A Nonparametric Framework for Quantifying Generative Inference on Neuromorphic Systems

TL;DR

This paper introduces a nonparametric method to evaluate and optimize generative inference performance on neuromorphic hardware, addressing the lack of suitable metrics for high-dimensional probabilistic models.

Contribution

It applies nonparametric goodness-of-fit testing to quantify generative performance and guide parameter selection for neuromorphic Gibbs sampling.

Findings

Effective quantification of generative performance on neuromorphic systems.

Guidelines for optimizing hardware resource usage during sampling.

Enhanced decision-making for parameter tuning in neuromorphic generative models.

Abstract

Restricted Boltzmann Machines and Deep Belief Networks have been successfully used in probabilistic generative model applications such as image occlusion removal, pattern completion and motion synthesis. Generative inference in such algorithms can be performed very efficiently on hardware using a Markov Chain Monte Carlo procedure called Gibbs sampling, where stochastic samples are drawn from noisy integrate and fire neurons implemented on neuromorphic substrates. Currently, no satisfactory metrics exist for evaluating the generative performance of such algorithms implemented on high-dimensional data for neuromorphic platforms. This paper demonstrates the application of nonparametric goodness-of-fit testing to both quantify the generative performance as well as provide decision-directed criteria for choosing the parameters of the neuromorphic Gibbs sampler and optimizing usage of hardware resources used during sampling.