Dynamically Scaled Temperature in Self-Supervised Contrastive Learning

In contemporary self-supervised contrastive algorithms such as SimCLR and MoCo, the task of balancing attraction between two semantically similar samples and repulsion between two samples of different classes is primarily affected by the presence of hard negative samples. While the InfoNCE loss has been shown to impose penalties based on hardness, the temperature hyperparameter is the key to regulate the penalties and the tradeoff between uniformity and tolerance. In this work, we focus our attention on improving the performance of InfoNCE loss in self-supervised learning by proposing a novel cosine similarity dependent temperature scaling function to effectively optimize the distribution of the samples in the feature space. We also provide mathematical analyzes to support the construction of such a dynamically scaled temperature function. Experimental evidence shows that the proposed framework outperforms the contrastive loss-based SSL algorithms.