Исследовательская группа

Прикладное машинное обучение и глубокое обучение

Scaling up and assuring reinforcement learning with abstract MDPs

April 24

While reinforcement learning (RL) had recent great successes in game AI and other decision making tasks, there are still two major challenges:(1) Scaling up RL to complex tasks(2) Assuring properties such as safety of the learning process and the learning result.In this talk, I will show how abstract Markov Decision Processes (MDPs), reward shaping, and quantitative verification can be used to tackle these challenges.

Докладчик: Дэниэл Куденко

Время начала: 18:30