Deep Voice 3, GANFather, Test Tube, Sound and Meaning, Proven Beauty, Word Embeddings in 157…

Deep Voice 3, GANFather, Test Tube, Sound and Meaning, Proven Beauty, Word Embeddings in 157 Languages,…

On People…
Deep learning course offered by François Fleuret (includes videos, pdf, etc.) — Link

This article explains what’s the relationship between meaning and sound, and why people feel annoyed by words like “moist” — Link

MIT Technology Review: The GANfather: The man who’s given machines the gift of imagination — Link

On Education and Research…
Baidu presents Deep Voice 3, a fully-convolutional attention-based neural text-to-speech system, which achieved state-of-the-art neural speech synthesis and is order of magnitude faster than current systems — Link

4 approaches to natural language processing: distributional appraoches, frame-based approaches, model-theoretical approach, interactive learning, … Link

[Paper] Deep contextualized word representations — Link

CS224n: NLP with Deep Learning from Stanford — Link

On Code and Data…
Fasttext releases pre-trained word embeddings in 157 languages, a resource which is useful for those working on multilingual problems or research — Link

Keras implementation of the “One pixel attack for fooling deep neural networks” — Link

Test Tube is a library to track and optimize deep learning experiments — Link

On Industry…
Proven Beauty, a startup that uses NLP to provide a personalized line of skin care products to its customers — Link

Baidu’s new algorithm learns to clone a voice, and it only needs less than a minute of audio data from the speaker — Link

Australian Securities and Investments Commission (ASIC) wants to use NLP for enforcing and regulating company and financial service laws — Link

Worthy Mentions…
Ian Goodfellow’s new work on “Adverserial Examples that Fool both Human and Computer Vision” — Link

Notes on building a single shot multibox detector with PyTorch — Link

[Article] Malicious use of Artificial Intelligence — Link

Don’t forget to spread the love ?

Source: Deep Learning on Medium