MILE: Memory-Interactive Learning Engine for Solving Mathematical Problems

Yuxuan Wu, Hideki Nakayama

Keywords: mathematical reasoning, symbolic reasoning, neural networks with memory

Abstract: Mathematical problem solving is a task that examines the capacity of machine learning models for performing logical reasoning. Existing work employed formulas as intermediate labels in this task to formulate the computing and reasoning processes and achieved remarkable performance. However, we are questioning the limitations of existing methods from two perspectives: the expressive capacity of formulas and the learning capacity of existing models. In this work, we proposed Memory-Interactive Learning Engine (MILE), a new framework for solving mathematical problems. Our main contribution in this work includes a new formula representing technique and a new decoding method. In our experiment on Math23K dataset, MILE outperformed existing methods on not only question answering accuracy but also robustness and generalization capacity.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning

TL;DR: A new learning framework interacting with memory embeddings for solving mathematical problems