Google’s project started in 2009, long before carmakers and most other companies seriously considered the technology, but Google has yet to launch an autonomous vehicle service for the public.

Xuedong Huang, the company’s chief speech scientist, reports that in a recent benchmark evaluation against the industry standard Switchboard speech recognition task, Microsoft researchers achieved a word error rate (WER) of 6.3 percent, the lowest in the industry.

A list of domains where recent Deep Learning research has made a big impact.

With new neural network architectures popping up every now and then, it’s hard to keep track of them all. Knowing all the abbreviations being thrown around (DCIGN, BiLSTM, DCGAN, anyone?) can be a bit overwhelming at first. Check out this cheat sheet.

Twitter Cortex built a framework for RL development. and is open sourcing torch-twrl to the world. torch-twrl aims to provide 1. A RL framework in Lua/Torch with minimal dependencies 2. Rapid development with well defined, modular code 3. Seamless integration with OpenAI’s RL benchmark framework Gym.

NVIDIA CEO Jen-Hsun Huang has announced the next generation of NVIDIA’s neural network inferencing cards, the Tesla P40 and Tesla P4. These cards are the direct successor to the current Tesla M40 and M4 products, and with the addition of the Pascal architecture, NVIDIA is promising a major leap in inferencing performance.

Official code for the Exploring the Limits of Language Modeling paper that describes a hybrid between character CNN, a large and deep LSTM, and a specific Softmax architecture to achieve a new state-of-the-art.

The Cars Overhead With Context (COWC) data set is a large set of annotated cars from overhead. It is useful for training a device such as a deep neural network to learn to detect and/or count cars. More information in the paper.

Several open source implementations of  DeepMind’s text-to-speech breakthrough model WaveNet have been popping up. Here are two: One in Tensorflow, and one using Keras.

A simple end-to-end model for speech recognition, combining a convolutional network based acoustic model and a graph decoding. By Ronan Collobert, Christian Puhrsch, Gabriel Synnaeve (Facebook AI Research)

The paper presents super-resolution generative adversarial network (SRGAN), capable of recovering photo-realistic natural images from 4 times downsampling by using perceptual loss function which consists of an adversarial loss and a content loss. (Twitter Research)

Scenarios from the real-time strategy game StarCraft as new benchmarks for reinforcement learning algorithms. The paper proposes micromanagement tasks, which present the problem of the short-term, low-level control of army members during a battle. By Nicolas Usunier, Gabriel Synnaeve, Zeming Lin, Soumith Chintala (Facebook AI Research)

A recently discovered method called feedback-alignment shows that the weights used for propagating the error backward don’t have to be symmetric with the weights used for propagation the activation forward. In this work, the feedback alignment principle is used for training hidden layers more independently from the rest of the network, and from a zero initial condition. By Arild Nøkland.