search

ctn-archive / nengo_theano

ABANDONED; see https://github.com/nengo/nengo instead
MIT License
3 stars 3 forks source link

Build time speedups #55

Closed studywolf closed 11 years ago

studywolf commented 11 years ago

ready to merge!

This drops the buildtime by a factor of 10-20. Just used numpy instead of creating theano functions for the encoders and random numbers in ensemble.py.

jaberg commented 11 years ago

Also, in the spirit of reducing up build times, I found it more natural to use numpy.linalg.lstsq() instead of the various dots and eigh stuff in ensemble_origin. It's probably faster too. Does it work in general?

A = neurons.accumulate(...)
decoders = lstsq(A, target_values)
studywolf commented 11 years ago

gaaahhhh wtf. the only thing i wanted to include was the very last commit, the one from today. that one only changes ensemble.py ... sigh. i'll just delete this and try again.

studywolf commented 11 years ago

good question! thought @tcstewar ?