Glossary
Clear, reusable definitions for responsible AI learning, governance, and implementation. In Arabic, English, French, and Spanish.
1444 terms
A distributed system that includes multiple software components running on multiple computers as a single system.
Open term1444 terms
Statistical way to compare two or more techniques to determine which performs better and to understand whether the difference is statistically significant.
Open termAI & DataSee "Abductive Reasoning."
Open termAI & DataA framework that combines logic programming with abductive reasoning to generate plausible explanations or hypotheses for observed data or events.
Open termAI & DataThe process of inferring probable conclusions based on known information. Also called "Abduction."
Open termAI & DataA scientific examination used to analyze the performance of a model by systematically removing or disabling some of its components and evaluating the impact of that on the performance.
Open termAI & DataA set of data values and related operations that are independent of any implementation.
Open termAI & DataA state of being accountable for something based on regulation, agreement, or assignment.
Open termAI & DataA rate of correct predictions a classification model makes.
Open termAI & DataA mechanism in reinforcement learning in which the agent transitions between the environment states.
Open termAI & DataA subfield of machine learning where AI agents learn to understand the effects of their actions within an environment.
Open termAI & DataA computer vision task that analyzes videos to identify human actions. Also called "Human Action Recognition" or "Human Activity Recognition (HAR)."
Open termAI & DataA process of deciding what to do next by an agent.
Open termAI & DataThe total set of possible actions an agent can perform within its operating environment.
Open termAI & DataA mathematical formula used in an artificial neural network to process weighted inputs of a node and generate output.
Open termAI & DataSee "Feature Map."
Open termAI & DataA semi-supervised learning approach where the algorithm selects some of the data it needs for learning.
Open termAI & DataMetadata that is automatically collected and continuously updated in real time from systems and events, then actively used to trigger actions, automate processes, and support operational and governance functions.
Open termAI & DataA method to adapt a model to specific tasks by selecting and annotating the most uncertain questions with task-specific prompts and chain-of-thought.
Open termAI & DataAn advanced method for advertising that targets ads at the audiences who are most likely receptive to the offerings. Also called "Targeted Advertising (or Targeted Ads)."
Open termAI & DataAn advanced gradient descent algorithm that performs gradient using knowledge of the data geometry from previous iterations.
Open termAI & DataA fine-tuning technique that adds small trainable layers to a pre-trained model and only updates the parameters of these layers to adapt the model to specific tasks.
Open termAI & DataAn algorithm that changes its behavior at run time based on available information and predefined criteria.
Open termAI & DataA boosting technique that combines multiple weak classifiers to create a strong classifier.
Open termAI & DataA flexible data governance approach focuses on continually adjusting to meet changes in business needs, regulatory requirements, and technologies.
Open termAI & DataA way to update and retrain machine learning models incrementally to adapt to new changes.
Open termAI & DataA technique that combines fuzzy logic and neural networks to enhance the performance and applicability of a prediction.
Open termAI & DataA retrieval mechanism that dynamically adjusts the search strategy based on context, user needs, and knowledge gaps to improve the relevance and quality of the results.
Open termAI & DataA function that never overestimates the cost of reaching the goal in a search algorithm.
Open termAI & DataSee "Driver Assistance."
Open termAI & DataThe inputs that have been deliberately designed to trick a machine learning model into making a mistake.
Open term30 of 1444 terms
