Continuous Histogram Loss: Beyond Neural Similarity
This work addresses the problem of similarity learning for researchers and practitioners by extending it beyond binary similarities, though it appears incremental as it builds on the existing Histogram loss.
The paper tackles the limitation of binary similarity in state-of-the-art similarity learning methods by introducing Continuous Histogram Loss (CHL), which generalizes Histogram loss to handle multiple-valued similarities continuously distributed within a range, enabling it to solve a wider range of tasks such as similarity learning, representation learning, and data visualization.
Similarity learning has gained a lot of attention from researches in recent years and tons of successful approaches have been recently proposed. However, the majority of the state-of-the-art similarity learning methods consider only a binary similarity. In this paper we introduce a new loss function called Continuous Histogram Loss (CHL) which generalizes recently proposed Histogram loss to multiple-valued similarities, i.e. allowing the acceptable values of similarity to be continuously distributed within some range. The novel loss function is computed by aggregating pairwise distances and similarities into 2D histograms in a differentiable manner and then computing the probability of condition that pairwise distances will not decrease as the similarities increase. The novel loss is capable of solving a wider range of tasks including similarity learning, representation learning and data visualization.