Learning Structured Latent Factors from Dependent Data:A Generative Model Framework from Information-Theoretic Perspective
This work addresses the fundamental challenge of learning controllable and generalizable representations with structural properties for machine learning practitioners, offering a flexible framework that can automatically estimate dependency structures from data.
The paper tackles the problem of learning structured latent factors from dependent data by introducing a generative model framework that enforces desired structural properties through mask variables and extends the multivariate information bottleneck theory. The result is a principled approach that unifies existing generative models and is applicable to tasks like multi-modal data modeling, algorithmic fairness, and invariant risk minimization.
Learning controllable and generalizable representation of multivariate data with desired structural properties remains a fundamental problem in machine learning. In this paper, we present a novel framework for learning generative models with various underlying structures in the latent space. We represent the inductive bias in the form of mask variables to model the dependency structure in the graphical model and extend the theory of multivariate information bottleneck to enforce it. Our model provides a principled approach to learn a set of semantically meaningful latent factors that reflect various types of desired structures like capturing correlation or encoding invariance, while also offering the flexibility to automatically estimate the dependency structure from data. We show that our framework unifies many existing generative models and can be applied to a variety of tasks including multi-modal data modeling, algorithmic fairness, and invariant risk minimization.