EVENTO
Optimizing the Energy Efficiency and Performance of Deep Learning Algorithms Using Hyperparameter Self-Tuning Techniques: A Step Towards Green AI
Tipo de evento: Seminário de Avaliação - Série A
The use of Artificial Intelligence (AI) algorithms has grown significantly in the last decade, and continues to grow at a fast pace. This expressive growth is mainly due to the Deep Learning (DL). However, DL algorithms require a high computational power to reach the desired levels of precision, often requiring thousands of machine hours on HPC, which limits access to these algorithms to small institutions or groups of researchers. In addition, the use of these computing resources generates high energy consumption, which has become a major problem, not just financially, but also ecologically. Recent studies indicate that the current trajectory of AI, with DL models trained on increasingly large data sets, is unsustainable and harmful in several ways, including its massive carbon footprint. Motivated by the importance of this topic, this work proposes to present a methodology capable of producing more efficient architectures of DL algorithms, both with regard to the consumption of computational resources, energy and time, with minimal loss of accuracy in predictions. Thus generating more sustainable algorithms (economically and ecologically) in search of a greener AI. In addition, also makes the training and use of these algorithms more inclusive, i.e., accessible to small groups or institutions so that they can enjoy the benefits offered by these algorithms. One approach that has shown promise in reducing these costs is the development of Automated Machine Learning (AutoML), which the goal is to automate the development of some parts, or even the entire pipeline, of an ML model. There are already several AutoML frameworks available. However, the most proposed approaches only seek to maximize the predictive performance in a given task without considering aspects related to energy efficiency. Thus, we propose develop and evaluate an AutoML approach to improve the efficiency of ML models. The term efficiency encompasses the accuracy of the models, the time to reach a solution, and its energy consumption. This leads to a multi-objective optimization problem which we propose to solve through the Genetic Algorithms (GA) focusing on the architecture and hyperparameter search. It is being developed to be part of an AutoML solution to increase ML pipeline automation's efficiency to train and run DL algorithms.Para assistir acesse:meet.google.com/pzk-hgys-oiq
Data Início: 09/12/2022 Hora: 09:00 Data Fim: 09/12/2022 Hora: 12:00
Local: LNCC - Laboratório Nacional de Computação Ciêntifica - Videoconferência
Aluno: Andre Muniz Yokoyama - - LNCC
Orientador: Bruno Richard Schulze - Laboratório Nacional de Computação Científica - LNCC Mariza Ferro - Universidade Federal Fluminense - UFF
Participante Banca Examinadora: Bruno Richard Schulze - Laboratório Nacional de Computação Científica - LNCC Eduardo Bezerra da Silva - Centro Federal de Educação Tecnológica Celso Suckow da Fonseca - CEFET-RJ Pablo Javier Blanco - Laboratório Nacional de Computação Científica - LNCC
Suplente Banca Examinadora: Fabio André Machado Porto - Laboratório Nacional de Computação Científica - LNCC
