Uses and Abuses of the Cross-Entropy Loss: Case Studies in Modern Deep Learning

Abstract

Modern deep learning is primarily an experimental science, in which empirical advances occasionally come at the expense of probabilistic rigor. Here we focus on one such example; namely the use of the categorical cross-entropy loss to model data that is not strictly categorical, but rather takes values on the simplex. This practice is standard in neural network architectures with label smoothing and actor-mimic reinforcement learning, amongst others. Drawing on the recently discovered continuous-categorical distribution, we propose probabilistically-inspired alternatives to these models, providing an approach that is more principled and theoretically appealing. Through careful experimentation, including an ablation study, we identify the potential for outperformance in these models, thereby highlighting the importance of a proper probabilistic treatment, as well as illustrating some of the failure modes thereof.

References in corpus (5)

• Distilling the Knowledge in a Neural Network
• L2 Regularization versus Batch and Weight Normalization
• Towards better decoding and language model integration in sequence to sequence models
• Batch Normalization is a Cause of Adversarial Vulnerability
• Rethinking the Usage of Batch Normalization and Dropout in the Training of Deep Neural Networks

Cited by in corpus (2)

• FundusQ-Net: a Regression Quality Assessment Deep Learning Algorithm for Fundus Images Quality Grading
• DSC-IITISM at FinCausal 2021: Combining POS tagging with Attention-based Contextual Representations for Identifying Causal Relationships in Financial Documents