Packet Routing Using Reinforcement Learning -
Theory and Simulation

This site contains a theoratical paper which present using reinforment learning as a method of routing packets in computer networks.
A running simulation of packet routing using reinforcement learning could be run on the 'nova' server from ~shlomy/learning/project/ by typing : "java learnnet.Network" and filling in the parameters.
The simulation enables comparison between different algorithms in different input patterns and network loads. The results are discussed in the paper.

Packet Routing Using Reinforcement Learning - The paper

Get this paper in postscript form

Page 001 Page 002 Page 003 Page 004 Page 005 Page 006 Page 007 Page 008 Page 009 Page 010 Page 011 Page 012 Page 013 Page 014 Page 015 Page 016 Page 017 Page 018 Page 019 Page 020 Page 021 Page 022