He/she will work on optimization techniques for fast retraining of deep neural networks. The project is within the context of sustainable AI, where the current paradigm of retraining from scratch is replaced by a methodology based on continual learning of networks. The focus will be on optimization techniques which prevent the stability gap (which temporarily reduces the network performance).
The tasks include:
(i) Implementation of algorithms for fast adaptation to new data.
(ii) the writing of the research results in a publication.
(iii) supervision of master and/or PhD students related to the project.
Closing date: 26/11/2024
Més posts de Recerca
Cap comentari:
Publica un comentari a l'entrada