Abstract

This talk consists of two shorter talks on specific probabilistic computing systems:

1. BayesDB, a Bayesian database table, lets users query many of the probable implications of their tabular data as naturally as SQL lets them query the data itself. With the Bayesian Query Language (BQL), a domain-specific probabilistic programming language for data tables, users with no statistics training can solve basic data science problems, such as detecting predictive relationships between variables, inferring missing values, simulating probable observations, and identifying statistically similar database entries. BayesDB is based on a nonparametric Bayesian machine learning technique for directly estimating the full multivariate (joint) distribution underlying high-dimensional, heterogeneously typed data. I will illustrate BayesDB and give an overview of current applications to datasets from econometrics and sociology. I will also discuss the potential for using BayesDB to improve the quality of the empirical reasoning performed by non-experts and begin to mitigate the shortage of analysts with expertise in statistics.

2. The brain interprets ambiguous sensory information faster and more reliably than modern computers, using neurons that are slower and less reliable than logic gates. But Bayesian inference, which underpins many computational models of perception and cognition, appears computationally challenging even given modern transistor speeds and energy budgets. I will show how to build fast Bayesian computing machines using intentionally stochastic, digital parts, narrowing this efficiency gap by multiple orders of magnitude.