Just updated the project to make it better at making new words.
cooooool.
My AI is 60% done scanning and storing the big.txt document into memory.
Wow! 60%? That's impressive!
It's been teaching itself all night, and it actually generates words sometimes!
Very nice!
Would it go over 100%?
No, the percentage is how far it scanned the doc and stored it in memory. BTW, it's running at about 54 items (stored in memory) per second.
I also added a feature that can teach it even more once it's done.
@warped_wart_wars What's weird is that it learns faster with turbo mode OFF... This means now it's at around 60 items (stored in memory) per second.
But what does that code do? I guess it's looking for something in the upper half of probabilities? But that seems kind of arbitrary; it should accept a pair of letters with a probability weighted by the probability of that pair in the training data.
The average variable is the average of the number of occurrences of one of the two-letter items in the listings list. It is used to filter out stuff that probably won't come up in a word.
As the amount of data that is stored increases, the average goes higher, allowing a better chance of an actual word being reported.
I see. The cost of that, I guess, is that you never generate the rare words.
Maybe. I'm not very sure, though.
I just now made it so it auto-adjusts depending on how much learning data it has.
You know people can reply multiple times and only view it once, right...?
When I got to >700k, it didn't even generate anything within 20 seconds.
Ha ha!
I'm starting the learning process over.