly enough to make humans laugh.
Echo State Network (ESN)
A recurrent neural network featuring a fixed, sparse hidden layer for efficient temporal pattern learning.
Embeddings
Dense representations of words or features as vectors, essential for capturing the semantic properties of data in machine learning.
Embodied Agent
An AI system interacting with its environment through a physical or virtual embodiment.
Embodied Cognitive Science
Interdisciplinary study of mind-body integration to explain intelligent behavior through holistic models and principles.
Ensemble Averaging
Combining multiple models to improve predictive performance in machine learning.
Ensemble Learning
The practice of combining multiple models to improve prediction accuracy, demonstrating the value of diverse data interpretations.
Entity Recognition
Identifying and classifying key elements in text into predefined categories, a specific task in NLP requiring fine-grained annotations.
Epoch (Machine Learning)
One complete pass of the full training dataset through the machine learning model.
Error-Driven Learning
Learning method minimizing error feedback through actions, a form of reinforcement learning.
Ethical Considerations in Annotation
Addressing moral and ethical issues related to data labeling, including privacy, consent, and the treatment of annotators.
Ethics of Artificial Intelligence
Moral principles guiding the development and application of AI technologies.
Evolutionary Algorithm
Optimization algorithms inspired by biological evolution, using selection, mutation, and recombination.
Evolutionary Computation
Algorithms inspired by biological evolution for solving global optimization problems in AI.
Evolving Classification Function
Dynamic classifiers in AI that adapt to new data in changing environments for classification tasks.
Existential Risk
Risk where AI advances could potentially lead to human extinction or global catastrophe.
Expert Annotation
Data labeling performed by individuals with specialized knowledge in the relevant domain, ensuring high-quality annotations.
Expert System
AI that mimics human expert decision-making using knowledge-based reasoning, typically with if-then rules.
Explainable AI (XAI)
Methods and techniques in artificial intelligence that make the results of the solution understandable by humans, vital for transparency and trust in AI applications.
F
F1 Score
A measure that combines precision and recall into a single metric, providing a balanced view of model performance.
Fast-and-Frugal Trees
Simple decision trees for rapid, efficient classification and decision-making with minimal information.
Feature Extraction
Process of deriving informative, non-redundant features from raw data to improve machine learning model performance.
Feature Learning
Automated discovery of data representations for improved feature detection or classification in AI/ML models.
Feature Selection
Process of choosing a subset of relevant features for effective model construction in AI/ML.
Federated Learning
Collaborative machine learning without centralizing data, enhancing privacy and reducing data transfer needs.
Few-shot Learning
Training models to understand new concepts with very few examples, critical for scenarios with limited data availability.
First-order Logic
A formal system using quantified variables over objects, distinguishing expressions with relations and quantifiers.
Fluent
A temporal condition or property in AI that can change over time, often represented in logical systems.
Formal Language
A structured set of strings generated from an alphabet based on predefined syntactic rules.
Forward Chaining
A reasoning method that starts with known data, applying inference rules to deduce new information until a goal is reached.
Frame
A data structure for organizing knowledge as stereotyped situations in AI systems.
Frame Language
A language for organizing knowledge using frames, emphasizing explicit representation over encapsulation.
Frame Problem
The challenge of updating beliefs to reflect only relevant changes after an action, without extensive reevaluation.
Friendly Artificial Intelligence (FAI)
AI designed to positively impact humanity, ensuring ethical behavior and safety in its interactions and decisions.
Futures Studies
An interdisciplinary field exploring possible, probable, and preferable futures and their underlying assumptions.
Fuzzy Control System
A system utilizing fuzzy logic to analyze continuous input values for nuanced, human-like reasoning in control processes.
Fuzzy Logic
A logic system handling degrees of truth between completely true and false, unlike binary Boolean logic.
Fuzzy Rule
A conditional statement in fuzzy logic that maps fuzzy inputs to a fuzzy output based on degrees of truth.
Fuzzy Set
A set where elements have degrees of membership, represented by values between 0 and 1.
G
Game Theory
Mathematical analysis of strategies in interactions among rational decision-makers.
General Game Playing
AI capability to understand and play multiple games without game-specific programming.
Generative Adversarial Network (GAN)
A machine learning framework where two neural networks, a generator and a discriminator, compete in a zero-sum game.
Generative Adversarial Networks (GANs)
An AI architecture for generating new data samples, using a system of competing neural networks.
Generative Artificial Intelligence
AI that creates new content by learning from data patterns, often using Transformer-based neural networks.
Generative Pretrained Transformer (GPT)
A transformer-based model that generates human-like text by predicting subsequent tokens after extensive pretraining.
Genetic Algorithm
An optimization technique inspired by natural selection, using mutation, crossover, and selection to evolve solutions.
Genetic Operator
Operators in genetic algorithms that manipulate candidate solutions to evolve towards optimal solutions.
Glowworm Swarm Optimization
An optimization algorithm inspired by the social behavior of glowworms for solving multimodal optimization problems.
Graph (Abstract Data Type)
A data structure representing networks of nodes interconnected by edges, used to model relationships.
Graph (Discrete Mathematics)
A mathematical structure consisting of vertices connected by edges, representing pairwise relationships.
Graph Database
A database using graph structures with nodes and edges for semantic queries, prioritizing data relationships.
Graph Theory
A mathematical field studying graphs to model relationships between interconnected objects or entities.
Graph Traversal
The systematic process of visiting, checking, or updating each vertex in a graph, often to discover paths or connections.
Ground Truth
Accurate, real-world data serving as a benchmark for training and evaluating AI models.
Grounding
Connecting abstract concepts to real-world instances, crucial for contextual understanding in AI.
H
Halting Problem
The challenge of determining if a given computer program will eventually halt or run indefinitely.
Heuristic
A problem-solving approach using shortcuts for speed over accuracy, often in search and optimization tasks.
Hidden Layer
Intermediate layers of neurons in neural networks, between input and output, for feature transformation.
Hidden Unit
A neuron within the hidden layer(s) of an artificial neural network, not directly connected to input/output.
Human-in-the-loop (HITL)
A framework integrating human expertise into AI workflows to enhance decision-making and model accuracy.
Hyper-heuristic
An approach automating heuristic selection and combination to solve diverse computational problems efficiently.
Hyperparameter Tuning
Optimizing machine learning model settings to improve performance and accuracy.
I
IEEE Computational Intelligence Society
IEEE society dedicated to computational paradigms inspired by biology and linguistics, such as neural networks and fuzzy systems.
Incremental Learning
Learning method where the model continuously updates its knowledge as new data becomes available.
Inference Engine
Applies logical rules to a knowledge base to deduce new information in AI systems.
Information Integration
Combining data from diverse sources into a cohesive, unified view, often involving disparate data types.
Information Processing Language (IPL)
Early programming language designed for advanced problem-solving, including list processing and dynamic memory allocation.
Instance Segmentation
Identifying and delineating each distinct object within an image, providing precise boundaries and labels.
Intelligence Amplification
Enhancing human intelligence through the strategic use of information technology.
Intelligence Explosion
Rapid enhancement of artificial intelligence to superintelligence through recursive self-improvement, potentially leading to singularity.
Intelligent Agent
An autonomous entity that acts towards goals in an environment using observation, actuators, and possibly learning.
Intelligent Control
Control techniques that employ AI methods such as neural networks and fuzzy logic for advanced system regulation.
Intelligent Personal Assistant
Software that performs tasks for users through verbal commands, often with speech interpretation and response capabilities.
Inter-annotator Agreement
The consistency of data labeling across different annotators, vital for reliable training data quality.
Inter-annotator Reliability
The consistency of annotations across different raters, crucial for the credibility of labeled datasets.
Interpretation
Assigning meaning to symbols in a formal language, crucial for understanding semantics in AI/ML.
Intrinsic Motivation
Motivation driven by the information content or inherent interest of the task, rather than external rewards.
IoU (Intersection over Union)
A metric assessing object detector accuracy by quantifying overlap between predicted and ground truth bounding boxes.
Issue Tree
A graphical representation dissecting a question into components to analyze root causes and potential solutions.
J
K
Kernel Method
Definition: Algorithms in machine learning for pattern analysis, notably including support vector machines (SVMs).
KL-ONE
A semantic network-based knowledge representation system focusing on structured inheritance networks.
Knowledge Acquisition
Process of defining rules and ontologies for knowledge-based systems, often involving domain experts.
Knowledge Engineering
The practice of building, maintaining, and using knowledge-based systems by integrating technical, scientific, and social aspects.
Knowledge Extraction
Creating knowledge from various data sources, aiming for machine-readable formats that support inferencing.
Knowledge Interchange Format (KIF)
A language for sharing and re-using knowledge across diverse knowledge-based systems.
Knowledge Representation and Reasoning
AI field focused on how to represent and reason about world knowledge to solve complex tasks.
Knowledge-based System
A system that uses explicit knowledge and reasoning to solve complex problems.
L
Label Noise
Inaccuracies in dataset annotations that can adversely affect machine learning model training and accuracy.
Label Propagation
Spreading known labels to unlabeled data in a dataset, leveraging data similarities in semi-supervised learning.
Label Skew
A mismatch between training data label distribution and real-world occurrence, affecting model generalization.
Labeling
Assigning identifiers to data points to denote their characteristics or categories, essential for supervised learning.
Labeling Cost Optimization
Reducing expenses associated with data annotation without compromising data quality for efficient model training.
Language Model
A statistical tool that predicts the next word in a sequence based on previous words.
Large Language Model
A highly parameterized language model, often based on the transformer architecture, requiring extensive data and computing for training.
Lazy Learning
A learning method where data generalization is delayed until a query is made.
Lisp (programming language)
A programming language family known for its fully parenthesized prefix notation and history in AI.
Logic Programming
Programming paradigm based on formal logic, expressing solutions through facts and rules within a problem domain.
Long Short-Term Memory (LSTM)
A recurrent neural network architecture designed for processing sequences and retaining long-term dependencies.
M
Machine Learning
Study of algorithms that improve automatically through experience, enabling systems to learn from data.
Machine Learning (ML)
AI algorithms that learn from data to make predictions or decisions, evolving without explicit programming.
Machine Listening
The study of algorithms for audio understanding and processing by machines.
Machine Perception
A computer system's ability to interpret data akin to human sensory understanding.
Machine Teaching
Guiding AI model learning through curated datasets and structured labeling to impart specific knowledge or behaviors.
Machine Vision
Imaging-based automatic inspection and analysis for applications like inspection, process control, and robot guidance.
Markov Chain
A stochastic model where each event's probability depends solely on the state achieved in the previous event.
Markov Decision Process (MDP)
A mathematical model for decision-making in situations with randomness and partial control by a decision-maker.
Mathematical Optimization
Selection of the best element from alternatives based on a criterion in mathematics, computer science, and operations research.
Mechanism Design
Engineering approach in economics and game theory to create mechanisms or incentives for desired outcomes in strategic settings.
Mechatronics
An interdisciplinary engineering field combining mechanical, electrical, and computer science for designing integrated systems.
Metabolic Network Reconstruction and Simulation
Correlates an organism's genome with its molecular physiology through in-depth models.
Metaheuristic
A high-level strategy guiding the search for optimal solutions in complex optimization problems with limited information.
Model Checking
Automated verification of a system model against specified correctness properties or safety requirements.
Model Fine-tuning
Adjusting a pre-trained model's parameters to optimize performance for a specific dataset or task.
Model-agnostic Annotation Techniques
Data labeling methods independent of machine learning model specifics, enhancing versatility in model training.
Modus Ponens
A fundamental rule of inference in propositional logic: "If P implies Q and P is true, then Q is true."
Modus Tollens
A logical rule: "If P implies Q and Q is false, then P must be false."
Monte Carlo Tree Search (MCTS)
A heuristic search algorithm used in decision processes, relying on random sampling to evaluate potential moves.
Multi-Agent System
A system composed of multiple interacting intelligent agents, capable of solving complex or distributed problems.
Multi-Swarm Optimization
An optimization technique using multiple sub-swarms in particle swarm optimization to focus on various solution regions.
Multi-task Learning
Simultaneously training a single model on multiple related tasks to leverage shared knowledge and improve performance.
Multimodal Data
Datasets incorporating various data types like text, images, and audio, enriching analysis and model training.
Mutation
A genetic operator in genetic algorithms that introduces diversity by randomly altering gene values in chromosomes.
Mycin
An early expert system for diagnosing infections and recommending antibiotics, using backward chaining in AI.
N
Naive Bayes Classifier
A simple probabilistic classifier based on applying Bayes' theorem with strong independence assumptions among features.
Naive Semantics
A basic approach to representing domain-specific knowledge, often in natural language processing within artificial intelligence.
Name Binding
The process of associating identifiers with specific entities like variables, functions, or objects in programming.
Named Graph
A Semantic Web concept for identifying RDF statement sets with URIs, enabling metadata descriptions like context and provenance.
Named-Entity Recognition (NER)
AI technique for identifying and classifying key information elements in text into predefined categories.
Natural Language Generation (NLG)
The AI process of converting structured data into human-readable text.
Natural Language Processing (NLP)
AI technology enabling computers to understand, interpret, and generate human language.
Network Motif
Recurrent, statistically significant sub-graphs within larger networks, indicative of underlying system processes or structures.
Neural Machine Translation (NMT)
Machine translation via deep learning, predicting word sequences with a unified neural network model.
Neural Network
Computing systems inspired by biological neural networks, modeling complex patterns through interconnected artificial neurons.
Neural Turing Machine (NTM)
An AI model blending neural networks with external memory, enabling algorithmic tasks through differentiable memory operations.
Neuro-fuzzy
Integrative systems combining neural networks' learning capabilities with fuzzy logic's reasoning for enhanced decision-making.
Neurocybernetics
Interdisciplinary study linking neuroscience and cybernetics to develop systems interfacing brains with computational devices.
Neuromorphic Engineering
Designing computer systems to mimic neuro-biological architectures for advanced computing capabilities.
Node
Fundamental unit in data structures and networks, holding data and possibly connecting to other nodes.
Nondeterministic Algorithm
An algorithm that can produce different outcomes for the same input on different executions.
Normalization
Scaling data features to a standard range or distribution to improve algorithm performance and stability.
Nouvelle AI
AI approach focusing on emergent intelligence from simple behaviors akin to biological organisms, contrasting classical AI.
NP (Nondeterministic Polynomial Time)
Complexity class of decision problems verifiable in polynomial time, not necessarily solvable within it.
NP-completeness
Complexity class of problems solvable and verifiable in polynomial time by nondeterministic algorithms, hard as any in NP.
NP-hardness
Classification of problems as hard as the most challenging in NP, not necessarily in NP, lacking efficient solutions.
O
Object Detection
Identifying and locating objects within images, often involving bounding boxes and class labels.
Occam's Razor
Principle favoring simpler solutions among competing hypotheses making the same predictions, minimizing assumptions.
Offline Learning
Machine learning where models are trained once with a fixed dataset, not updated with new data post-training.
One-hot Encoding
Converting categorical variables into binary vectors for machine learning model compatibility.
Online Machine Learning
Learning method where models update incrementally with each new data point, adapting to new patterns over time.
Ontology
A structured representation of knowledge as a set of concepts within a domain, and the relationships between those concepts.
Ontology Learning
Automated or semi-automated process of creating ontologies by extracting domain-specific terms and their interrelations from text.
Ontology-based Annotation
Leveraging a structured knowledge framework to guide data labeling, enhancing consistency and semantic richness.
Open Mind Common Sense (OMCS)
AI project at MIT Media Lab to build a large commonsense knowledge base from widespread web contributions.
Open-source Software
Software with publicly accessible source code, allowing modification, study, and distribution by anyone.
OpenAI
AI research lab focusing on developing friendly AI to benefit humanity, operating as a for-profit under a non-profit parent.
OpenCog
Open-source project aimed at creating a framework for embodied cognition and human-equivalent AGI.
Out-of-Distribution Detection
Identifying data points that significantly differ from the training data distribution, critical for model reliability.
Overfitting and Underfitting
Overfitting is when a model learns noise from the training data, impairing its generalization. Underfitting is when a model is too simple to capture the underlying data pattern.
P
Partial Order Reduction
Technique reducing state-space in model checking by exploiting commutativity of concurrent transitions.
Partially Observable Markov Decision Process (POMDP)
A framework modeling decision-making under uncertainty with limited observability of the system state.
Particle Swarm Optimization (PSO)
Algorithm optimizing problems by simulating social behavior of particles moving towards optimal solutions.
Pathfinding
Computing the shortest route between two points, heavily based on algorithms like Dijkstra's.
Pattern Recognition
Automatic identification of regularities in data to classify or take actions using computer algorithms.
Pose Estimation
Determining the configuration of objects or body parts in spatial dimensions from image or video data.
Precision and Recall
Metrics evaluating a model's predictions: Precision measures exactness, while Recall assesses completeness.
Predicate Logic
A formal system utilizing quantified variables over objects, enabling complex statements and relations in logic.
Predictive Analytics
Techniques analyzing current/historical data to predict future events, using data mining, predictive modeling, and ML.
Principal Component Analysis (PCA)
A statistical technique transforming correlated variables into a set of linearly uncorrelated variables called principal components.
Principle of Rationality
The assumption that agents act in the most adequate way according to their understanding of a situation.
Probabilistic Programming
Programming paradigm for specifying probabilistic models and performing automatic inference.
Production System
A system in AI comprising rules and mechanisms for decision-making based on world states.
Programming Language
Formal language for creating instructions that computers execute to perform specific tasks.
Prolog
A declarative programming language for AI and computational linguistics, based on first-order logic.
Propositional Calculus
A branch of logic focusing on propositions and their connections through logical connectives.
Python
High-level programming language known for code readability and support for multiple programming paradigms.
Q
Qualification Problem
The challenge of enumerating all preconditions for an action's intended effect in AI systems.
Quality Assurance in Annotation
Ensuring the accuracy and consistency of data labeling through systematic review and validation processes.
Quantifier
Logic symbols specifying the quantity of specimens satisfying a formula, commonly "for all" and "there exists".
Quantum Computing
Computing based on quantum-mechanical phenomena like superposition and entanglement for enhanced computational power.
Query Language
Computer languages used to make queries in databases and information systems for data retrieval.
R
R Programming Language
Language and environment for statistical computing and graphics, widely used in data analysis and research.
Radial Basis Function Network
A neural network using radial basis functions as activation functions for tasks like approximation and classification.
Random Forest
Ensemble method using multiple decision trees for classification, regression, and other tasks, improving over single trees.
Reasoning System
Software that deduces conclusions from knowledge using logical methods like deduction and induction in AI.
Recurrent Neural Network (RNN)
Neural networks with memory, processing sequences by leveraging their internal state for dynamic temporal behavior.
Recurrent Neural Networks (RNNs)
AI models ideal for sequential data analysis, capturing temporal information through internal memory.
Region Connection Calculus (RCC)
A framework for qualitative spatial reasoning, describing regions by their interrelations.
Regularization
Techniques to prevent overfitting by penalizing model complexity, enhancing generalization to unseen data.
Reinforcement Learning (RL)
ML paradigm where agents learn to make decisions by optimizing cumulative rewards through trial and error.
Reinforcement Learning from Human Feedback (RLHF)
Training AI with human feedback to shape behavior, enhancing qualities like truthfulness or safety in generated content.
Reservoir Computing
A computation framework using a fixed dynamical system for input processing and a trainable output layer.
Resource Description Framework (RDF)
A W3C standard for modeling and sharing web resources metadata using various serialization formats.
Restricted Boltzmann Machine (RBM)
A generative neural network for learning input distributions through a layer of visible and hidden units.
Rete Algorithm
An efficient pattern matching algorithm for rule-based systems to apply rules to facts in a knowledge base.
Robotics
The interdisciplinary field focusing on the design, construction, and operation of robots using various engineering disciplines.
Rule-Based System
A system that uses predefined, human-crafted rules to interpret and process information for decision-making in AI.
S
Satisfiability
The property of a logical formula being true under at least one interpretation within mathematical logic.
Scalable Annotation
Data labeling methods designed to efficiently expand with increasing data volumes, crucial for big data projects.
Search Algorithm
Algorithms designed to retrieve information from data structures or explore problem domains for solutions.
Selection
In genetic algorithms, the process of choosing genomes from a population for breeding based on fitness.
Selective Linear Definite clause resolution (SLD resolution)
An inference rule in logic programming, refining resolution for sound, complete reasoning with Horn clauses.
Self-Management
Autonomous operation of computer systems without human intervention, crucial in AI and ML applications.
Self-Supervised Learning
Learning from data without explicit labels, using inherent data structures to generate supervisory signals.
Semantic Annotation
Enriching data with context-specific metadata that clarifies its meaning, facilitating deeper AI understanding.
Semantic Network
A graphical representation of concepts and their interrelations used for knowledge representation in AI.
Semantic Query
Queries leveraging context and associations in data to retrieve both explicit and inferred information.
Semantic Reasoner
Software that infers logical consequences from facts or axioms, often using ontology and description logic.
Semantic Segmentation
Assigning a class label to each pixel in an image, delineating objects by their semantic meaning.
Semantics
The study of meaning in programming languages, focusing on the relationship between syntax and program behavior.
Semi-supervised Learning
Combining labeled and unlabeled data to train models, enhancing learning efficiency and data usage.
Sensor Fusion
Integrating data from multiple sensors to reduce uncertainty and improve information accuracy in AI/ML systems.
Separation Logic
An extension of Hoare logic for reasoning about programs, particularly for mutable data structures.
Sequence-to-Sequence Models
AI models transforming input sequences into output sequences, crucial for tasks like translation or text summarization.
Similarity Learning
Machine learning area focused on learning a function to assess similarity or relatedness between objects.
Simulated Annealing
A probabilistic technique for approximating the global optimum in large optimization problem spaces.
Situated Approach
Designing AI agents to effectively interact within their specific environments, emphasizing perceptual and motor skills.
Situation Calculus
A formalism in logic for representing and reasoning about dynamic systems and their changes over time.
Software
Data or instructions that direct a computer's operation, distinct from the physical hardware.
Software Engineering
The systematic application of engineering principles to the design, development, maintenance, testing, and evaluation of software.
SPARQL
A query language for databases to retrieve and manipulate data in RDF format, used in semantic web technologies.
Spatial-Temporal Reasoning
AI domain focused on understanding and predicting spatial and temporal dynamics within data.
Speech Recognition
AI technology that converts spoken language into text.
Spiking Neural Network (SNN)
A type of neural network that simulates time-dependent neural events, mimicking natural brain dynamics.
Stanford Research Institute Problem Solver (STRIPS)
An early form of automated planner for AI, pivotal in developing languages for expressing automated planning problems.
State
A representation of the status of a system or environment at a specific point in time, capturing relevant variables.
Statistical Classification
Machine learning method for assigning categories to data points based on their features and training data.
Statistical Relational Learning (SRL)
AI field combining relational data models with statistical methods to handle uncertainty and complex structures.
Stochastic Optimization
Optimization techniques using randomness to solve problems with uncertain or variable elements.
Stochastic Semantic Analysis
A method in NLP that employs probabilistic models to understand the meaning of word segments in language.
Subject-Matter Expert (SME)
An individual with deep knowledge and expertise in a specific field, demonstrated through education, certification, or experience.
Superintelligence
Hypothetical AI that vastly exceeds the cognitive performance of humans in all domains of interest.
Supervised Learning
Machine learning task where a model is trained on labeled data to predict outputs from inputs.
Swarm Intelligence
Collective intelligence emerging from simple agents coordinating and interacting within decentralized, self-organized systems.
Symbolic Artificial Intelligence
AI approach using human-readable symbols to represent problems, logic, and search for problem-solving.
Synthetic Annotation
Creating data labels automatically through simulations or generative models, enhancing datasets without manual labeling.
Synthetic Data Generation
Creating artificial data mimicking real data's statistical properties for training models or ensuring privacy.
Synthetic Intelligence
Intelligence in machines designed to be genuine and autonomous, not merely an imitation of human intelligence.
Systems Neuroscience
Neuroscience subfield studying neural circuits' structure, function, and their integration into brain systems.
T
Technological Singularity
Hypothetical future point where technological growth becomes uncontrollable, leading to transformative changes to civilization.
Temporal Difference Learning
Reinforcement learning methods that update value estimates based on the difference between subsequent predictions.
Tensor Network Theory
Mathematical framework modeling the transformation of sensory data into motor responses by neuronal networks using tensors.
TensorFlow
Open-source library for numerical computation and large-scale machine learning, using data flow graphs.
Theoretical Computer Science
Branch of computer science that deals with the mathematical aspects of computing and computational theory.
Theory of Computation
Branch of computer science studying the limits and capabilities of computing machines and algorithms.
Thompson Sampling
Probabilistic method for balancing exploration and exploitation in multi-armed bandit problems by sampling from belief distributions.
Time Complexity
Measure of the amount of computational time an algorithm takes to complete as a function of input size.
Tokenization
Breaking down text into smaller units (tokens) for analysis in NLP, crucial for data preprocessing.
Transfer Annotation
Reusing existing dataset annotations in new, but similar, contexts to minimize additional labeling work.
Transfer Learning
Adapting a pre-trained model for a new, but related, task to reduce the need for extensive new data.
Transformer
Deep learning model using self-attention to process sequential data, significant in natural language processing and beyond.
Transhumanism
Movement advocating for enhancing human capabilities through advanced technologies.
Transition System
Conceptual model describing discrete systems' states and transitions, used in computational theory.
Tree Traversal
Process of visiting each node in a tree data structure in a specific order, exactly once.
True Quantified Boolean Formula (TQBF)
A formal language of fully quantified Boolean formulas where every variable is bound and the formula is true.
Turing Machine
Abstract computational model capable of simulating any algorithm via tape manipulation based on predefined rules.
Turing Test
Test assessing a machine's ability to exhibit human-like intelligence in conversations, devised by Alan Turing.
Type System
Framework within programming languages for assigning types to program constructs, enhancing reliability and performance.
U
V
W
Watson
IBM's AI system for answering questions in natural language, developed under the DeepQA project.
Weak AI
AI designed to perform specific, narrow tasks without possessing consciousness, understanding, or self-awareness.
Weak Supervision
Utilizing imprecise or noisy labels to train models, reducing reliance on extensively labeled datasets.
World Wide Web Consortium (W3C)
International organization that develops open standards to ensure the long-term growth of the Web.