|April 11 · Issue #80 · View online |
We’re back with a new set of links covering the latest news in deep learning.
Happy reading and hacking!
If you like receiving this newsletter and would like to support our work, you can do so by sharing this issue with friends and colleagues who might find it interesting. Thanks!
| China Now Has the Most Valuable AI Startup in the World |
The latest investment round of the SenseTime Group created some impressive numbers. The use case of mass surveillance for authorities based on live camera feeds is a little worrying though.
| NVIDIA And ARM Partnership To Bring Deep Learning Technology To IoT Devices |
Nvidia and ARM have teamed up to bring deep learning to your smartphone and other edge devices.
| Can a "Google AI" Build Your Genome Sequence? |
The latest article from a new publication
focused on questioning hype and panic in AI: A new AI-powered tool from Google promises more-accurate genome sequences, but its impact on genomics research remains to be seen.
| Deep Learning Studio is Better More Powerful Than Ever (sponsored) |
Open, Free and No-coding deep learning platform got an upgrade. Developers, researchers and students love our platform. Register today (It’s Free)!
| Infinia ML - Machine Learning Solutions for Business |
Get to know Infinia ML, a team of machine learning experts focused on delivering business impact. They help companies solve some of their toughest data challenges and unlock their biggest data opportunities.
| Lessons Learned Reproducing a Deep Reinforcement Learning Paper |
This extremely extensive and well-written article describes the author’s journey into reproducing a recent paper on deep reinforcement learning. The image below demonstrates the main issue quite well and the article covers all the minor issues along the way. Definitely a recommended read!
| The 1cycle policy |
Very nice article on tuning hyperparameters using the 1cycle policy. The author achieved the results of a paper using almost a 10th of the original training time.
| How to (quickly) build a deep learning image dataset |
A great new series on building your own Pokedex (an automatic tool for categorizing Pokemon) using deep learning. Includes building a dataset (this article), training a neural network and deployment to a mobile device.
| Caffe2 Merges With PyTorch |
Facebook seems to slim down their machine learning framework portfolio and has started merging Caffe2 into Pytorch. We’ll see where this goes, but it will hopefully ease the usual conversion pain.
| Secure Computations as Dataflow Programs |
Although very advanced, both in terms of TensorFlow and cryptography knowledge, this article inspects how to do optimised machine learning on encrypted data.
| Reverse Engineering Core ML |
Staying on the security topic, the author asks the question, how safe his deployed CoreML models are and if other developers would be able to reconstruct the graph and weights he tediously trained.
| Data Scientist |
Deep NLP and Vision for Business at Infinia ML.
| Data Scientist Intern |
Deep Learning for Business at Infinia ML.
| Differentiable Plasticity: A New Method for Learning to Learn |
Uber has found a way to create ‘plastic neural networks’, which adjust neurons even after training has finished in order to better tackle upcoming tasks.
| Interoceptive robustness through environment-mediated morphological development |
The authors show how robustness can be achieved by evolving the geometry of soft robots, their control systems, and how their material properties develop in response to one particular interoceptive stimulus (engineering stress) during their lifetimes.
| Guiding Neural Machine Translation with Retrieved Translation Pieces |
In this paper, the authors propose a simple, fast, and effective method for recalling previously seen translation examples and incorporating them into the NMT decoding process.