Integrating Prior Knowledge in Contrastive Learning with Kernel
This work addresses representation quality in contrastive learning for domains like natural and medical images, offering an incremental improvement by incorporating prior knowledge.
The paper tackles the problem of improving unsupervised contrastive learning by integrating prior knowledge from generative models or weak attributes into positive and negative sampling, resulting in a novel loss that outperforms other approaches in weakly supervised scenarios.
Data augmentation is a crucial component in unsupervised contrastive learning (CL). It determines how positive samples are defined and, ultimately, the quality of the learned representation. In this work, we open the door to new perspectives for CL by integrating prior knowledge, given either by generative models -- viewed as prior representations -- or weak attributes in the positive and negative sampling. To this end, we use kernel theory to propose a novel loss, called decoupled uniformity, that i) allows the integration of prior knowledge and ii) removes the negative-positive coupling in the original InfoNCE loss. We draw a connection between contrastive learning and conditional mean embedding theory to derive tight bounds on the downstream classification loss. In an unsupervised setting, we empirically demonstrate that CL benefits from generative models to improve its representation both on natural and medical images. In a weakly supervised scenario, our framework outperforms other unconditional and conditional CL approaches.