Learning Module - Abstract Documentation
Purpose and Responsibility
Implements the core adaptive learning algorithms including Bayesian updating, strategy mixture models, hierarchical learning, and transfer learning. Provides the cognitive modeling framework for personalized learning adaptation.
Key Data Structures and Relationships
Core Learning Components
- adaptive: Adaptive scheduling with Expected Information Gain optimization
- bayesian: Bayesian cognitive modeling and uncertainty quantification
- learner: Core learner model with node embeddings and operation proficiencies
- hierarchical_bayes: Hierarchical Bayesian models for population learning
- strategy_mixture: Multiple strategy learning and mixture model fitting
- transfer_learning: Cross-domain knowledge transfer and generalization
- macro_learning: Meta-learning and strategy discovery algorithms
Cognitive Architecture
LearnerModel → NodeEmbeddings + OperationProficiencies + MemoryStrengths
AdaptiveScheduler → BayesianModel + TaskSelection + ModelUpdating
Main Data Flows and Transformations
Learning Pipeline
- Task Selection: EIG-based optimization for maximum learning efficiency
- Response Processing: Bayesian updating of cognitive model parameters
- Strategy Adaptation: Dynamic strategy mixture based on performance patterns
- Transfer Integration: Cross-domain knowledge application and generalization
Model Evolution
- Parameter Updating: Continuous refinement of cognitive model parameters
- Uncertainty Reduction: Information gain-based learning optimization
- Strategy Discovery: Automatic identification of effective learning strategies
- Knowledge Transfer: Application of learned patterns to new domains
Core Algorithms and Business Logic Abstractions
- Bayesian Inference: Posterior updating with evidence integration
- Information Theory: Expected Information Gain calculation for task selection
- Mixture Modeling: Multiple strategy identification and weighting
- Hierarchical Modeling: Individual differences within population structure