Projects Mutual Information Estimation With Contrastive Fenchel-Legendre Optimization Analyzing the limitations of the mutual information (MI) variational bound InfoNCE, with application to deep representation learning Proposing a novel unified framework for contrastive MI estimation, derived simple, powerful algorithm named Fenchel-Legendre Optimization (FLO) Meta-Learning with Information Theoretical Objectives Deriving a mutual information based meta-learning framework based on information theoretic generalization theory Applying our work to the prompt tuning and few-shot learning applications Association-based Optimal Subpopulation Selection for Multivariate Data Identifying a subset of observations for which the variables are strongly associated Proposing a semiparametric statistical approach for the optimal selection of subpopulations based on the patterns of associations in multivariate data Muti-Layer Sliced Designs for Online Experimentation Gauging the platform factor effects (e.g., device platforms), webpage factor effects and their interactions in online experiments for user experience and click-through-rate (CTR) improvement Customized design schemes with the ordered importance of sliced factors