|April 26 · Issue #38 · View online |
Hi and welcome to a new week in deep learning,
We hope you’ll enjoy reading as much as we did and would appreciate you sharing this newsletter with your friends and colleagues.
See you next week!
| Deep Learning for Photo Editing |
In this article by one of your trusty curators, an image editing startup shares the highs and lows of their deep learning journey and gives some insights on the different strategies, architectural decisions, and successes.
| NVIDIA Tesla P100 GPU: Be prepared for the AI revolution - Cloud computing news |
IBM has started to add Tesla P100 GPUs to their datacenters and offers them in their cloud solutions. So if you are in desperate desire of power for your training, you can now move on to P100 GPUs.
| Forget AI. The real revolution could be IA |
An interesting take on the current ‘AI will take over world’ trend, arguing that AI will be more used for augmenting, as it will probably not (yet) surpass humans. Great read and definitely something to think about.
| An API to copy the voice of anyone |
Style transfer on images has been done in a multitude of ways, so now we apparently move on to style transfer for speech. Just feed the model with a sample snippet and you’re able to generate new samples using your own voice.
| The Race To Build An AI Chip For Everything Just Got Real |
An interesting look at the AI chip market, that came to live when the machine learning hype took off. Googles Tensor Processing Unit is one of the more known examples, but still not available to the public and more and more companies are trying to make a fortune with chips specialized for deep learning workloads.
| AI Grant |
Building something AI related and want to release it under an MIT or Apache 2.0 license? Why not get 5000$ to do it? The deadline is in a few days, but you’ve still got a chance to enter this great contest.
| The End of Human Doctors |
The introduction to an upcoming series of articles, that will explore if deep learning is going to be a threat for today’s doctors, written by a radiologist.
| A Brief History of CNNs in Image Segmentation: From R-CNN to Mask R-CNN |
This article tells the history of CNNs in image segmentation and especially Facebook’s way to the Mask R-CNN that’s powering the impressive image understanding techniques they demonstrated last week. Well written and definitely worth a look.
| Alternatives to a Degree to Prove Yourself in Deep Learning |
Some thoughts and tips on how to get into deep learning and, more important, prove to others that you are experienced in the field afterward. Includes an extensive list of things to keep an eye on and may help you get a nice deep learning job.
| CMU CS: 15-883: Computational Models of Neural Systems |
A course that was held in 2015 and covers information processing in real neural systems from a computer science perspective. It focuses on well explored parts of the brain and models these using patterns from computer science.
| The GAN Zoo – Deep Hunt |
Confused about all the different GANs that have come up recently? Avinash Hindupur has got you covered with an overview of all named GAN variants in a single list.
| Failures of Deep Learning |
In this recent talk, Shai Shalev-Shwartz gives theoretical insights on some of the limitations in deep learning by explaining three families of problems that lead to failing algorithms.
| Forge: Neural Network toolkit for Metal |
Matthijs Hollemans announced a new framework, that is supposed to ease the implementation of neural networks using Metal Performance Shaders on iOS. The framework features a Keras like API, additional operations like depthwise convolution and sample implementations for LeNet-5, Inception-v3 and Googles recently published MobileNets.
| ritchieng/the-incredible-pytorch |
This curated list of tutorials, papers, projects, communities and more relating to PyTorch is a great starting point if you want to dive deeper into PyTorch, but great for getting an overview as well.
| deepmind/dnc |
A TensorFlow implementation of the Differentiable Neural Computer from DeepMind.
| Google's Cloud Vision API Is Not Robust To Noise |
In this paper, the authors try to evaluate the robustness of the Google Cloud Vision API and find that adding noise to images can lead to very different results. Their suggested countermeasure is adding a noise filter to the APIs inputs.