|February 15 · Issue #28 · View online |
We just crossed 3000 subscribers and would like to thank you again for all of your support. As always, if you’d like to see our community grow, simply share this issue with friends and colleagues.
Happy reading, coding and thinking and see you next week!
| Machine Learning @Scale 2017 Recap |
Facebook hosted their machine learning at scale conference in New York and has now made videos of the talks public. Major companies share insights on how they deploy models and handle their research and development. Especially Joaquin Candelas demonstration of Facebooks machine learning backend is worth a look.
| AI learns to solve quantum state of many particles at once |
Deep learning is making its way into physics. Scientists at the ETH Zurich have started experimenting with neural nets to model quantum states.
| Applying machine learning to physics could be the way to build the first quantum computer |
And more quantum physics: An extensive overview of possible applications, already existing approaches and current research on the use of deep learning for complex physics problems.
| The Future of Artificial Intelligence |
An interesting article on the future of general artificial intelligence or ‘strong AI’ and its ethical and moral limitations. Salted with quotes on the dark future and the obligatory terminator image.
| General AI Challenge |
The General AI Challenge is made up of multiple rounds, each designed to tackle a crucial research problem in human-level AI development. GoodAI will give out $5mil in prize money over the following years. Here’s your chance to have an active hand in developing safe and beneficial general-purpose artificial intelligence.
| TensorFlow HowTo: A Universal Approximator Inside a Neural Net |
Morgan Giraud gives a gentle introduction to TensorFlow by building a universal approximator using the framework.
| Abusing Generative Adversarial Networks to Make 8-bit Pixel Art |
A detailed explanation of GANs and how to train them. And while learning about those networks you get to see pixel art that was generated by a neural network trained on Zelda and Super Mario levels.
| Generative Adversarial Networks (GANs) in 50 lines of code (PyTorch) |
If you are now eager to implement your own GAN take a look at this article, where Dev Nag shows you how to create a rather simple GAN using the new PyTorch library.
| Open Sourcing TensorFlowOnSpark |
Following CaffeOnSpark, which was released last year, Yahoo just announced TensorFlowOnSpark, a framework that enables distributed TensorFlow execution on Spark and Hadoop clusters. Converting existing TensorFlow code is supposed to take less than 10 lines of code.
| The AWS Deep Learning AMI, Now with Ubuntu |
Amazon has introduced an additional Ubuntu version of their deep learning AMI.
| Understanding Agent Cooperation |
DeepMind tries to model the behaviour of multiple agents in common scenarios similar to the prisoners dilemma. The idea is to study how the agents cooperate and if they work together or against each other. Features great videos of the agents behaviour over time.
| Comparative Study of CNN and RNN for Natural Language Processing |
This paper compares CNNs and RNNs and their different use cases in NLP.
| Attribute-Controlled Face Photo Synthesis from Simple Line Drawing |
Need some faces? Just draw a line face, implement the system in this paper and you are good to go.