Deep Learning, the Curse of Dimensionality, and Autoencoders (kdnuggets.com) 57 points by jonbaer 11y ago ↗ HN
[–] galapago 11y ago ↗ > Autoencoders are an extremely exciting new approach to unsupervised learningThey are not new. [–] jostmey 11y ago ↗ Yeah, but some of methods that make the technique successful are less than ten years old.
[–] jostmey 11y ago ↗ Yeah, but some of methods that make the technique successful are less than ten years old.
[–] mercurial 11y ago ↗ Seems to me that you would seriously reduce the size of the problem by first running a face-recognition algorithm, and then only analyzing the face... [–] chaosfactor 11y ago ↗ You've created a nice little chicken and egg problem for yourself. [–] im3w1l 11y ago ↗ I think he means first detect that there is a face at all, and where in the image it is. When you know that, do further processing. [–] mercurial 11y ago ↗ That's the idea.
[–] chaosfactor 11y ago ↗ You've created a nice little chicken and egg problem for yourself. [–] im3w1l 11y ago ↗ I think he means first detect that there is a face at all, and where in the image it is. When you know that, do further processing. [–] mercurial 11y ago ↗ That's the idea.
[–] im3w1l 11y ago ↗ I think he means first detect that there is a face at all, and where in the image it is. When you know that, do further processing. [–] mercurial 11y ago ↗ That's the idea.
[–] bayesianhorse 11y ago ↗ Sounds a bit like the age old method of producing medical professionals: Make them read and reproduce large amounts of medical facts.
[–] idunning 11y ago ↗ Original content is at http://nikhilbuduma.com/2015/03/10/the-curse-of-dimensionali...
8 comments
[ 2.7 ms ] story [ 35.8 ms ] threadThey are not new.