Automated Algorithm Design

Goals

This project aims to revolutionize algorithm development by creating an automated framework that evolves hybrid algorithms outperforming existing methods. Using Multi-Objective Genetic Programming (MOGP), it combines advanced basis functions operating on vectors, matrices, images, and videos to design human-readable, competitive algorithms directly from data. MOGP generates a set of Pareto optimal solutions, allowing researchers to choose algorithms best suited to specific objectives and changing conditions. This approach frees researchers to guide optimization strategically and derive inspiration for new basis functions, fundamentally enhancing algorithm design in the era of big data and complex multi-objective challenges.

Issues Involved or Addressed

This project has many areas that should be investigated, including: improving the speed of evolutionary processes, integration of new basis functions from other domains, cloud computing, processing of big data sets, and application to the domains of interest.

Partners/Sponsors

Google

Methods and Technologies

  • Machine Learning
  • Multiple Objective Optimization
  • Python
  • GPI/Cuda Programming
  • Software Testing
  • Signal/Image/Video Processing
  • Cloud/Cluster Computing
  • C++
  • Open Source Software Development
  • Multi-domain Applications

Majors Sought

Computing: Algorithms, Combinatorics and Optimization, Analytics, Computer Science

Engineering: Aerospace Engineering, Analytics, Bioinformatics, Computer Engineering, Electrical Engineering, Industrial Engineering, Mechanical Engineering

Preferred Interests and Preparation

EE, CmpE, CS, AE, ME, ISyE, BME, HCI, Algorithms, Combinatorics, & Optimization, Bioinformatics
Background/interest in optimization, machine learning, signal processing, image processing, python programming.

Advisor

Jason Zutty
Jason Zutty
jason.zutty@gtri.gatech.edu

Day, Time & Location

Full Team Meeting:
5:00-5:50 Monday/Wednesday
Klaus 1440

Subteam meetings scheduled after classes begin.