日本語索引
読み方から用語を探せます。
あ
- アクタークリティック (Wikipedia)
- Actor–critic
A method combining an actor that updates the policy with a critic that evaluates the actor using value estimates. - アルゴリズム (Wikipedia)
- Algorithm
A defined procedure used to solve a problem or perform a computation. - アルゴリズム影響評価 (Wikipedia)
- Algorithmic impact assessment
A structured assessment of an automated system's expected effects, risks, and safeguards. - アルゴリズム偏り (Wikipedia)
- Algorithmic bias
Systematic and unfair differences in algorithmic outcomes affecting people or groups. - 異常検知 (Wikipedia)
- Anomaly detection
Identifying unusual patterns or outliers in data. - 一般化 (Wikipedia)
- Generalization
A model’s ability to perform well on data it did not see during training. - 意味的類似度 (Wikipedia)
- Semantic similarity
A measure of how closely two items match in meaning rather than surface form. - 医用画像 (Wikipedia)
- Medical imaging
Techniques for creating and interpreting images of the body for clinical purposes. - 医療分野のAI (Wikipedia)
- AI in healthcare
The application of AI to clinical care, medical operations, research, and public health. - 埋め込み (Wikipedia)
- Embedding
A learned numerical representation that places items in a vector space. - エピソード (Wikipedia)
- Episode
One complete sequence of interaction from an initial state until a terminal state. - AIエージェント (Wikipedia)
- AI agent
An AI system that observes context, chooses actions, and works toward a goal. - AIガバナンス (Wikipedia)
- AI governance
The structures, responsibilities, and processes used to direct and oversee AI systems. - AI規制と政策 (Wikipedia)
- AI regulation and policy
Rules and public policies governing the development, deployment, and use of AI. - AIの安全性と評価 (Wikipedia)
- AI safety and evaluation
The practice of measuring AI capabilities, limitations, and risks and reducing the likelihood or severity of harmful outcomes. - AI倫理 (Wikipedia)
- AI ethics
The study and practice of addressing moral issues arising from AI design, deployment, and use. - エージェント (Wikipedia)
- Agent
The decision-making entity that observes the environment, chooses actions, and receives rewards.
か
- 回帰 (Wikipedia)
- Regression
A supervised-learning task that estimates a continuous numerical target. - 解釈可能性 (Wikipedia)
- Interpretability
The degree to which a person can understand how a model produces an output. - 確率分布 (Wikipedia)
- Probability distribution
A mathematical description of the probabilities assigned to possible outcomes. - 確率変数 (Wikipedia)
- Random variable
A variable whose possible numerical values are outcomes of a random process. - 確率モデル (Wikipedia)
- Probabilistic model
A model that represents variables and outcomes using probability distributions. - Hidden layer
A neural-network layer located between the input and output layers. - 価値関数 (Wikipedia)
- Value function
A function that estimates the expected return from a state, or from taking an action in a state, under a policy. - 価値ベース手法 (Wikipedia)
- Value-based method
A method that learns value estimates and derives action choices from them. - 活性化関数 (Wikipedia)
- Activation function
A function that transforms a neuron's combined input and introduces nonlinear behavior. - 活用 (Wikipedia)
- Exploitation
Choosing actions currently believed to produce the highest return. - 下流タスク (Wikipedia)
- Downstream task
A specific application for which a pretrained model is adapted or evaluated. - 環境 (Wikipedia)
- Environment
Everything outside the agent that responds to its actions and supplies states and rewards. - 画像セグメンテーション (Wikipedia)
- Image segmentation
The task of assigning image pixels to meaningful regions or classes. - 画像分類 (Wikipedia)
- Image classification
Assigning a label or category to an image. - 頑健性 (Wikipedia)
- Robustness
The ability of a system to maintain acceptable behavior under variation, noise, or attack. - 機械学習 (Wikipedia)
- Machine learning
A field of AI in which systems learn patterns or behavior from data and experience. - 基盤モデル (Wikipedia)
- Foundation model
A model trained broadly at scale that can be adapted to many downstream tasks. - Q学習 (Wikipedia)
- Q-learning
A model-free, value-based algorithm that learns the expected return for taking each action in each state. - 強化学習 (Wikipedia)
- Reinforcement learning
A machine-learning approach in which an agent learns behavior through interaction and feedback from rewards. - 教師あり学習 (Wikipedia)
- Supervised learning
A machine-learning approach in which a model is trained on labeled examples. - 教師なし学習 (Wikipedia)
- Unsupervised learning
A machine-learning approach in which a model learns patterns from unlabeled data. - 金融分野のAI (Wikipedia)
- AI in finance
The application of AI to financial analysis, decisions, services, and risk management. - クラスタリング (Wikipedia)
- Clustering
Grouping similar data points according to their features. - クラスラベル (Wikipedia)
- Class label
The category name or identifier predicted in a classification task. - 訓練 (Wikipedia)
- Training
The process of adjusting a model’s parameters using data. - 訓練データ (Wikipedia)
- Training data
Examples used to fit a model's parameters or learned behavior. - 訓練・検証・テストデータ (Wikipedia)
- Training, validation, and test data
Separate data subsets used to fit a model, tune choices, and estimate final performance. - グラウンディング (Wikipedia)
- Grounding
Connecting a model's output to supplied evidence, data, or observable context. - グラフ (Wikipedia)
- Graph
A structure consisting of entities represented as vertices and relationships represented as edges. - グラフニューラルネットワーク (Wikipedia)
- Graph neural network
A neural network designed to learn from nodes, edges, and relationships in graph-structured data. - 計画 (Wikipedia)
- Planning
Selecting and ordering actions intended to move from a current state toward a goal. - 検索拡張生成 (Wikipedia)
- Retrieval-augmented generation
A generation method that retrieves external information and supplies it to a generative model as context. - 検証データ (Wikipedia)
- Validation data
Data used during development to compare settings or select models without fitting their parameters directly. - 言語モデル (Wikipedia)
- Language model
A model that assigns probabilities to sequences of language elements or predicts language content. - 交差検証 (Wikipedia)
- Cross-validation
A resampling method that evaluates a model across multiple training and validation partitions. - 行動 (Wikipedia)
- Action
A choice made by the agent that can affect the environment and future rewards. - 勾配降下法 (Wikipedia)
- Gradient descent
An iterative optimization method that moves parameters in the direction that reduces an objective. - 公平性 (Wikipedia)
- Fairness
The study and management of unjustified differences in how AI systems affect people or groups. - 固有表現抽出 (Wikipedia)
- Named-entity recognition
The task of locating and categorizing names such as people, organizations, and places in text. - コンテキストウィンドウ (Wikipedia)
- Context window
The amount of input and generated content a model can consider in one interaction. - コンピュータビジョン (Wikipedia)
- Computer vision
The field that enables computers to extract information from images and video. - 誤差逆伝播法 (Wikipedia)
- Backpropagation
An algorithm that computes gradients by propagating output error backward through a neural network.
さ
- 最適化アルゴリズム (Wikipedia)
- Optimization algorithm
A procedure for finding parameter values that improve or minimize an objective. - サポートセット (Wikipedia)
- Support set
The small labeled example set supplied for adaptation in a few-shot learning task. - サンプリング (Wikipedia)
- Sampling
The process of drawing examples from a probability distribution or generative model. - 視覚言語モデル (Wikipedia)
- Vision-language model
A model trained to represent and reason across visual and linguistic information. - 市場セグメンテーション (Wikipedia)
- Market segmentation
Dividing a market into groups with shared characteristics or behavior. - 自然言語処理 (Wikipedia)
- Natural language processing
The field concerned with computational processing and generation of human language. - 収益 (Wikipedia)
- Return
The total reward accumulated from a time step onward, usually with future rewards discounted. - 終端状態 (Wikipedia)
- Terminal state
A state that ends an episode and after which no further action is taken in that episode. - 出力ラベル (Wikipedia)
- Output label
The expected category or value associated with an input. - 少数ショット学習 (Wikipedia)
- Few-shot learning
Learning or adapting from only a small number of labeled examples. - 深層強化学習 (Wikipedia)
- Deep reinforcement learning
Reinforcement learning that uses deep neural networks to represent policies, value functions, or environment models. - 信用スコアリング (Wikipedia)
- Credit scoring
Estimating the likelihood that a borrower will repay debt. - 時間ステップ (Wikipedia)
- Time step
One discrete point in an interaction sequence at which the agent observes, acts, and receives feedback. - 次元削減 (Wikipedia)
- Dimensionality reduction
Reducing the number of features while retaining essential structure or information. - 自己位置推定 (Wikipedia)
- Localization
Estimating the position and orientation of a robot or autonomous system. - 自己教師あり学習 (Wikipedia)
- Self-supervised learning
Learning in which supervisory signals are derived from the structure of the input data itself. - 自己注意機構 (Wikipedia)
- Self-attention
Attention in which elements of one sequence attend to other elements in that same sequence. - 事後確率 (Wikipedia)
- Posterior probability
A probability updated after combining prior belief with observed evidence. - 事前確率 (Wikipedia)
- Prior probability
A probability expressing belief before incorporating the current evidence. - 事前学習 (Wikipedia)
- Pretraining
Initial training on broad data before a model is adapted for a more specific purpose. - 事前学習済みモデル (Wikipedia)
- Pretrained model
A model whose parameters were learned before adaptation to the current task. - 自動化バイアス (Wikipedia)
- Automation bias
The tendency to favor suggestions from automated systems even when contrary evidence exists. - 状態 (Wikipedia)
- State
A representation of the information needed to describe the environment at a particular time. - 状態遷移 (Wikipedia)
- State transition
The change from one state to another after the agent takes an action. - 情報検索 (Wikipedia)
- Information retrieval
Finding relevant documents or records in response to an information need. - 自律システム (Wikipedia)
- Autonomous system
A system that can operate and make decisions with limited direct human control. - 人工知能 (Wikipedia)
- Artificial intelligence
The field concerned with building machines that perform tasks associated with human intelligence. - 人工ニューロン (Wikipedia)
- Artificial neuron
A mathematical unit that combines inputs and applies an activation function to produce an output. - 推論 (Wikipedia)
- Inference
Using a trained model to produce an output from new input. - スパムメール検出 (Wikipedia)
- Spam email detection
Classifying email as spam or not spam. - 生成モデル (Wikipedia)
- Generative model
A model that learns a data distribution and can produce new samples resembling its training data. - 説明可能なAI (Wikipedia)
- Explainable AI
Methods and practices intended to make AI behavior understandable to people. - 説明責任 (Wikipedia)
- Accountability
Responsibility for decisions, impacts, oversight, and remedies associated with an AI system. - センサフュージョン (Wikipedia)
- Sensor fusion
Combining measurements from multiple sensors to obtain a more reliable estimate. - 潜在変数 (Wikipedia)
- Latent variable
An unobserved variable used by a model to explain patterns in observed data. - 損失関数 (Wikipedia)
- Loss function
A formula that measures the difference between predicted and actual outputs.
た
- 対照学習 (Wikipedia)
- Contrastive learning
Learning representations by bringing related examples closer and separating unrelated examples. - 対象タスク (Wikipedia)
- Target task
The task to which transferred knowledge or a pretrained model is being applied. - 多クラス分類 (Wikipedia)
- Multiclass classification
Classification in which an example is selected from more than two possible classes. - 探索 (Wikipedia)
- Exploration
Trying actions whose outcomes are uncertain in order to discover potentially better behavior. - 大規模言語モデル (Wikipedia)
- Large language model
A language model with many parameters trained on large text collections for broad language capabilities. - 代理タスク (Wikipedia)
- Pretext task
A task constructed from unlabeled data to provide a learning signal for representation learning. - 知的エージェント (Wikipedia)
- Intelligent agent
A system that perceives an environment and acts to pursue objectives. - 注意機構 (Wikipedia)
- Attention
A mechanism that assigns different weights to input elements when computing a representation. - ツール使用 (Wikipedia)
- Tool use
The ability of an AI system to invoke external functions, software, or services to complete work. - 敵対的機械学習 (Wikipedia)
- Adversarial machine learning
The study of attacks on machine-learning systems and defenses against them. - 敵対的サンプル (Wikipedia)
- Adversarial example
An input deliberately modified to cause a model to make an incorrect prediction. - テストデータ (Wikipedia)
- Test data
Held-out data used to estimate a finished model's performance on unseen examples. - 転移学習 (Wikipedia)
- Transfer learning
Reusing knowledge learned for one task or domain to improve learning on another. - ディープラーニング (Wikipedia)
- Deep learning
Machine learning based on neural networks with multiple learned representation layers. - データ (Wikipedia)
- Data
Recorded observations, measurements, symbols, or examples used for analysis and learning. - データ可視化 (Wikipedia)
- Data visualization
Representing data graphically to reveal patterns, trends, and insights. - データセット (Wikipedia)
- Dataset
An organized collection of related data used for analysis or model development. - データとデータセット (Wikipedia)
- Data and datasets
The examples and organized collections used for analysis and machine learning. - データドリフト (Wikipedia)
- Data drift
A change over time in the statistical properties of data received by a deployed system. - データ分割 (Wikipedia)
- Data split
The process of partitioning a dataset into distinct subsets for model development and evaluation. - データポイズニング (Wikipedia)
- Data poisoning
An attack that corrupts training data to influence a model's learned behavior. - データリーク (Wikipedia)
- Data leakage
The unintended use of information during training that would not be available when the model is deployed. - 特徴選択 (Wikipedia)
- Feature selection
Selecting a useful subset of available input features. - 特徴抽出 (Wikipedia)
- Feature extraction
Transforming raw data into informative derived features. - 特徴量 (Wikipedia)
- Feature
A measurable input property supplied to a machine-learning model. - 特徴量エンジニアリング (Wikipedia)
- Feature engineering
The process of creating or transforming model inputs to improve learning and performance. - 特徴量重要度 (Wikipedia)
- Feature importance
A measure of how strongly an input feature influences a model's predictions. - トランスフォーマー (Wikipedia)
- Transformer
A neural-network architecture that processes relationships between tokens primarily through attention mechanisms. - トークン (Wikipedia)
- Token
A unit of text or other input that a model processes as one element of a sequence. - トークン化 (Wikipedia)
- Tokenization
The process of dividing input into units that a language-processing system can handle. - 動作計画 (Wikipedia)
- Motion planning
Computing a feasible sequence of movements from a starting configuration to a goal.
な
- 二値分類 (Wikipedia)
- Binary classification
Classification in which each example is assigned to one of two classes. - 入力データ (Wikipedia)
- Input data
The raw information or features supplied to a model. - ニューラルネットワーク (Wikipedia)
- Neural network
A model composed of connected processing units arranged in layers. - 人間とAIの相互作用 (Wikipedia)
- Human–AI interaction
The study and design of communication, cooperation, and control between people and AI systems.
は
- ハルシネーション (Wikipedia)
- Hallucination
Generated content that is unsupported, false, or inconsistent with the supplied evidence. - ヒューマン・イン・ザ・ループ (Wikipedia)
- Human in the loop
A design in which people provide oversight, decisions, labels, or feedback within an automated process. - 評価指標 (Wikipedia)
- Evaluation metric
A numerical measure used to assess a model's predictions or behavior. - 表現学習 (Wikipedia)
- Representation learning
Learning useful features or representations directly from data. - ファインチューニング (Wikipedia)
- Fine-tuning
Further training a pretrained model on task-specific data or objectives. - 不正検知 (Wikipedia)
- Fraud detection
Identifying transactions or behavior that may involve deception or unauthorized activity. - 物体検出 (Wikipedia)
- Object detection
The task of locating and classifying objects within an image or video. - 分類 (Wikipedia)
- Classification
The task of assigning an input to one or more discrete categories. - ベイズ学習 (Wikipedia)
- Bayesian learning
Learning that applies Bayes' theorem to update uncertainty about models or parameters. - ベクトル (Wikipedia)
- Vector
An ordered collection of numbers used to represent position, direction, or learned features. - ベクトルデータベース (Wikipedia)
- Vector database
A database designed to store embeddings and retrieve nearby vectors efficiently. - ベンチマーク (Wikipedia)
- Benchmark
A standardized task, dataset, or measurement used to compare systems. - 方策 (Wikipedia)
- Policy
A rule or probability distribution that determines which action an agent selects in each state. - 方策勾配 (Wikipedia)
- Policy gradient
A family of policy-based methods that improves a policy by following the gradient of expected return. - 方策ベース手法 (Wikipedia)
- Policy-based method
A method that optimizes a policy directly instead of deriving it only from learned values. - 報酬 (Wikipedia)
- Reward
A numerical feedback signal that indicates the immediate desirability of an outcome.
ま
- マルコフ決定過程 (Wikipedia)
- Markov Decision Process (MDP)
A mathematical framework for sequential decisions defined by states, actions, transition probabilities, rewards, and a discount factor. - マルチモーダル埋め込み (Wikipedia)
- Multimodal embedding
A representation that places information from different modalities in a shared or aligned vector space. - マルチモーダルAI (Wikipedia)
- Multimodal AI
AI that processes or combines more than one kind of data, such as text, images, audio, or video. - 未見データ (Wikipedia)
- Unseen data
Data that a model did not encounter during training. - メタラーニング (Wikipedia)
- Meta-learning
Learning methods that improve a system's ability to learn new tasks from limited experience. - メッセージパッシング (Wikipedia)
- Message passing
A graph-learning procedure in which nodes update their representations using information from neighboring nodes. - メモリ (Wikipedia)
- Memory
Stored information that an agent can retain and use across steps or interactions. - 目的関数 (Wikipedia)
- Objective function
A numerical function that an optimization process seeks to minimize or maximize. - 目的変数 (Wikipedia)
- Target variable
The outcome that a supervised model is trained to predict. - 目標値 (Wikipedia)
- Target value
The expected output used in supervised learning and absent from unsupervised training data. - モダリティ (Wikipedia)
- Modality
A particular form or channel of information, such as text, vision, or audio. - モデル (Wikipedia)
- Model
A mathematical representation that processes input data to produce predictions, classifications, or decisions. - モデル監視 (Wikipedia)
- Model monitoring
Tracking a deployed model's behavior, performance, inputs, and operational health. - モデル集約 (Wikipedia)
- Model aggregation
The process of combining model updates received from multiple federated clients. - モデル展開 (Wikipedia)
- Model deployment
Making a trained model available in a production system or application. - モデル評価 (Wikipedia)
- Model evaluation
The systematic measurement of how well a model performs for its intended purpose. - モデルフリー強化学習 (Wikipedia)
- Model-free reinforcement learning
Reinforcement learning that learns values or policies directly from experience without learning a model for planning. - モデルベース強化学習 (Wikipedia)
- Model-based reinforcement learning
Reinforcement learning that uses a learned or supplied model of environment dynamics to plan or improve behavior.
や
- 予測 (Wikipedia)
- Prediction
An output produced by a trained model. - 予測誤差 (Wikipedia)
- Prediction error
The difference between a predicted value and its target value.
ら
- ラベル (Wikipedia)
- Label
A known category or target attached to an example in labeled data. - ラベル付きデータセット (Wikipedia)
- Labeled dataset
A collection in which each input is paired with its expected output or target. - 臨床意思決定支援 (Wikipedia)
- Clinical decision support
Software that supplies patient-specific information to assist healthcare decisions. - レッドチーミング (Wikipedia)
- Red teaming
Adversarial testing that searches deliberately for failures, vulnerabilities, or harmful behavior. - 連合学習 (Wikipedia)
- Federated learning
Distributed learning in which participants train locally and share model updates rather than raw data. - 連合クライアント (Wikipedia)
- Federated client
A participating device or organization that performs local training in federated learning. - 連続値 (Wikipedia)
- Continuous value
A numerical value that can vary across an interval rather than belonging to a discrete class. - ロボット知覚 (Wikipedia)
- Robot perception
The process by which a robot interprets sensor data to understand itself and its surroundings. - ロボティクス分野のAI (Wikipedia)
- AI in robotics
The use of AI for robot perception, decision-making, learning, and control.
わ
- 割引率 (Wikipedia)
- Discount factor
A value between zero and one that controls how strongly future rewards affect present decisions.