This is an old revision of the document!


Basics of Reinforcement Learning

Overview

Starting from the idea of learning as the basis for intelligence, reinforcement learning (RL) has been proposed as the adaptive learning-based algorithm for decision making in animals emphasizing that the elements we encounter in this framework have empirical evidence accumulated from neuroscience studies. This workshop is concerned with setting RL in the framework of Markov Decision Processes, learning how to obtain goal-directed behavior of an agent by means of pure interaction with an environment. It is mixed with practical examples of learning with it. It will also briefly overview some of the recent success stories concerning RL and some research challenges and frontiers in this field of research.

Goals

  1. Basics of the framework of reinforcement learning
  2. How to write experiments in python and test the learning algorithms used for goal-directed behavior of an agent in an environment.
  3. How to scale this framework to high-dimensional input and output spaces, making it more similar to how animals learn.
  4. Interact with a deep learning framework for function approximation and apply this to the same task of learning from trial and error behavior.

Tutorials

Python - basic introduction http://cs231n.github.io/python-numpy-tutorial/

Environment setup for the practical session - download ipython notebooks and place them in your Google Drive, use colab to view or edit them.

Ipython notebooks:
Intro colab: https://github.com/ioanachelu/PracticalSessions/blob/master/introductory/Intro_Colab.ipynb
Intro numpy: https://github.com/ioanachelu/PracticalSessions/blob/master/introductory/Intro_Numpy.ipynb
Intro matplotlib: https://github.com/ioanachelu/PracticalSessions/blob/master/introductory/Intro_Plotting.ipynb (optional dar recomandat)
Intro tensorflow and sonnet: https://github.com/ioanachelu/PracticalSessions/blob/master/introductory/Intro_Tensorflow_and_Sonnet.ipynb (optional dar recomandat)

Registration

To be added.

sesiuni/basics-of-reinforcement-learning.1533836545.txt.gz · Last modified: 2018/08/09 20:42 by amacovei