Multi Task Learning
Multi-task learning (MTL) aims to improve model efficiency and generalization by training a single model to perform multiple related tasks simultaneously. Current research focuses on addressing challenges like task interference and optimization difficulties, exploring architectures such as Mixture-of-Experts (MoE), low-rank adaptors, and hierarchical models to enhance performance and efficiency across diverse tasks. MTL's significance lies in its potential to improve resource utilization and create more robust and adaptable AI systems, with applications spanning various fields including natural language processing, computer vision, and scientific modeling.
Papers
GDOD: Effective Gradient Descent using Orthogonal Decomposition for Multi-Task Learning
Xin Dong, Ruize Wu, Chao Xiong, Hai Li, Lei Cheng, Yong He, Shiyou Qian, Jian Cao, Linjian Mo
Learning Vision-based Robotic Manipulation Tasks Sequentially in Offline Reinforcement Learning Settings
Sudhir Pratap Yadav, Rajendra Nagar, Suril V. Shah