|April 25 · Issue #82 · View online |
Hey and welcome to another week in deep learning!
This weeks issue is sponsored by Lambda Labs
, who provide specialized workstations and servers for deep learning. If you pull the trigger, mention DLWEEKLY for $400 USD discount on any Lambda Quad!
Happy reading and hacking!
| A.I. Researchers Are Making More Than $1 Million, Even at a Nonprofit |
Based on tax forms filed by OpenAI, the New York Times took a look at the A.I. job market and salary landscape. Considering that OpenAI can’t give away equity, the salaries are pretty impressive.
| Scientists plan huge European AI hub to compete with US |
In an attempt to stop the ‘brain drain’, young researchers leaving universities to work at the big companies, European scientists have created an open letter
. The letter suggests creating a European Lab for Learning & Intelligent Systems (ELLIS) to attract researchers.
| Google's New AI Head Is So Smart He Doesn't Need AI |
An interesting interview with Jeff Dean, Googles new AI lead, covering his career, upcoming challenges and the role of artificial intelligence in Googles roadmap.
| Lessons from My First Two Years of AI Research |
This collection of learnings, anecdotes, tips and experiences from two years in research is a great read if you’re interested in getting into the field or just want to know how where the inspiration for all those papers comes from. It even includes practical tips on how to efficiently explore new ideas.
| Executing gradient descent on the earth |
Inspired by the ‘lost in the mountains’ analogy used to explain gradient descent, Chris Foster decided to tackle the obvious question: How well does gradient descent perform on the earth’s surface?
| Generating image segmentation datasets with Unreal Engine 4 |
Harnessing 3D engines for data synthesis has been in our minds for a long time and Unity
has already started working on related features, but we haven’t seen any articles on how to actually do it yet. Jeff is here to help and shows us how to generate labelled data for image segmentation using UE4.
| Deep Learning and a New Programming Paradigm |
This article explains the concept of Automatic Differentiation, why it may change the way we write software in the future and covers the challenges and issues.
Are you looking to enhance your business with Machine Learning? ML Conference introduces you to the world of ML tools, programming languages and technologies. Meet the world’s Machine Learning and AI masterminds and join ML Conference 2018
| PyTorch 0.4.0 |
The latest release introduces Windows support, 24 distributions with cdf, variance etc., dtypes, zero-dimensional Tensors, Tensor-Variable merge,, faster distributed, perf and bug fixes and CuDNN 7.1.
| BeatGAN2.0 |
This project aims to explore the ability of the DCGAN architecture to generate complex audio patterns, namely drum beats. It’s well documented and fun to explore.
| Stealing Machine Learning Models via Prediction APIs |
The title probably already explains the idea: This paper collects techniques to ‘steal’ models that are exposed through a query interface without actually knowing anything about the models architecture.
| Natural and Effective Obfuscation by Head Inpainting |
Instead of blurring or removing heads, the authors propose ‘head inpainting’ to anonymize images and protect identities.
| Deep Probabilistic Programming Languages: A Qualitative Study |
Deep probabilistic programming languages try to combine the advantages of deep learning with those of probabilistic programming languages. This paper explains deep probabilistic programming languages and characterizes their current strengths and weaknesses.