A General Approach to Automatic Programming using Occam's Razor, Compression and Self-inspection

摘要

This paper describes a novel general method for automatic programming which can be seen as a generalization of techniques such as genetic programming and ADATE. The approach builds on the assumption that data compression can be used as a metaphor for cognition and intelligence. The proof-of-concept system is evaluated on sequence prediction problems. Inspired by Solomonoff's induction methods, the general method outlined takes an information theoretical view on automatic model building. Even though much work on theory and evaluation is still needed, the firm theoretical foundation holds promise for an efficient method and advantageous scaling properties.