Increasing Depth of Neural Networks for Life-long Learning

TL;DR

This paper introduces a novel continual learning method that increases neural network depth by adding new layers for new tasks, enabling knowledge transfer and avoiding forgetting without requiring a memory buffer.

Contribution

It proposes a dynamic, depth-increasing neural network approach inspired by Progressive Neural Networks, reducing memory usage while maintaining performance in lifelong learning scenarios.

Findings

Performs on par with existing methods on Split CIFAR and Tiny ImageNet.

Outperforms Experience Replay in a single-dataset task setup.

Avoids catastrophic forgetting without a rehearsal buffer.

Abstract

Purpose: We propose a novel method for continual learning based on the increasing depth of neural networks. This work explores whether extending neural network depth may be beneficial in a life-long learning setting. Methods: We propose a novel approach based on adding new layers on top of existing ones to enable the forward transfer of knowledge and adapting previously learned representations. We employ a method of determining the most similar tasks for selecting the best location in our network to add new nodes with trainable parameters. This approach allows for creating a tree-like model, where each node is a set of neural network parameters dedicated to a specific task. The Progressive Neural Network concept inspires the proposed method. Therefore, it benefits from dynamic changes in network structure. However, Progressive Neural Network allocates a lot of memory for the whole network structure during the learning process. The proposed method alleviates this by adding only part of a network for a new task and utilizing a subset of previously trained weights. At the same time, we may retain the benefit of PNN, such as no forgetting guaranteed by design, without needing a memory buffer. Results: Experiments on Split CIFAR and Split Tiny ImageNet show that the proposed algorithm is on par with other continual learning methods. In a more challenging setup with a single computer vision dataset as a separate task, our method outperforms Experience Replay. Conclusion: It is compatible with commonly used computer vision architectures and does not require a custom network structure. As an adaptation to changing data distribution is made by expanding the architecture, there is no need to utilize a rehearsal buffer. For this reason, our method could be used for sensitive applications where data privacy must be considered.