Hacker News new | past | comments | ask | show | jobs | submit login
Training Recurrent Neural Networks [pdf] (utoronto.ca)
88 points by e19293001 on Oct 6, 2016 | hide | past | web | favorite | 8 comments



I found this from neural-networks course in coursera[0]. The author of this paper had been discussed as an example of what recurrent neural nets can now do.

Here's the description from the slide:

     •  Ilya Sutskever (2011) trained a special type of recurrent
      neural net to predict the next character in a sequence.

     •  After training for a long time on a string of half a billion
      characters from English Wikipedia, he got it to generate new
      text.

           – It generates by predicting the probability distribution
            for the next character and then sampling a character from
            this distribution.

           – The next slide shows an example of the kind of text it
            generates.

     Notice how much it knows!
     
     Some text generated one character at a time by Ilya Sutskever’s
     recurrent neural network:
     
     In 1974 Northern Denver had been overshadowed by CNL, and several
     Irish intelligence agencies in the Mediterranean region. However,
     on the Victoria, Kings Hebrew stated that Charles decided to
     escape during an alliance. The mansion house was completed in
     1882, the second in its bridge are omitted, while closing is the
     proton reticulum composed below it aims, such that it is the
     blurring of appearing on any well-paid type of box printer.
[0] - https://www.coursera.org/learn/neural-networks/


You probably know this already, but if you are just looking for an illustrated demonstration of what RNNs are capable of in the text domain, probably the best brief article is the post by Andrej Karpathy about [The Unreasonable Effectiveness of Recurrent Neural Networks](http://karpathy.github.io/2015/05/21/rnn-effectiveness/).


Also needs 2013 designation,

Especially important as neural net knowledge seems to be evolving quickly.

And perhaps someone can explain why paper matters relative to the plethora of papers and approaches "out there".


> The certificate is only valid for the following names: www.cs.toronto.edu, cs.toronto.edu


You can just change the URL to https://www.cs.toronto.edu/~ilya/pubs/ilya_sutskever_phd_the... to get valid https.


Could you explain how this works?


The University of Toronto owns both the domains toronto.edu and utoronto.ca. Some sub-sites work on both domains, and some don't. Apparently they misconfigured their servers so that they are using the toronto.edu certificate on a utoronto.ca site.


Considering that it is from 2013 and from Hinton team then are the Restricted Boltzmann Machines actually necessary for this?




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: