ICMS 2020 Session

Artificial Intelligence and Mathematical Software

Accepted Talk

Stephen Forrest (Maplesoft)
Deep Learning in Maple
Abstract: The 2018 release of the MAaple computer algebra system included a new package, DeepLearning, which was built around a repackaged distribution of Google TensorFlow; this represented the first effort at making machine learning tools generally available within official distributions of Maple. While the core algorithms of Maple do not yet make direct use of DeepLearning, its availability within the system has enabled some experiments with the goal of replacing existing heuristics crafted with machine-generated ones. We report here on the further integration of DeepLearning within Maple and, in particular, on some ongoing work towards building a neural network-based expression simplifier.