These are chat archives for thunder-project/thunder

27th
Mar 2015
Noah Young
@npyoung
Mar 27 2015 00:32
@freeman-lab got it, thanks. Now that I have the latest code I'm noticing that thunder.NMF's docstring lists a w0 argument that doesn't seem to be implemented in the actual __init__. Is this feature going to be available soon?
Gilles Vanwalleghem
@Yassum
Mar 27 2015 03:04
@jwittenbach For the design matrix, I initiate it using numpy.zeros((1,360),dtype=np.int), then put the relevant values to 1. If I use this, it gives me a singular matrix error. It works if I make it a numpy.zeros((2,360),dtype=np.int) though.
Anyway I am trying to do the direction tuning now and it gives me errors when I try to access the params RDD. I'll keep trying to figure that out.
Jason Wittenbach
@jwittenbach
Mar 27 2015 03:20
@Yassum Something along these lines works for a single binary regressor:
series = tsc.loadExample('fish-series')
L = len(series.index)
X = np.zeros((1, L))
X[0,20:30] = 1
results = thunder.RegressionModel.load(X, 'linear').fit(series)
results.first()