|September 28 · Issue #59 · View online |
Hi and welcome to another issue of Deep Learning Weekly!
We’re happy to have crossed the 6000 subscriber mark this week and want to thank all of you for your support! If you want to see us hit the 10k, share this newsletter with colleagues, friends, and family!
See you next week!
| NVIDIA Deep Learning Accelerator |
In an attempt to further push the development of specialized deep learning hardware, Nvidia has released an open-source architecture specification. This includes all necessary documents, definitions, and code and currently focuses on IoT devices and applications. An interesting move and we’ll see the results in future hardware.
| Jeff Dean’s Talk on Large-Scale Deep Learning |
Jeff Dean, Google Brain lead engineer, gave a general talk about the applications of machine and deep learning that Google Brain is currently working on. You can watch the full talk
or just get an overview using Pavel Surmenok summary in the linked post.
| Microsoft launches new machine learning tools |
Microsoft did just announce a new set of machine learning tools, based around their Azure cloud services. Furthermore, there are some new features in Visual Studio Code that should make working with TensorFlow, Theano, Keras etc. easier.
| Introducing faster GPUs for Google Compute Engine |
Google has once again updated the available GPUs in their Compute Engine. These finally include Nvidias P100s, which should allow training of even larger models on even larger datasets.
| Machine Learning Glossary |
This glossary of machine and deep learning terms seems like a really valuable resource when trying to understand new papers, tutorials or even actual code. Definitely worth a bookmark!
| Using categorical data in machine learning with Python |
This article explores the use of deep learning on categorical data and covers basic, as well as more advanced techniques. Take a look, if you have such data and don’t know how to make any use of them.
| Essential Cheat Sheets for Machine Learning and Deep Learning Engineers |
From the more general glossary created by Google to a very specialized, but extremely handy collection of cheat sheets from Kailash Ahirwar. These cover all major frameworks and concepts and should join the glossary in your bookmarks.
| baidu/mobile-deep-learning |
Baidu has released an extremely small and fast framework, focused on CNNs thats currently in use in their iOS apps. Although quite specialized this may come in handy for iOS app development, as it utilizes the iPhones GPU, has a minimal footprint and should yield great performance.
| Fabrik – A platform to build deep learning models online |
Fabrik (Previously IDE) looks like an interesting platform to build, visualize and train deep learning models using a collaborative platform.
| SerpentAI: Game Agent Framework |
This framework allows you to train intelligent agents using video games and features a plugin based system to allow a high flexibility. In theory you are able to train agents written in any framework using any video game you want, which sounds pretty exciting!
| CUDA Toolkit 9.0 Release Notes |
Version 9.0 of the CUDA toolkit has been released.
| The Consciousness Prior |
This quite mind-bending paper from Yoshua Bengio tries to bridge the gap between deep learning & symbolic AI, while taking into account the potential importance of language.
| The loss surface of deep and wide neural networks |
The authors of this paper try to verify some common assumptions about our constant search for the global minimum in the loss functions of neural networks.