General Cyclical Training of Neural Networks

Abstract

This paper describes the principle of "General Cyclical Training" in machine learning, where training starts and ends with "easy training" and the "hard training" happens during the middle epochs. We propose several manifestations for training neural networks, including algorithmic examples (via hyper-parameters and loss functions), data-based examples, and model-based examples. Specifically, we introduce several novel techniques: cyclical weight decay, cyclical batch size, cyclical focal loss, cyclical softmax temperature, cyclical data augmentation, cyclical gradient clipping, and cyclical semi-supervised learning. In addition, we demonstrate that cyclical weight decay, cyclical softmax temperature, and cyclical gradient clipping (as three examples of this principle) are beneficial in the test accuracy performance of a trained model. Furthermore, we discuss model-based examples (such as pretraining and knowledge distillation) from the perspective of general cyclical training and recommend some changes to the typical training methodology. In summary, this paper defines the general cyclical training concept and discusses several specific ways in which this concept can be applied to training neural networks. In the spirit of reproducibility, the code used in our experiments is available at https://github.com/lnsmith54/CFL.

Skeleton-Based Action Recognition Through Contrasting Two-Stream Spatial-Temporal Networks

“…Combining the focus loss with the general cyclical training principle, the cyclical focal loss [52] is proposed. In [53], the general cyclical training of a neural network is considered as a combination of curriculum learning in the early epochs with fine-tuning at the end of training. Specifically, the easy and confident training samples are used at the start and end stages of network training, and the hard training samples are processed at the middle stage of training.…”