A Nonparametric Framework for Quantifying Generative Inference on Neuromorphic Systems
This addresses a problem for researchers and engineers working on neuromorphic computing by providing a framework to assess and improve generative inference, though it is incremental as it applies existing statistical methods to a specific domain.
The paper tackles the lack of metrics for evaluating generative performance of algorithms like Restricted Boltzmann Machines on neuromorphic systems, and demonstrates that nonparametric goodness-of-fit testing can quantify this performance and optimize hardware resource usage.
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.