| Wanderbot applet |
|
This applet shows how fast is the learning process, when you use QCON Framework. The bot is controlled by one-layer neural network with random weights at start. It learns using the Connectionist Q-learning algorithm. It is rewarded for going straightforward and punished for collisions. The little circles behind the bot show the perception sensors. When such circle detects an orange obstacle, it gives "1" to the brain's input; "0" otherwise.
Only registered users can write comments. Powered by AkoComment 2.0! |
||
