Question 8.5.3 from the second mini project

Question 8.5.3 from the second mini project

by Pierre Louis Jean Marie Gaudillière -
Number of replies: 1

Hello!

I have a problem when I am trying to solve how many letters we can put into the list before the system becomes unstable. I had the idea to do it experimentally, that is to say to create a list of letter, store the list in the network, and then to take each letter of the list and let the system evolved for a few steps in order to check if every letter was a stable point. Hence I am doing a "for" loop but starting with the second letter of the list, many of the neurons values have already been flipped, even at step zero. I attached to this post the code and the results.

Thank you for your help and happy easter!

Pierre Louis Gaudillière

In reply to Pierre Louis Jean Marie Gaudillière

Re: Question 8.5.3 from the second mini project

by Chiara Gastaldi -

Hello!

First of all, a little reminder:  you shouldn't share your code.

The "experimental" approach that you are trying can definitively help in understanding what's new when patterns are letters, compared to the previous exercise. However, what you should focus on is this difference from a network dynamics point of view. What makes letters different from the previous patterns? These are the underlying questions of the letter exercise and you should show that you understood it in the comment part.