Position: The Future of Bayesian Prediction Is Prior-Fitted

TL;DR

This paper advocates for the future prominence of Prior-data Fitted Networks (PFNs) and amortized inference in Bayesian modeling, emphasizing their efficiency in low-data scenarios and potential for advancing complex applications.

Contribution

It highlights the expansion of PFNs to complex domains, argues for their central role in Bayesian inference, and discusses future research directions and limitations.

Findings

PFNs effectively leverage prior information from artificial datasets.

PFNs enable efficient low-data Bayesian inference.

The paper identifies key challenges and future directions for PFNs.

Abstract

Training neural networks on randomly generated artificial datasets yields Bayesian models that capture the prior defined by the dataset-generating distribution. Prior-data Fitted Networks (PFNs) are a class of methods designed to leverage this insight. In an era of rapidly increasing computational resources for pre-training and a near stagnation in the generation of new real-world data in many applications, PFNs are poised to play a more important role across a wide range of applications. They enable the efficient allocation of pre-training compute to low-data scenarios. Originally applied to small Bayesian modeling tasks, the field of PFNs has significantly expanded to address more complex domains and larger datasets. This position paper argues that PFNs and other amortized inference approaches represent the future of Bayesian inference, leveraging amortized learning to tackle data-scarce problems. We thus believe they are a fruitful area of research. In this position paper, we explore their potential and directions to address their current limitations.