Links to Online Demos
 

The following is a sampling of some online demonstrations of neural net training algorithms. I encourage you to at least play around with some of these. My favourite, is the TensorFlow playground. Let me know if you find any broken links.






Boltzmann Machine

This is Hinton’s Neural Network Simulation. It shows what I believe is a Restricted Boltzmann Machine used to learn handwritten numerals. The simulation can be used in two ways as shown below.




















































Associative Memory

A Java implementation of a fully working associative memory. The demonstration is based upon storage and recall of images.


Channel Assignment in Cellular Networks

Reinforcement Learning is used to address the channel assignment problem encountered in cellular networks.


Q-learning

An applet that shows the Q-Learning applied to a crawling robot


Hopfield Net

An applet demonstrating the Hopfield Net

Select here to regenerate a number

Select one of these images to attempt recognition

Position of white box above represents the numbers 0-9

or

TensorFlow Playground

This is a a wonderful, interactive neural net. Play around with the number of layers,

hidden neurons and other parameters to see how they affect learning.

Deep Learning Demos in Javascript using Keras

Lots of cool deep neural nets to try here!


Grid World Online Demo

You can set up your own grid world demo as in Sutton’s book, including if you like

the cliff-walking problem.


VGG16

A 16 layer deep neural net image classifier built by the Visual Geometry Group at Oxford.


Article: Deep Reinforcement Learning

A useful technical introduction to the topic of deep reinforcement learning.


Article: Q-Learning with TensorFlow

An article that describes how to build RL agents using TensorFlow.


Video: Introduction to RL with Function Approximation

By Rich Sutton, this video is well worth watching! He offers some useful tips if you are finding that your RL agent is not converging. Related slides.


Article: Generating Image Descriptions


Article: The Unreasonable Effectiveness of Recurrent Neural Nets

This is article is lots of fun to read!


Google AI Experiments

Fun hands-on activities in AI!


t-Stochastic Neighbourhood Embedding

Demo using MNIST data


Cool 3d Demo of Digit Recognition