Concrete Type Inference for Code Optimization using Machine Learning with SMT Solving

Explore a groundbreaking approach to concrete type inference for dynamically typed languages in this 18-minute conference talk from OOPSLA2 2023. Discover how researchers from Georgia Institute of Technology combine machine learning models, including GPT-4, with SMT solving to enable code optimization without requiring programmer-provided type information. Learn about the three algorithms developed, including a hybrid approach that significantly outperforms individual methods. Examine experimental results showing impressive performance improvements, with geometric mean speedups of 26.4× using Numba and 62.2× using the Intrepydd optimizing compiler. Understand the potential impact on programmer productivity and resource efficiency in Python applications. Access supplementary materials and related research through provided links and ORCID identifiers.

[OOPSLA23] Concrete Type Inference for Code Optimization using Machine Learning with SMT S...