Learn More
Potential-based shaping was designed as a way of introducing background knowledge into model-free reinforcement-learning algorithms. By identifying states that are likely to have high value, this approach can decrease experience complexity—the number of trials needed to find near-optimal behavior. An orthogonal way of decreasing experience complexity is to(More)
Probabilistic inference procedures are usually coded painstakingly from scratch, for each target model and each inference algorithm. We reduce this coding effort by generating inference procedures from models automatically. We make this code generation modular by decomposing inference algorithms into reusable program transformations. These source-to-source(More)
We present Hakaru, a new probabilistic programming system that allows composable reuse of distributions, queries, and inference algorithms, all expressed in a single language of measures. The system implements two automatic and semantics-preserving program transformations—disintegration, which calculates conditional distributions , and simplification, which(More)
—We draw a formal connection between using synthetic training data to optimize neural network parameters and approximate, Bayesian, model-based reasoning. In particular, training a neural network using synthetic data can be viewed as learning a proposal distribution generator for approximate inference in the synthetic-data generative model. We demonstrate(More)
A promising approach to implementing black-box inference is to search a space of inference strategies. However, a search space has not been defined to date that includes combinations of exact and approximate inference methods, such as Rao-Blackwellized Metropolis-Hastings and Gibbs sampling of continuous distributions. We present a handful of inference(More)
In recent years, declarative programming languages specialized for probabilistic modeling has emerged as distinct class of languages. These languages are predominantly written by researchers in the machine learning field and concentrate on generalized MCMC inference algorithm. Unfortunately, all these languages are too slow for practical adoption. In my(More)
  • 1