Machine Intelligence and Symbolic Computation
at
Applications of Computer Algebra (ACA) 2024
University of Havana
Havana, Cuba, June 25 - 28, 2024

Recently, Machine Learning (ML) methods have achieved spectacular results in solving difficult and highly relevant problems like natural language (NL) translation, NL dialogue generation, object identification in photos, etc. In this session, the interaction of ML methods with Symbolic Computation (SC) methods will be studied:

- the application of ML methods to improve SC methods,

- the application of SC methods to improve ML methods,

- the collaboration of SC methods and ML to create new powerful algorithms for hard problems.

Examples of topics falling into the scope of the session are given below. Talks focusing exclusively on either ML or SC will not be considered.

Organizers

• Bruno Buchberger (RISC, Johannes Kepler University, Austria)

• Bo Huang (Beihang University, China)

• Dongchen Jiang (Beijing Forestry University, China)

• Dongming Wang (Beihang University, China and CNRS, France)

• Jing Yang (Guangxi Minzu University, China)

Examples of Topics

- semi-automating mathematical theory exploration (concept formation, conjecture formation, proving, etc.) by combining SC and ML (LLMs) methods,

- semi-automating software development from NL specifications by combining SC and ML (LLMs) methods,

- combining ML and SC methods in systematic testing,

- translation between formal mathematical text and NL mathematical text using ML (e.g. relation extraction from NL text),

- mathematical knowledge management (e.g. formulae retrieval in math knowledge bases) using ML,

- generating suggestions for next steps (predicting useful lemmas, proof strategies, resolvents, etc.) in theorem proving (e.g. resolution proving) by ML,

- integrating mathematical algorithms (algorithm libraries, mathematical software systems) into LLMs (and vice versa),

- math education supported by combined SC and ML methods,

- collaborative math writing supported by LLMs,

- SC inside training algorithms, e.g. symbolic regression and fitting (learning appropriate formal expressions for modeling labeled data),

- symbolic data augmentation to generate new training data,

- incorporating symbolic constraints (expressing known logical or physical rules) into ML training,

- explaining ML models by symbolic expressions that (hopefully) express structure in the domains modeled (“explainable AI”),

- using ML for complexity formulae for hard SC problems.

If you are interested in attending ACA 2024 and in proposing a talk at this session, please send an abstract of your talk to Bo Huang. Please use this LaTeX template for your abstract and send both the LaTeX source and a compiled PDF version. We suggest that abstracts be no more than 2 pages including references.

Talks

• The AI Mathematician: From Physics, to Geometry, to Number Theory

  Yang-Hui He (University of Oxford, UK)

• A Dataset for Suggesting Variable Orderings for Cylindrical Algebraic Decompositions

  Rui-Juan Jing (Jiangsu University, China)

• Formal Representation of Mathematical Texts

  Dongchen Jiang (Beijing Forestry University, China)

• A Framework of Plain Geometric Knowledge Discovery

  Weifeng Shang (Beihang University, China)

Important Dates

Submission of talks: May 15, 2024

Registration notification: June 1, 2024 (all registration is on-site)

Go to:

ACA 2024 main page

Conferences on Applications of Computer Algebra main page