LODAP: On-device incremental learning via lightweight operations and data pruning

Abstract

Incremental learning, which enables a deployed model to continually learn new classes over time, is becoming increasingly crucial for industrial edge systems, where communicating with remote servers for computation-intensive training is often difficult. As edge devices are expected to continue learning more classes after deployment, designing efficient on-device learning frameworks is essential. In this paper, we propose LODAP, a lightweight on-device incremental learning framework for edge systems. The key part of LODAP is a new module, namely Efficient Incremental Module (EIM), which combines normal convolutions with lightweight operations. During incremental learning, EIM employs adapters to effectively and efficiently learn features for new classes to improve the accuracy of incremental learning while reducing model complexity and training overhead. To further improve efficiency, LODAP integrates a data pruning strategy that removes redundant training samples, significantly lowering the training cost. We conducted extensive experiments on the CIFAR-100, Tiny-ImageNet, and CUB-200-2011 datasets. Experimental results show that LODAP improves the accuracy by up to 4.32% over existing methods while reducing around 50% of model complexity. In addition, evaluations on real edge systems demonstrate its applicability for on-device machine learning. The code is available at https://github.com/duanbiqing/LODAP.