Up to Speed on Deep Learning: October Update

Sharing some of the latest research, announcements, and resources on deep learning.

By Isaac Madan (email)

Continuing our series of deep learning updates, we pulled together some of the awesome resources that have emerged since our last post. In case you missed it, here are our past updates: August, July, June (part 1, part 2, part 3, part 4), May, April (part 1, part 2), March part 1, February, November, September part 2 & October part 1, September part 1, August (part 1,part 2), July (part 1, part 2), June, and the original set of 20+ resources we outlined in April 2016. As always, this list is not comprehensive, so let us know if there’s something we should add, or if you’re interested in discussing this area further.

Research & Announcements

PassGAN: A Deep Learning Approach for Password Guessing by Hitaj et al of Stevens Institute of Technology. Using a GAN to guess LinkedIn passwords based on a corpus of previously leaked passwords — they were able to crack 27% of the 143 million leaked passwords in their data set when combined with a traditional password guessing tool HashCat.

How far are we from solving the 2D & 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks) by Adrian Bulat and Georgios Tzimiropoulos of the University of Nottingham. A deep-dive into the application of deep learning to face alignment, or facial landmark localization — meaning identifying the geometric structure of faces in images. Original paper here.

Information Theory of Deep Learning by Naftali Tishby. Explanation of the information bottleneck, as an explanation as to why deep learning works. Tishby argues that deep neural networks learn according to a procedure called the “information bottleneck,” which he and two collaborators first described in purely theoretical terms in 1999. The idea is that a network rids noisy input data of extraneous details as if by squeezing the information through a bottleneck, retaining only the features most relevant to general concepts. Youtube video here.

Deep neural networks are more accurate than humans at detecting sexual orientation from facial images by Kosinski et al of Stanford. In a controversial new study that is setting the internet ablaze, researchers claim that artificial intelligence can be used to accurately detect someone’s sexual orientation. The research utilized deep neural networks to examine over 35,000 publicly-available dating site photos where sexual orientation was indicated. The preliminary research was published in the Journal of Personality and Social Psychology, authored by Michal Kosinski, a professor at Stanford University Graduate School of Business.

Learning to Optimize with Reinforcement Learning by Ke Li of UC Berkeley. A deep dive into optimization algorithms for machine learning. There is a paradox in the current [machine learning] paradigm: the algorithms that power machine learning are still designed manually. This raises a natural question: can we learn these algorithms instead? This could open up exciting possibilities: we could find new algorithms that perform better than manually designed algorithms, which could in turn improve learning capability.

Deep Learning Techniques for Music Generation — A Survey by Briot et al of Sorbonne. This book is a survey and an analysis of different ways of using deep learning (deep artificial neural networks) to generate musical content.

Fast automated analysis of strong gravitational lenses with convolutional neural networks by Hezaveh et al of Stanford. Application of deep learning in analyzing space imagery. Physicists at Stanford University have developed a new technique of using neural networks for analyzing gravitational lenses in distant space (Motherboard).

Resources, Tutorials & Data

Teachable Machine by Google. Fun, quick way to get introduced to machine learning. This experiment lets anyone explore how machine learning works, in a fun, hands-on way. You can teach a machine to using your camera, live in the browser — no coding required. You train a neural network locally on your device, without sending any images to a server. That’s how it responds so quickly to you.

HR Analytics: Using Machine Learning to Predict Employee Turnover by Matt Dancho. Tutorial & example of applying machine learning to solve a business problem. With advancements in machine learning (ML), we can now get both better predictive performance and better explanations of what critical features are linked to employee attrition. In this post, we’ll use two cutting edge techniques.

Medical Data for Machine Learning by Andrew Beam of Harvard. Curated list of publicly available medical data for machine learning.

Heart Disease Diagnosis with Deep Learning by Chuck-Hou Yee of Insight. Explanation of how to perform automatic segmentation of the right ventricle in cardiac MRI imagery via deep learning.

Realtime Driver Drowsiness Detection (Sleep Detection) by Taha Emara. Tutorial to do video eye monitoring via deep learning, which can be used to detect driver drowsiness.

Higher-Level APIs in TensorFlow by Peter Roelants of Onfido. How to use Estimator, Experiment and Dataset to train models.

Explained Simply: How DeepMind taught AI to play video games by Aman Agarwal. Comprehensive breakdown & simplified explanation of how AlphaGo works.


By Isaac Madan. Isaac is an investor at Venrock (email). If you’re interested in deep learning or there are resources I should share in a future newsletter, I’d love to hear from you.

Leave a Reply

Your email address will not be published. Required fields are marked *