Linear Models of Computation and Program Learning

TL;DR

This paper explores linear computation models like probabilistic sampling and generalized animation, arguing they facilitate easier program learning compared to traditional deterministic approaches, and discusses recent advances and theoretical connections.

Contribution

It introduces the perspective that linear models of computation can improve program learning and connects probabilistic programming with vector semantics and non-monotonic inference.

Findings

Linear models enable more tractable program learning.

Recent advances in higher-order probabilistic programming support this approach.

Connections between inconsistency, inference, and vector semantics are discussed.

Abstract

We consider two classes of computations which admit taking linear combinations of execution runs: probabilistic sampling and generalized animation. We argue that the task of program learning should be more tractable for these architectures than for conventional deterministic programs. We look at the recent advances in the "sampling the samplers" paradigm in higher-order probabilistic programming. We also discuss connections between partial inconsistency, non-monotonic inference, and vector semantics.