MUTE: Inter-class Ambiguity Driven Multi-hot Target Encoding for Deep Neural Network Design

Abstract

Target encoding is an effective technique to boost performance of classical and deep neural networks based classification models. However, the existing target encoding approaches require significant increase in the learning capacity, thus demand higher computation power and more training data. In this paper, we present a novel and efficient target encoding method, Inter-class Ambiguity Driven Multi-hot Target Encoding (MUTE), to improve both generalizability and robustness of a classification model by understanding the inter-class characteristics of a target dataset. By evaluating ambiguity between the target classes in a dataset, MUTE strategically optimizes the Hamming distances among target encoding. Such optimized target encoding offers higher classification strength for neural network models with negligible computation overhead and without increasing the model size. When MUTE is applied to the popular image classification networks and datasets, our experimental results show that MUTE offers better generalization and defense against the noises and adversarial attacks over the existing solutions.