Adaptive Variational Continual Learning via Task-Heuristic Modelling

Overview

This repository hosts the project Adaptive Variational Continual Learning via Task-Heuristic Modelling. The project aims to develop a self-balancing Variational Continual Learning (VCL) framework that adapts to new tasks using heuristic modeling.

Table of Contents

• Overview
• Results
• Installation
• Usage
• Notebooks
• License

Results

The root directory contains notebooks designed to reproduce the relevant original VCL results (vcl_reproduce.ipynb) and to facilitate further exploration of a self-balancing Variational Continual Learning framework (autovcl_experiment.ipynb). If the images are not properly rendered in the notebooks, please refer to the two PDF documents located in results/pdf.

Installation

To install the necessary dependencies, run the following command:

pip install -r requirements.txt

Usage

To run the experiments, use the following notebooks:

Notebooks

• vcl_reproduce.ipynb: Reproduces the original Variational Continual Learning (VCL) results.
• autovcl_experiment.ipynb: Explores the self-balancing Variational Continual Learning framework.

License

This project is licensed under the MIT License. See the LICENSE file for further details.