AI-ordlista
Den kompletta ordlistan över AI
Query by Committee
Approche où un comité de modèles vote pour identifier les échantillons avec le plus grand désaccord entre les membres.
Uncertainty Sampling
Strategy selecting samples for which the current model is least certain about its predictions.
Active Learning by Pool
Method where the algorithm selects the most informative samples from a pool of unlabeled data.
Active Learning on Data Streams
Approach where each sample arrives sequentially and the system instantly decides whether to label it or not.
Weighted Density
Strategy combining model uncertainty with data density to avoid outliers and favor representative regions.
Strategies Based on Margins
Selection based on the margin between the most probable classes, favoring samples near the decision boundary.
Active Learning for Deep Learning
Adaptation of active learning strategies specifically optimized for deep neural network architectures.
Adversarial Active Learning
Use of generative adversarial models to create or select samples maximizing classifier uncertainty.
Active Learning with Multiple Annotators
Strategies optimizing sample selection and assignment to annotators based on their expertise and cost.
Budget-aware Active Learning
Approaches integrating budget constraints and variable annotation costs into the selection strategy.
Active Reinforcement Learning
Use of reinforcement learning agents to learn the optimal sample selection policy.
Active Learning for NLP
Specialized strategies for natural language processing, handling the specificities of textual and sequential data.