PyBrain - Python Machine Learning Library, Reinforcement Learning with complex Robots and more

This video demonstrates what problem domains can be successfully learned via Reinforcement Learning with PyBrain.
PyBrain also supports all other kinds of Machine Learning domains like Supervised Learning, Artificial Evolution and Unsupervised Learning.

A lot of examples and HowTo's as well as a thorough documentation make it very easy to create your own simulations within PyBrain or to connect your problem to PyBrain to successfully learn useful behaviors.

PyBrain supports you with several models for behavior learning: Artificial Neural Networks, Recurrent Neural Networks, Gaussian Processes ...

Visit www.pybrain.org for more information.

PyBrain is a cooperative project by IDSIA and the CogBotLab Group of the Technische Universität München.
Both groups are led by Prof. Jürgen Schmidhuber:
http://www.idsia.ch/~juergen/

Related Publications:
follows

Musik:
follows

Category:

Science & Technology

Tags:

• Machine Learning
• Reinforcement Learning
• Artificial Intelligence
• Neural Networks
• Recurrent Neural Networks
• LSTM
• Robotic

[Epistemophilos]

Which ODE/simulator did you use for the simulated examples?

[NeuroDad2]

@Epistemophilos We used PyODE, but with an older python version. Sadly there is an unsolved garbage collector bug with PyODE and Python 2.6 or newer.

Most people I know that work with PyBrain use nowadays Bullet for the simulations.

[NeuroDad2]

Thx :)

[gabrius99]

@NeuroDad2 How do they manage to find a Bullet library for python?

[darkhaxx0r]

Robotic masturbation montage @ 4:47

[Epistemophilos]

@NeuroDad2 Ok. What about Gazebo - is that less suited?

[Epistemophilos]

@Epistemophilos Ok, I see from the pybrain documentation that it's probably pyode and opengl.

[KilianKool]

cool! :-)