Profile cover photo
Profile photo
Ian Goodfellow
1,962 followers
1,962 followers
About
Posts

I’m speaking at the GTC 2016
April 4-7 2016, Silicon Valley
Register with code FF16S20 to save up to $300. 
Add a comment...

Post has attachment
Christian and I have mostly studied adversarial examples from the point of view that they indicate areas for improvement in machine learning systems, and largely left the security implications unexplored. Nicolas Papernot is now tackling the security issues. He just finished demonstrating that he can control the output of a remotely hosted machine learning model by exploiting the cross-model generalization property of adversarial examples, even if he doesn't know what kind of model the remote host is using. It works in a correctly blinded experiment where he really doesn't know what model the target is using. It also works against non-deep machine learning models, like nearest neighbors. http://arxiv.org/abs/1602.02697
Add a comment...

Post has attachment

Post has shared content
Many more cool results with GANs!
After the release of DCGAN (and code), which is a neural network that can generate natural images (collaboration between Alec Radford, Luke Metz and me), people from the community have generated their own manga characters, flowers and now their own letters from reading chinese books. Fun stuff. All these generations are the hallucinations of a neural network and are not real.

Our paper: http://arxiv.org/abs/1511.06434
Our code:
https://github.com/Newmu/dcgan_code (Theano)
https://github.com/soumith/dcgan.torch (Torch)
https://github.com/mattya/chainer-DCGAN (manga)
http://genekogan.com/works/a-book-from-the-sky.html (chinese characters)
https://twitter.com/vintermann/status/675599478494208000 (flowers)
Looking for more fun stuff from the community.
PhotoPhotoPhotoPhotoPhoto
12/15/15
6 Photos - View album
Add a comment...

Post has attachment
Very cool new results on image generation using GANs from +Alec Radford , Luke Metz, and +Soumith Chintala .

Their latest GANs can do visual analogies like MAN WITH GLASSES - MAN + WOMAN = WOMAN WITH GLASSES.

https://github.com/Newmu/dcgan_code
http://arxiv.org/abs/1511.06434
Add a comment...

Post has attachment
I had the pleasure of hosting +Tianqi Chen  during his internship at Google Brain this summer. Tianqi's intern project was to develop a technique called Net2Net for rapidly transferring knowledge from one neural network to another. This can accelerate the training of a very large neural network and can reduce the amount of time spent retraining models from scratch while experimenting with many different model architectures. Together with +Jon Shlens we wrote a paper about Net2Net that we have submitted to ICLR 2016: http://arxiv.org/abs/1511.05641
Add a comment...

Post has shared content
Very proud to be open-sourcing TensorFlow, Google's newest Deep Learning framework! TensorFlow is both a production-grade C++ backend, which runs on Intel CPUs, NVidia GPUs, Android, iOS and OSX, and a very simple and research-friendly Python front-end that interfaces with Numpy, iPython Notebooks, and all the familiar Python-based scientific tooling that we love. TensorFlow is what we use every day in the Google Brain team, and while it's still very early days and there are a ton of rough edges to be ironed out, I'm excited about the opportunity to build a community of researchers, developers and infrastructure providers around it. Try it out!
Add a comment...

My slides from my talk "Practical Methodology for Deploying Machine Learning" at the Learn AI with the Best conference today: https://drive.google.com/file/d/0B64011x02sIkVGZlRk1iand0YkU/view?usp=sharing
Add a comment...

Post has attachment
Add a comment...

Post has attachment
Textbook feedback request:
https://drive.google.com/file/d/0B64011x02sIkbHNXQ0VrWHVGNTg/view?usp=sharing

Is Figure 13.2 useful? If not, should we cut it, or should we improve it in some way?
Add a comment...
Wait while more posts are being loaded