Alex Graves
cs.toronto.edu/~graves · Personal blog · on HN since 2013
Alex Graves, a machine learning researcher, hosts papers and demos on recurrent neural networks, including handwriting generation and CTC loss.
What HN saysHN commenters link his handwriting generation demo as an early RNN example and recommend his CTC and variational noise papers as readable and useful.
24HN people shared it
14linked it in comments
10stories submitted
298story points
12 months agolast shared
2013people per year2026
Most-shared posts
- 1Handwriting Generation with Recurrent Neural Networks12 people · 270 points · first shared 2013 · 80 comments on HN
- 2Recurrent neural network handwriting generation demo3 people · 8 points · first shared 2013 · discussion on HN
- 3Supervised Sequence Labelling with Recurrent Neural Networks [pdf]2 people · 3 points · first shared 2016 · discussion on HN
- 4Icml 2006.pdf2 people · first shared 2019
- 5Recurrent neural network handwriting generation demo1 people · 14 points · first shared 2015 · 3 comments on HN
- 6Recurrent Neural Network Handwriting Generation Demo1 people · 2 points · first shared 2014 · discussion on HN
- 7Neural Network Handwriting Generation1 people · 1 points · first shared 2015 · discussion on HN
- 8Handwriting1 people · first shared 2015
- 9Nips 2011.pdf1 people · first shared 2025
- 10Handwriting1 people · first shared 2014
- 11Handwriting1 people · first shared 2015
What HN people wrote when they linked it
fxtentacle · 2024-06-12
The constant fighting between Schmidhuber and LeCunn should be well known in here. But in my opinion, both of them have been involved with fantastic papers that are both painless to read and full of useful info:
CTC loss: https://www.cs.toronto.edu/~graves/icml_2006.pdf
DeepLearning: https://hal.science/hal-04206682/document
And both of these papers are new and on more recent research, so it's not just old papers that are good. It's just that there is a flood of bad new papers drowning out a constant signal with additional noise.
kastnerkyle · 2014-09-24
I think neural networks (RNNs to be specific) have the potential to do this - coupled with these handwriting bots it would be pretty hard to tell which is which. See the research of Alex Graves http://www.cs.toronto.edu/~graves/handwriting.html
This type of research is also pretty amazing for speech synthesis.
Houshalter · 2014-05-28
Perhaps you could get around the high error rate of speech recognition by having it ask the user again when it's not certain, and use prior probabilities on what they likely said ("wreck a nice beach" is less likely than "recognize speech".)
I wonder if a generative model would be acceptable at producing speech. It works really great on handwriting (http://www.cs.toronto.edu/~graves/handwriting.cgi?text=This+is+amazing…
For the underlying AI, this is a great demo of an AI assistant that can do things like you describe (https://www.youtube.com/watch?v=54HpAmzaIbs) . Good natural language processing, it can learn new facts and even asks questions about things it wants to learn more about.
albertzeyer · 2025-11-18
The random noise is added to the model parameters, not the inputs, or not?
This reminds me of variational noise (https://www.cs.toronto.edu/~graves/nips_2011.pdf).
If it is random noise on the input, it would be like many of the SSL methods, e.g. DINO (https://arxiv.org/abs/2104.14294), right?
phonebucket · 2023-01-26
Alex Graves (now at DeepMind) put a similar demo on his University of Toronto page a long time back: https://www.cs.toronto.edu/~graves/handwriting.html
Wistar · 2021-08-11
Maybe something more along the lines of this AI handwriting generator?
https://www.cs.toronto.edu/~graves/handwriting.html
GirkovArpa · 2021-07-22
Hey HN — This is a simple web app that allows you to draw an alphabet, type some text, and then have the text written in your own handwriting. I’m happy to hear feedback or answer questions about how it works!
P.S. Drawing the entire alphabet isn't needed, only the letters you need.
Compare to:
https://www.calligrapher.ai/
https://www.cs.toronto.edu/~graves/handwriting.html
alpineidyll3 · 2020-10-10
This was one of graves' cool demos back when he was with hinton.
https://www.cs.toronto.edu/~graves/handwriting.html
blcArmadillo · 2019-03-13
I'm guessing Connectionist Temporal Classification: https://www.cs.toronto.edu/~graves/icml_2006.pdf
dharma1 · 2016-06-12
Reminds me of the guy in Her. I guess this could be automated with a neural network - http://www.cs.toronto.edu/~graves/handwriting.html
vixsomnis · 2015-07-24
And here's the actual site where you can try it out: http://www.cs.toronto.edu/~graves/handwriting.cgi?text=
poizan42 · 2015-07-23
I'm slightly alarmed by urls such as: http://www.cs.toronto.edu/~graves/handwriting.cgi?text=Hej&style=.…
Relative paths allowed from user input is usually a HUGE warning sign. Are you sure I can't make it open arbitrary files? What happens to your cgi script if it reads a file in the wrong format?
detaro · 2015-07-23
Links to http://www.cs.toronto.edu/~graves/handwriting.html, which was discussed here: https://news.ycombinator.com/item?id=9931041
darkmighty · 2015-05-12
I was very impressed with this handwriting demo (and the accompaning paper):
http://www.cs.toronto.edu/~graves/handwriting.html
Using a recurrent architecture.
To be honest I haven't played with NNs, but it puzzles me as to why the non-recurrent approach is so prevalent for complex tasks. I mean, it's the basic combinatorial circuit vs sequential circuits, which we all know are much more suited for complex or large outputs. Where's everything we learned from synchronous logic synthesis?