Novel theoretical development in mixture models and latent variable models

Mixture models and latent variable frameworks underpin modern statistical learning, yet significant theoretical challenges remain in understanding identifiability, optimal inference, and scalable computation. The session is timely given the surge of interest in mixture models, network modeling, and high-dimensional latent-variable learning in statistics, machine learning, and data science. It appeals to researchers interested in theoretical guarantees, scalable algorithms, and applications where latent structure plays a central role. By bringing together diverse but complementary perspectives, this session fosters cross-field exchange and encourages further development of rigorous foundations for mixture and latent variable models.