Using Python & Numpy to reproduce this paper:
Földiák, P. Forming sparse representations by local anti-Hebbian learning. Biol. Cybern. 64, 165–170 (1990). https://doi.org/10.1007/BF02331346
See the Jupyter Notebook for details.
The feedforward weights of the 16 neurons visualized as a 8x8 grid of pixels.
Initial neurons:
Neurons after 600 trials:
Neurons after 1200 trials:


