RAMP: Recognition parametrisation by Amortised Message Passing

2026-07-21

AI Summary

This work addresses the challenge of efficiently inferring latent variable models in unsupervised learning when dealing with high-dimensional, complex data—scenarios where existing approaches often suffer from computational intractability or restrictive distributional assumptions. To overcome these limitations, the authors propose RAMP, a novel method that implicitly captures latent structure through a learnable, nonlinear amortized message-passing mechanism. By seamlessly integrating recognition networks with message passing, RAMP transcends the constraints of traditional explicit graphical models and approximate inference schemes. Within a likelihood-driven learning framework, RAMP achieves scalable posterior inference without sacrificing expressive power, enabling accurate and efficient recovery of latent distributions in high-dimensional settings. This approach substantially enhances both model flexibility and computational efficiency.

Abstract

A central aim of unsupervised learning is to uncover latent factors that explain dependencies among observations. Probabilistic models typically achieve this by introducing multiple latent variables linked through a graph of conditional relationships, with distributional parameters and their dependence learnt from data. Learning relies either on distributional choices that allow tractable belief propagation, or on approximations that scale poorly with model size and complexity. We build on the recently developed recognition-parametrised modelling paradigm to propose an alternative approach: RAMP, a method that implicitly defines latent structure by learning a flexible, nonlinear, amortised message-passing framework. We show that RAMP enables efficient likelihood-based recovery of latent-variable distributions within expressive nonlinear models acting on complex high-dimensional data.

Problem

• unsupervised learning
• latent factors
• probabilistic models
• message passing
• scalability

Innovation

• amortised message passing
• recognition parametrisation
• latent variable models
• nonlinear probabilistic modeling
• unsupervised learning