-This plots character based on how ‘similar’ they are. -Notice how numbers are grouped together, as well as most common letters. There is no real explanation why the characters are exactly where they are. Why does it think an asterisk is more similar to a number than a period for instance? If I ran same code again, it would produce different results because of randomness involved in the process. You would still see the same overall trends though, such as numbers being close together. ***Nearly all of the code inside is not used when run; the code was used to pre-calculate the positions of the characters on the screen.
Details: -This uses the GLoVe algorithm as well as an autoencoder, 100% in Scratch. -The model was trained on Leo Tolstoy’s “War and Peace” as well as H.G. Wells’ “The War of the Worlds” Note: having gone through the code again, I realized this has some flaws! I implemented the correct code in Python, but will not fix the Scratch code in the foreseeable future because it still produces good results.
