Learning to Discover Efficient Mathematical Identities

Abstract

In this paper we explore how machine learning techniques can be applied to the discovery of efficient mathematical identities. We introduce an attribute grammar framework for representing symbolic expressions. Given a set of grammar rules we build trees that combine different rules, looking for branches which yield compositions that are analytically equivalent to a target expression, but of lower computational complexity. However, as the size of the trees grows exponentially with the complexity of the target expression, brute force search is impractical for all but the simplest of expressions. Consequently, we introduce two novel learning approaches that are able to learn from simpler expressions to guide the tree search. The first of these is a simple n-gram model, the other being a recursive neural-network. We show how these approaches enable us to derive complex identities, beyond reach of brute-force search, or human derivation.

References in corpus (4)

• Efficient Estimation of Word Representations in Vector Space
• Improving neural networks by preventing co-adaptation of feature detectors
• Exact solutions to the nonlinear dynamics of learning in deep linear neural networks
• Can recursive neural tensor networks learn logical reasoning?

Cited by in corpus (18)

• Learning to Perform Local Rewriting for Combinatorial Optimization
• Protein Secondary Structure Prediction with Long Short Term Memory Networks
• Learning Program Embeddings to Propagate Feedback on Student Code
• Dynamic Neural Program Embedding for Program Repair
• Deep Learning for Symbolic Mathematics
• Learning Equations for Extrapolation and Control
• Learning Continuous Semantic Representations of Symbolic Expressions
• A Review of Emerging Research Directions in Abstract Visual Reasoning
• Attending to Mathematical Language with Transformers
• Learning to superoptimize programs
• Visual Learning of Arithmetic Operations
• Can Neural Networks Understand Logical Entailment?
• Informed Equation Learning
• Compositional Generalization with Tree Stack Memory Units
• Improving the Robustness to Data Inconsistency between Training and Testing for Code Completion by Hierarchical Language Model
• Improve Language Modelling for Code Completion through Statement Level Language Model based on Statement Embedding Generated by BiLSTM
• Generating Symbolic Reasoning Problems with Transformer GANs
• Pretrained Language Models are Symbolic Mathematics Solvers too!