core.distribution

Overview

Provides classes and utilities for representing probability distributions over neural network parameters.

Key Components

AbstractVariable: a base interface for distribution variables (e.g., GaussianVariable).
Utility functions for constructing, copying, and computing KL divergences between distributions.

These distributions form the core representation of uncertainty in a PAC-Bayes model.

View Source

 1"""
 2## Overview
 3Provides classes and utilities for representing probability distributions
 4over neural network parameters.
 5
 6## Key Components
 7- **AbstractVariable**: a base interface for distribution variables
 8  (e.g., GaussianVariable).
 9- **Utility functions** for constructing, copying, and computing KL divergences
10  between distributions.
11
12These distributions form the core representation of uncertainty in a
13PAC-Bayes model.
14"""
15
16from core.distribution.AbstractVariable import AbstractVariable
17from core.distribution.GaussianVariable import GaussianVariable
18from core.distribution.LaplaceVariable import LaplaceVariable