Geoffrey Hinton has joined Google

Post by abrichr » Wed, Mar 13, 2013, 7:15 PM UTC

drpgq wrote: post #85

ninjin wrote: post #38

Earlier quoted context omitted.

In a way, yes. Research in industry has the advantage that you don't have to "waste" your time applying for grants in order to fund your own research group. A lot of people seem to live in an illusion that professors actually spend most of their time doing research, they are the managers of research groups more than anything else and growing those research groups is hard work and takes up close to all of your time (t…

I won't put it up here, but those that are curious can look up Hinton's salary at the University of Toronto by looking at the Ontario government sunshine list website.

Link for the lazy: http://www.fin.gov.on.ca/en/publications/salarydisclosure/20...

Post by srean » Wed, Mar 13, 2013, 8:50 PM UTC

IanCal wrote: post #81

srean wrote: post #53

HN'ers may have a good laugh at these taken from Yan LeCun's page. LeCun made it a tradition to have Hinton jokes in the lines of Chuck Norris ones (or more appropriately Doug McCllroy ones). A few will recall that neural networks all but died from the US after Minsky's damning book. Hinton gave backpropagation which is one of the foundational pillars of feed-forward neural net algorithms. With his new thrust on what…

I like the quotes, however: > With his new thrust on whats called "deep belief networks" he is challenging his own early seminal contribution in the field I don't know if I agree with that, he still uses backprop. Backprop has always been known to have problems when you scale to millions of connections, and his work on RBMs/DBNs is really quite old. What was novel more recently was showing that the contrastive diverg…

Mostly agreed, one difference that I would like to highlight is that errors are not always backpropagated across all the layers. In addition to contrastive divergence the breakthrough has been that you can get away with unsupervised learning (like with autoencoders) in the layers.

On the comment that RBMs are new, now I have to come to accept that if one looks hard enough almost all things are old, only the name changes !

Re: Geoffrey Hinton has joined Google

Re: Geoffrey Hinton has joined Google