|January 11 · Issue #23 · View online |
Hello and welcome to another issue of Deep Learning Weekly,
this week we have Deep Learning predictions for 2017, a new course on Reinforcement Learning, lots of educational articles on Deep Learning in NLP, practical tips for debugging your TensorFlow code and three brilliant academic papers. In other news, NVIDIA ramps up move into autonomous vehicle space through Audi partnership and mystery online Go champion is revealed to be an updated version of AlphaGo.
| AI & Deep Learning predictions for 2017 |
Maluuba’s research team share their perspectives on the trends, initiatives
and applications of AI that they think will be most transformative in 2017
| Nvidia shows self-driving supercomputer at CES, names it Xavier - Roadshow |
At CES 2017, Nvidia took the wraps off a new computer designed to enable self-driving cars, and announced partnerships with Audi, Bosch, ZF, Here and Zenrin.
| CS 294 Deep Reinforcement Learning, Spring 2017 |
Starting January 18 the lectures of this Berkeley course will be recorded and you can follow it on the course website. If you want to discuss this course with other learner join our slack study group
| king - man + woman is queen; but why? |
Word vectors are great at encoding semantic relationships between words. This post will give you a solid intuition why exactly that is so.
| Applying Neural Network and Local Laplace Filter Methods to Very High Resolution Satellite Imagery… |
Great application a CNNs to detect destroyed housing units
| Deep Text Correcter |
Learn how to train a LSTM in TensorFlow to automatically correct small grammatical errors in conversation English. The author presents an interesting method for generating input/output text data by randomly perturbing correct sentences (e.g. removing a preposition).
| A Practical Guide for Debugging Tensorflow Codes |
Bugs?! I don’t write bugs, but for all you poor souls out there who are pulling your hair out while debugging their TensorFlow code these slides are a godsend.
| Cognitive Computing, Deep Learning, AI & More - Educate Yourself Podcast |
Podcast with Rob High, IBM Fellow, Vice President, and CTO of IBM Watson discusses cognitive computing, deep learning, Watson and education and developer opportunities via Watson as SaaS.
| GitHub - AutoDiff DAG builder, built from scratch on top of numpy and C |
essence - AutoDiff DAG builder, built from scratch on top of numpy and C. Very nice and readable code that offers a great way to understand the internals of Deep Learning frameworks such as TensorFlow or Theano.
| DeepStack: Expert-Level Artificial Intelligence in No-Limit Poker |
This paper introduces DeepStack an algorithm for imperfect information settings such as poker. It becomes the first program to beat professional poker players in heads-up no-limit Texas hold'em.
| Two-Bit Networks for Deep Learning on Resource-Constrained Embedded Devices |
There is a growing need to run Deep Learning algorithm on computationally and memory constrained IoT devices. The authors present Two-Bit Networks (TBNs) for model compression of CNNs with edge weights constrained to (-2, -1, 1, 2), which can be encoded with two bits thereby drastically reducing memory usage and improving computational efficiency.
| Google Reveals Secret Test of AI Bot to Beat top Go Players |
Updated version of DeepMind’s AlphaGo program behind mystery online competitor “Master(P)” who has been causing quite a stir in the world of online Go defeating current world number one, 19-year-old Ke Jie.
| Reinforcement Learning Using Quantum Boltzmann Machines |
In this paper the authors investigate the intriguing question whether quantum annealers with select chip layouts can outperform classical computers in reinforcement learning tasks.