Learning rate and regularization parameters were cut to integers before given to libFM

jfloff / pywFM

pywFM is a Python wrapper for Steffen Rendle's factorization machines library libFM

https://pypi.python.org/pypi/pywFM

MIT License

250 stars 43 forks source link

Learning rate and regularization parameters were cut to integers before given to libFM #8

Closed capdaha closed 8 years ago

capdaha commented 8 years ago

Learning rate and regularization parameters are, in general, float numbers. On creating argument for libFM they are cut to integers, because %d is the integer conversion type. It causes SGD to not learn, when learning rate is chosen less than 1. Changed to the float conversion type %f.

jfloff commented 8 years ago

Sorry for the delay, I was on vacation.

Great find! Thank you for the contribution!