End-to-End Incremental Learning

TL;DR

This paper presents an end-to-end incremental learning method for deep neural networks that mitigates catastrophic forgetting by using a distillation loss and a small exemplar set, enabling scalable learning of new classes.

Contribution

It introduces a novel end-to-end framework combining distillation and cross-entropy losses for incremental learning, maintaining data representation and classification jointly.

Findings

Achieves state-of-the-art results on CIFAR-100 and ImageNet datasets.

Effectively reduces catastrophic forgetting in incremental learning.

Uses only a small exemplar set from old classes for training.

Abstract

Although deep learning approaches have stood out in recent years due to their state-of-the-art results, they continue to suffer from catastrophic forgetting, a dramatic decrease in overall performance when training with new classes added incrementally. This is due to current neural network architectures requiring the entire dataset, consisting of all the samples from the old as well as the new classes, to update the model -a requirement that becomes easily unsustainable as the number of classes grows. We address this issue with our approach to learn deep neural networks incrementally, using new data and only a small exemplar set corresponding to samples from the old classes. This is based on a loss composed of a distillation measure to retain the knowledge acquired from the old classes, and a cross-entropy loss to learn the new classes. Our incremental training is achieved while keeping the entire framework end-to-end, i.e., learning the data representation and the classifier jointly, unlike recent methods with no such guarantees. We evaluate our method extensively on the CIFAR-100 and ImageNet (ILSVRC 2012) image classification datasets, and show state-of-the-art performance.