search

pthomaid / machinelearning

MIT License
0 stars 1 forks source link

i am waiting for content! :P #1

Open Mourgolikos opened 8 years ago

Mourgolikos commented 8 years ago

:dancers: :dancers: :dancers: :dancers: :dancers: :lollipop:

pthomaid commented 8 years ago

i am waiting for feedback! :P

Mourgolikos commented 8 years ago

Oh! I see a loop-party! :D

Oh... come on, one-letter variable names? nevermind... Challenge accepted! :P

pthomaid commented 8 years ago

numpy version with no loops uploaded! Really slow though..

Mourgolikos commented 8 years ago

you may use multiple threads in order to create each network in it's own thread. An impressive increase in execution speed expected.

Tip:

from multiprocessing.dummy import Pool

#make the Pool of workers
pool = Pool(2) # Threads count

networkCollection = pool.map(functionName, variables)

# some magic here

pool.close()
pool.join()

I am sure you got the idea ;)

Mourgolikos commented 8 years ago

Furthermore on the previous version without numpy (backpropagation.py), if i understand correctly, you are doing some operations, between 2D and 1D lists, in such a way that the function 

zip()

info about the zip() comes in handy. ;)