Smalltalk - Re: Agile Artificial Intelligence first trained Perceptron

Smalltalk › Pharo › Pharo Smalltalk Users

Re: Agile Artificial Intelligence first trained Perceptron

Posted by Pharo Smalltalk Users mailing list on Oct 05, 2020; 1:29am
URL: https://forum.world.st/Agile-Artificial-Intelligence-tp5119233p5122993.html

Hello,

Sorry to reply late. I just saw this email.

Each training modify the weights and bias of the perceptron by a small quantity. It is just that, in this very particular case, from 31 or 32 is the drop that makes all things change...

Cheers,

Alexandre

--
_,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:
Alexandre Bergel http://www.bergel.eu
^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.

On 21-08-2020, at 01:31, bentai <[hidden email]> wrote:

I noticed that if I trained the OR gate with 31 iterations, the first test
would fail.

If I train with 32 iterations, then the test for #(0 0) would pass.

What is the explanation for this "tipping point"?

The other tests pass, even if timeRepeat := 1.

I just decided to experiment a bit.

--
Sent from: http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html