only 300 vectors

[the-Fish2]

solution: make it an array initially? more compressed than list issue - pca transform + matplot require list, array respectively

[the-Fish2]

Perhaps do a random sampling

[the-Fish2]

also make like a universal variable for all the places where I write 300

[the-Fish2]

• expanding to total # about 30 days

[the-Fish2]

*correction: 3.47 days

also A TON of repetition therefore dp store an array of arrays for connections between x and y and then add to matrix else yeah

[the-Fish2]

maybe also fill in the matrix with diagonal of like 1s and then it can't be the cos similarity answer?? idk- but then won't need to iterate through cluster

[the-Fish2]

try running birch on 2-D?

if birch is more efficient than my clustering it works out easier (waaaaaaay faster; 0.2 vs 30 secs)

[the-Fish2]

MAYBE THIS MATRIX SHOULD BE MORE LIKE A GRAPH. Like, if distance(w1, w2) = 10, and distance(w2, w3) = 6, then distance(w1, w3) is NOT less than 4 at best maybe store 'min bound's ? liiiike (depending on cosine dist similarity)

[the-Fish2]

or ignore points that are included in a cluster with other points

[the-Fish2]

afkljawrjwek fine run birch gives 44 dim

[the-Fish2]

current: 1000 vectors