|January 17 · Issue #24 · View online |
this weeks issue is bringing you a detailed explanation on how to recognize traffic lights and win 5000$, an extensive set of machine learning rules from Google, Pinterest’s latest post on their deep learning usage, great podcasts and the last 2016 in review article from the Google Brain team.
We also want to take this opportunity and thank all of you for helping us cross the 2000 subscriber mark this week! As always, if you want to help us grow this great community of Deep Learning enthusiasts, simply share this issue with friends and colleagues.
Thanks and see you next week!
| Applying deep learning to Related Pins |
Pinterest sheds some light on the use of deep learning in their related pin recommendation system. Which they obviously had to call Pin2Vec.
| Microsoft acquires deep learning startup Maluuba |
Another deep learning startup gets acquired by one of the major players. This time Microsoft bought a lot of NLP and deep learning expertise presumably to further improve Cortana.
| Recognizing Traffic Lights With Deep Learning |
A very detailed post on building the system that won the Nexar Traffic Light Recognition Challenge. Great explanations and s great read!
| Machine Intelligence, from University to Industry |
Great podcast with Cameron Schuler from the Alberta Machine Intelligence Institute (AMII) talking about the evolution of AI, the envelope-pushing research at the university of Alberta, reinforcement learning and the importance of independent research institutions which are not owned by a major tech company.
| Deep Learning for Self-Driving Cars |
This class is an introduction to the practice of deep learning through the applied theme of building a self-driving car. It’s open to beginners designed for those who are new to machine learning.
| How Deep Learning Will Reshape Our Cities by The AI Podcast |
Another podcast, but this time on the future of our cities and the role of deep learning.
| GitHub - poolio/unrolled_gan: Unrolled Generative Adversarial Networks |
A TensorFlow implementation of Unrolled Generative Adversarial Networks.
| Google Tensorflow chooses Keras |
According to Keras author Francois Chollet Google is choosing Keras as an abstraction layer for TensorFlow.
| GitHub - deepdrive/deepdrive-universe: Run self-driving car agents in GTAV Universe |
OpenAI open sourced their integration with Grand Theft Auto V, all you need to do now is purchase a copy of the game and you can start training your AI in the world of GTA.
| The Google Brain team — Looking Back on 2016 |
We are not done with reviews of 2016 yet, so here is another one, this time from the Google Brain team. They present their achievements in all major fields and give some hints what we can expect this year.
| Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning |
Tang et al. make a surprising find: a simple generalization of the classic count-based approach can reach near state-of-the-art performance on various high-dimensional and/or continuous deep RL benchmarks.
| Generating Focussed Molecule Libraries for Drug Discovery with Recurrent Neural Networks |
Fascinating paper that details the use of an LSTM to generate novel molecules targeted at drug discovery.
| Rules of Machine Learning |
Googles extensive best practices for building and deploying machine learning systems in production. Great tips and valuable hints when to stop or what to focus on when creating the latest and greatest machine learning applications.