Matlab codes for solving the following tasks using the k-winner-take-all neural network architecture: Control tasks:
- Puddle World
- Mountain Car
- Acrobot task
The codes also include:
- Regular back-propogation (1 hidden layer without sparse coding)
- Linear (no hidden layer)
- Q-table architectures.