On Apr 6, 2012, at 10:19 , Andreas Mueller wrote: > On 04/06/2012 08:04 AM, xinfan meng wrote: >> >> >> On Fri, Apr 6, 2012 at 1:57 PM, David Warde-Farley >> <[email protected]> wrote: >> On 2012-04-05, at 5:17 PM, Vlad Niculae <[email protected]> wrote: >> >> > >> > http://ufldl.stanford.edu/wiki/images/8/84/SelfTaughtFeatures.png >> > >> > It is easy to set up the skeleton of such an example and when the >> > implementation is good it will magically run :) >> >> I have stared at way too many MNIST weight visualizations over the years, >> and this one strikes me as "too good" to have been produced using supervised >> SGD. Based on the filename it sounds like some form of "self taught >> learning", which would imply an unsupervised architecture of some kind. >> >> Interesting. Does good visualization leads to good prediction? >> > No ;) > But it looks oh so good in the paper?! > > Btw, I looked at the exercise sheet an the visualisation of the weights there > looks quite different > and more like what I would expect.
This is correct and so is DWF. I couldn't find the correct image at the time so I linked to the autoencoder image. It still looks like pen strokes, though more noisy. I still think it is a nice example. In this case actually, I remember that increasing max_iter would make the image look better (less noisy) and the accuracy slightly increase. Vlad > ------------------------------------------------------------------------------ > For Developers, A Lot Can Happen In A Second. > Boundary is the first to Know...and Tell You. > Monitor Your Applications in Ultra-Fine Resolution. Try it FREE! > http://p.sf.net/sfu/Boundary-d2dvs2_______________________________________________ > Scikit-learn-general mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/scikit-learn-general ------------------------------------------------------------------------------ For Developers, A Lot Can Happen In A Second. Boundary is the first to Know...and Tell You. Monitor Your Applications in Ultra-Fine Resolution. Try it FREE! http://p.sf.net/sfu/Boundary-d2dvs2 _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
