Exploring loss function topology with cyclical learning rates
This work provides insights into neural network behavior for researchers, but it is incremental as it builds on existing methods like cyclical learning rates.
The authors investigated the loss function topology of residual networks using cyclical learning rates and linear network interpolation, discovering phenomena like counterintuitive loss changes and rapid training, and demonstrated that cyclical learning rates can achieve higher testing accuracy than traditional training with large learning rates.
We present observations and discussion of previously unreported phenomena discovered while training residual networks. The goal of this work is to better understand the nature of neural networks through the examination of these new empirical results. These behaviors were identified through the application of Cyclical Learning Rates (CLR) and linear network interpolation. Among these behaviors are counterintuitive increases and decreases in training loss and instances of rapid training. For example, we demonstrate how CLR can produce greater testing accuracy than traditional training despite using large learning rates. Files to replicate these results are available at https://github.com/lnsmith54/exploring-loss