Amortized Probabilistic Conditioning for Optimization, Simulation and Inference

TL;DR

The paper introduces ACE, a transformer-based meta-learning model that explicitly models latent variables, enabling flexible conditioning and prediction in probabilistic tasks across various domains.

Contribution

It proposes ACE, a novel model that allows explicit representation and conditioning on latent variables, enhancing flexibility in probabilistic meta-learning tasks.

Findings

ACE demonstrates improved performance in image completion and classification.

ACE effectively supports Bayesian optimization and simulation-based inference.

The model offers flexible conditioning on observed data and priors at runtime.

Abstract

Amortized meta-learning methods based on pre-training have propelled fields like natural language processing and vision. Transformer-based neural processes and their variants are leading models for probabilistic meta-learning with a tractable objective. Often trained on synthetic data, these models implicitly capture essential latent information in the data-generation process. However, existing methods do not allow users to flexibly inject (condition on) and extract (predict) this probabilistic latent information at runtime, which is key to many tasks. We introduce the Amortized Conditioning Engine (ACE), a new transformer-based meta-learning model that explicitly represents latent variables of interest. ACE affords conditioning on both observed data and interpretable latent variables, the inclusion of priors at runtime, and outputs predictive distributions for discrete and continuous data and latents. We show ACE's modeling flexibility and performance in diverse tasks such as image completion and classification, Bayesian optimization, and simulation-based inference.