Debugging using Orthogonal Gradient Descent

TL;DR

This paper introduces a method for debugging neural networks by unlearning and relearning behaviors using Orthogonal Gradient Descent, avoiding full retraining and improving model correction efficiency.

Contribution

It applies a modified continual learning algorithm, OGD, to neural network debugging, enabling behavior correction without retraining from scratch.

Findings

Successfully unlearns undesirable behaviors in models

Relearns correct behaviors after unlearning

Maintains overall model performance during debugging

Abstract

In this report we consider the following problem: Given a trained model that is partially faulty, can we correct its behaviour without having to train the model from scratch? In other words, can we ``debug" neural networks similar to how we address bugs in our mathematical models and standard computer code. We base our approach on the hypothesis that debugging can be treated as a two-task continual learning problem. In particular, we employ a modified version of a continual learning algorithm called Orthogonal Gradient Descent (OGD) to demonstrate, via two simple experiments on the MNIST dataset, that we can in-fact \textit{unlearn} the undesirable behaviour while retaining the general performance of the model, and we can additionally \textit{relearn} the appropriate behaviour, both without having to train the model from scratch.