Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

A Model or 603 Exemplars: Towards Memory-Efficient Class-Incremental Learning

About

Real-world applications require the classification model to adapt to new classes without forgetting old ones. Correspondingly, Class-Incremental Learning (CIL) aims to train a model with limited memory size to meet this requirement. Typical CIL methods tend to save representative exemplars from former classes to resist forgetting, while recent works find that storing models from history can substantially boost the performance. However, the stored models are not counted into the memory budget, which implicitly results in unfair comparisons. We find that when counting the model size into the total budget and comparing methods with aligned memory size, saving models do not consistently work, especially for the case with limited memory budgets. As a result, we need to holistically evaluate different CIL methods at different memory scales and simultaneously consider accuracy and memory size for measurement. On the other hand, we dive deeply into the construction of the memory buffer for memory efficiency. By analyzing the effect of different layers in the network, we find that shallow and deep layers have different characteristics in CIL. Motivated by this, we propose a simple yet effective baseline, denoted as MEMO for Memory-efficient Expandable MOdel. MEMO extends specialized layers based on the shared generalized representations, efficiently extracting diverse representations with modest cost and maintaining representative exemplars. Extensive experiments on benchmark datasets validate MEMO's competitive performance. Code is available at: https://github.com/wangkiw/ICLR23-MEMO

Da-Wei Zhou, Qi-Wei Wang, Han-Jia Ye, De-Chuan Zhan• 2022

Related benchmarks

TaskDatasetResultRank
Class-incremental learningCIFAR-100
Averaged Incremental Accuracy75.83
281
Class-incremental learningCIFAR-100
Average Accuracy87.1
150
Class-incremental learningImageNet-R
Last Accuracy65.7
147
Class-incremental learningCIFAR100 (test)
Avg Acc68.99
116
Class-incremental learningImageNet A
Average Accuracy39.2
110
Class-incremental learningCIFAR-100 10 (test)
Average Top-1 Accuracy61.68
105
Class-incremental learningImageNet-R B0 Inc20
Last Accuracy66.62
98
Class-incremental learningImageNet-100
Avg Acc81.11
82
Class-incremental learningCIFAR100 10 Tasks
Accuracy84.08
66
Class-incremental learningCUB200
Last Accuracy82.7
64
Showing 10 of 60 rows

Other info

Follow for update