Advances in Contrastive Learning and Human-Expert Collaboration

The field of machine learning is witnessing significant developments in contrastive learning and human-expert collaboration. Researchers are exploring new frameworks and techniques to improve the performance of contrastive learning methods, including the use of probabilistic variational contrastive learning, projection perspectives, and meta-learning representations. Additionally, there is a growing interest in incorporating human expertise into machine learning models, with studies investigating the use of Bayesian inference for correlated human experts and classifiers. These innovations have the potential to enhance the accuracy and efficiency of machine learning models in various applications, including medical classification and image recognition. Noteworthy papers include: Bayesian Inference for Corated Human Experts and Classifiers, which develops a general Bayesian framework for querying human experts and leveraging class probability estimates of pre-trained classifiers. Probabilistic Variational Contrastive Learning, which proposes a decoder-free framework that maximizes the evidence lower bound by interpreting the InfoNCE loss as a surrogate reconstruction term and adding a KL divergence regularizer to a uniform prior on the unit hypersphere.

Sources

Bayesian Inference for Correlated Human Experts and Classifiers

Any-Class Presence Likelihood for Robust Multi-Label Classification with Abundant Negative Data

Inverting Black-Box Face Recognition Systems via Zero-Order Optimization in Eigenface Space

On the Similarities of Embeddings in Contrastive Learning

A theoretical framework for self-supervised contrastive learning for continuous dependent data

Generalizing Supervised Contrastive learning: A Projection Perspective

Probabilistic Variational Contrastive Learning

Meta-learning Representations for Learning from Multiple Annotators

Built with on top of