[Scikit-learn-general] my silence

Just a quick note that I've been silent lately because I've been Busy With Life, but also because github was notifying an email address hosted at my previous employer, which was deactivated a fortnight ago. If there were issues that sought my particular attention, please let me know. --

Re: [Scikit-learn-general] [GSoC2015 metric learning]

Apparently, the problem was with multiplying big 4D tensors: grad = np.sum(proba[:, :, np.newaxis, np.newaxis] * outer, axis=1).sum(axis= 0) I don't know the details, but line-prof said that most of the time was spent here (and the other similar line), probably it was due to copies. I recalled you

Re: [Scikit-learn-general] [GSoC2015 metric learning]

On Sun, May 31, 2015 at 7:25 PM, Artem wrote: > I added a simple benchmark > > that > compares NCA-assisted 1NN with default one (with Euclidean distance) on > Wine dataset (it was one of the datasets rep

Re: [Scikit-learn-general] [GSoC2015 metric learning]

I added a simple benchmark that compares NCA-assisted 1NN with default one (with Euclidean distance) on Wine dataset (it was one of the datasets reported in the NCA paper). See my output here