I think that I may have found something that is even better than Neural Net Lottery Picker and Percep-Lotron.
It is called Neural-Lotto. You will have to Google it to see what I mean. Do not forget the dash in the name when you Google it.
It just went live in January of this year.
It has eight parameters. They are as follows:
Learning Rate --- from 0 to 1
Neurons / Layers --- from 10 to 990
Layers --- from 1 to 999
Stepping --- from 1 to 200
History --- from 10 to 200
Epochs -- from 100 to 10,000
Algorithm --- from 1 to 5
Momentum -- from 0 to 1
I have been experimenting with the settings. I have not yet had a chance to study the results.
Now that I have told all on this forum / website of the existence of this thing, I hope that someone will reciprocate by posting the results of his or her experiments. I am particularly interested in knowing the best possible settings for a Pick 5 game such as Mass Cash which I play.