|
Download the Free Connectionist Q-learning Java Framework v.1.0 :
(java source code included)
Download the examples on how to use it (see the demo applets ). Please, ensure you've placed those jars together with qLeraningFramework.jar in one directory: qPatterns.jar - if you set Temporal Difference parameters to zero, the framework can act as a standard neural network trained with backpropagation. Here used as pattern recognition network. Name
| Size
(click for download) | Screenshot | Description | TATRY
C++Builder 3 | <disabled>
|  | Take a breath and walk through real Polish Tatra mountains on your computer.
On the screenshot you can see:
Kozia Przelecz, Hala Gasienicowa, Koscielec and Swinica.
See how it matches the reality!!! | Etology
Turbo Pascal | 50 kB
+code | 
| Wonderful ecosystem simulation with predator - prey dependence. 180 animals * 160 neural connections. After few days of evolution, I obtained animals capable of flocking and screaming(coloured circles) in case of seeing predator coming too close.
Map.txt is the file that describe the environment. You can simply open it in notepad and change food caloricity, mutation rate or trees position(preys hide there from predators). For more info click here | Tetribot
C++
| 50 kB
+code |  | Is a neural network able to play Tetris?
The following program tries to answer this question. It presents 14 bots playing Tetris simultaneously. Each bot consists of 250-neuron-net (each neuron has 5 dendrites). After he ends his game, program counts his evaluation and decides, whetrher he has to be removed or not. Removed bot is replaced by an offspring of two best bots (crossing-over + mutation). Press "s" to slow down the simulation. | Evolrace
Turbo Pascal
and C++ | 78 kB
C++ code | 
| Evolution of Formula 1 cars. 100 cars run a race each time.
The winner has 99 mutated offsprings which start in the next race. Each car controlled by a neural net. You can draw your own track and runtime-modify cars` neural nets. | Tunnel
Turbo Pascal | 63 kB
+code | < | Tunnel is a predator-prey simulation. Both of them have their own neural networks, whose evolution you can observe simultaneously. Notice, that the network is divided into four different nets, each one used for a different hungry level. For example if the animal is hungry, it can learn to run towards food (on the chart "food" white line is higher than the others), otherwise it seeks for the partner. | Sierpinski3D
Turbo Pascal | 27 kB
+code |  | The 3D version of Sierpinski triangle!
You can walk into it using arrow keys.
Page Up/Page Down - vertical movement | 3D brick
Turbo Pascal | 49 kB
+code |  | The 3D kinematics simulator I mentioned in Future plans.
Use 2,8,4,6 keys to move the brick. | Keos
Delphi 1 | 417 kB |  | Program for learning words, especially for Polish users. Tested on my Latin, English and German lessons. |
|
