| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
Is there a ROS node/facility to implement machine learning via Artificial Neural Network (ANN) ? I know, I can write my own C++ and Python codes. However since topics as navigation, SLAM and face recognition has some inbuilt resources at ROS, so I was hoping that ANN may also have such a node/package.
I used PyBrain to create a reinforcement learning application that was using gazebo simulator with pr2, and had rewards based on the topics published by the simulator and the robot nodes.
I believe that the ANN doesn't need to be part of ROS, you just need to implement your publishers subscribers.
By the way, pybrain is pretty powerful, and the ANN has an ok documentation, but the reinforcement learning classes are awfully documented.
No, as it was just an experiment for a course at my AI master, but I'll gladly send it to you by email if you are interested.
@Arkapravo OpenCV also includes a fairly extensive collection of machine learning routines including neural networks. See http://opencv.itseez.com/modules/ml/doc/ml.html.
Sure just let me know your email address. Recently I've been experimenting with what Arkapravo suggests, and I find opencv Machine Learning library much more powerful than pybrain and better documented. I would go with that one
Ok, thanks. My email adress is bojan_andonovski@yahoo.de
I'd recommend this: http://scikit-learn.org/stable/
@Victor could you send me the same files if you do not mind? emadmahrousawwad at gmail dot com
So far as I know, there is not. However, I'd be stunned if there weren't existing python or C++ libraries for these kinds of operations; writing ROS nodes that use such libraries is easy.
(In fact, making it easy to interface with other libraries is the _point_ of ROS)
Asked: 2012-03-11 03:25:36 -0500
Seen: 11,904 times
Last updated: Mar 12 '12
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
