Reproduce “Learning Continuous Semantic Representations of Symbolic Expressions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Miltiadis Allamanis, Pankajan Chanthirasegaran, Pushmeet Kohli, Charles Sutton
EQNET learns vectors for symbolic expressions, bringing equivalent expressions closer despite different syntax. It recursively combines vectors along expression trees and uses subexpression autoencoding as training regularization. On Boolean and polynomial benchmarks, neighborhoods generally reflected equivalence better than comparison models. On BOOLL5 unseen classes, score₅ was 75.2%, versus 0.5% for a two-layer TREENN; this is not general equivalence checking.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.