Skip to content

Learning_Mechanisms.md

Latest commit

 

History

History
47 lines (26 loc) · 2.73 KB

Learning_Mechanisms.md

File metadata and controls

47 lines (26 loc) · 2.73 KB
graph LR
    LearningRule["LearningRule"]
    MCCLearningRule["MCCLearningRule"]
    WeightPostProcessor["WeightPostProcessor"]
    LearningRule -- "is base for" --> MCCLearningRule
    MCCLearningRule -- "extends" --> LearningRule
    LearningRule -- "sends output to" --> WeightPostProcessor
    MCCLearningRule -- "sends output to" --> WeightPostProcessor
Loading

CodeBoardingDemoContact

Details

The Learning Mechanisms subsystem is a critical part of bindsnet, responsible for adapting synaptic weights based on various learning rules. It adheres to the project's architectural bias towards a modular and extensible scientific computing library for SNNs.

LearningRule

This component serves as the abstract foundation for various synaptic weight update rules. It provides a generalized interface and common mechanisms, acting as a dispatcher that directs updates to specific connection types (e.g., 1D, 2D, 3D, convolutional) via internal helper methods. It defines the core update method as a central entry point for weight adjustments.

Related Classes/Methods:

MCCLearningRule

A concrete and specialized implementation of the Mean-Covariance Control (MCC) learning algorithm. It manages its own specific connection updates and contains the unique logic for weight adjustment according to the MCC rule, extending or overriding the base LearningRule's update mechanism.

Related Classes/Methods:

WeightPostProcessor

This component handles operations applied to synaptic weights after the primary learning rule has been executed. This can include normalization, clipping, or other constraints to ensure weight stability or adherence to specific criteria.

Related Classes/Methods:

FAQ