POTEL, M. Pierre (2023) Stylebot: Learning to imitate diverse human behaviors through offline reinforcement learning. PFE - Project Graduation, ENSTA.

[img]PDF
Restricted to Repository staff only

614Kb

Abstract

This report aims at describing the process of style imitation learning, i.e., learning to capture the diversity of the behaviors in a dataset in an unsupervised way and learning to regenerate them. Two approaches are proposed to solve this problem, one based on variational autoencoders, and one based on decision transformers.

Item Type:Thesis (PFE - Project Graduation)
Uncontrolled Keywords:Reinforcement learning, offline reinforcement learning, variational inference, deep learning, generative models
Subjects:Information and Communication Sciences and Technologies
Mathematics and Applications
ID Code:9881
Deposited By:Pierre Potel
Deposited On:14 nov. 2023 14:46
Dernière modification:14 nov. 2023 14:46

Repository Staff Only: item control page