Parameter-Efficient Transfer Learning
Project Description
There are lots of pretrained models are released recently, this project aims to design efficient transfer learning algorithms to adapt these pretrained models to downstream applications.
Supervisor
CHEN, Long
Quota
3
Course type
UROP1000
UROP1100
UROP2100
UROP3100
UROP3200
UROP4100
Applicant's Roles
Design algorithms for parameter-efficient transfer learning.
Prerequisite: Basic deep learning and machine learning knowledge, strong Python programming ability
Prerequisite: Basic deep learning and machine learning knowledge, strong Python programming ability
Applicant's Learning Objectives
Understand the state-of-the-art methods for this direction, and try to propose another effective solution.
Complexity of the project
Moderate