This demo is meant to show a genetic machine learning algorithm in action by training it to solve a simple maze. The agent (represented by a blue dot) must go from the green square to the red square without hitting any barriers. There are multiple presets of parameters for the genetic algorithm so you can see how the parameters affect the results. Below is the maze in question, which will display the best agent of each generation attempting to solve the maze. Below that are graphs of performance metrics of the genetic algorithm. Refer to the tutorials page for more information on the results of each preset. The parameters corresponding to each preset are as follows:
| Preset | Model Size | # of Generations | Population Size | Selection Proportion | Mutation Probability | Mutation Factor |
|---|---|---|---|---|---|---|
| Standard | 20 | 20 | 100 | 0.3 | 0.01 | 0.3 |
| Better | 50 | 20 | 100 | 0.3 | 0.01 | 0.3 |
| Low Population | 20 | 20 | 20 | 0.3 | 0.01 | 0.3 |
| High Mutation | 20 | 20 | 100 | 0.3 | 0.3 | 0.7 |