Titulo Estágio
Automatic Design of Efficient Machine Learning Models
Áreas de especialidade
Sistemas Inteligentes
Engenharia de Software
Local do Estágio
CISUC
Enquadramento
Nowadays ML models are becoming larger which makes them highly inefficient energy-wise. It has been estimated that training GPT-3 consumed 1287000000 Wh which emitted 552 tons CO2e [1] which is the equivalent of 123 vehicles driven for one year or roughly the amount of electricity consumed by 120 American Households during a year. These energy costs make Deep ML inaccessible to most individuals and companies (even large ones), hindering its applicability, and making it energetically unsustainable globally.
1 - Patterson, David, et al. "The carbon footprint of machine learning training will plateau, then shrink." Computer 55.7 (2022): 18-28.
Objetivo
The main goal of this work will be to explore the use of NeuroEvolution to create Energy-Efficient models. In concrete, we will analyse the usage of carbon footprint scores to guide the evolution of methods that use less energy.
Plano de Trabalhos - Semestre 1
1 - Literature Review
2 - Framework Selection/Proposal
3 - Preliminary Implementation
4 - Writing of the intermediate report
Plano de Trabalhos - Semestre 2
5 - Analysis of the results
6 - Framework Refinement
7 - Validation of the results
8 - Scientific Article with the main results
9 - Writing of the thesis
Condições
The work is to be conducted at the ECOS and CMS groups of CISUC. A workplace will be provided as well as the required computational resources.
There is a possibility of the student being awarded a scholarship (Bolsa de Investigação para Licenciado) for at least 6 months. The scholarship will follow the Fundação para a Ciência e Tecnologia (FCT) monthly stipend guidelines.
Orientador
Nuno Lourenço / Penousal Machado
naml@dei.uc.pt 📩