ODS breakfast in Paris! See you this Saturday at 10:30 at Malongo Café, 50 Rue Saint-André des Arts.

​​TabNine showed deep learning code autocomplete tool based on GPT-2 architecture.

Video demonstrates the concept. Hopefully, it will allow us to write code with less bugs, not more.

Link: https://tabnine.com/blog/deep
Something relatively similar by Microsoft: https://visualstudio.microsoft.com/ru/services/intellicode

#GPT2 #TabNine #autocomplete #product #NLP #NLU #codegeneration

Great collection of practical rules for routine DS engineering / research job.

Machine Learning in a company is 10% Data Science & 90% other challenges, this pdf provides a great deal of principals and solutions to deal with them.

We can only recommend saving this post to your Saved Messages by forwarding it to yourself.

Link: http://martin.zinkevich.org/rules_of_ml/rules_of_ml.pdf

#cheatsheet #advice #practical #common #shouldbesaved

​​YouTokenToMe, new tool for text tokenisation from VK team

Meet new enhanced tokenisation tool on steroids. Works 7-10 times faster alphabetic languages and 40 to 50 times faster on logographic languages, than alternatives.

Under the hood (watch source) there is C++ implementation with python bindings, using Byte Pair Encoding (BPE) algorithm. YouTokenToMe beats #SentencePiece by Google and #fastBPE, created by a researcher from Facebook AI Research in terms of speed.

Github: https://github.com/vkcom/YouTokenToMe
Medium: https://medium.com/@vktech/youtokentome-a-tool-for-quick-text-tokenization-from-the-vk-team-aa6341215c5a
Byte Pair Encoding: https://arxiv.org/abs/1508.07909

This improvement for everyday used toolset deserves minimum 50 claps at Medium and a Star on github!
Let’s give VK research team appreciation from the community they deserve 👏!!

What’s wrong with transformer architecture: an overview

How the Transformers broke NLP leaderboards and why that can be bad for industry.

Link: https://hackingsemantics.xyz/2019/leaderboards/

#NLP #overview #transformer #BERT #XLNet

​​Simultaneous food and facial recognition at a Foxconn factory canteen, Shenzhen China

#video #foodlearning #facerecogniction #dl #cv #foxconn

​​Deep Learning Image Segmentation for Ecommerce Catalogue Visual Search

Microsoft’s article on image segmentation

Link: https://www.microsoft.com/developerblog/2018/04/18/deep-learning-image-segmentation-for-ecommerce-catalogue-visual-search/

#CV #DL #Segmentation #Microsoft

​​Google AI research on learning better simulation methods for partial differential equations

New research shows how machine learning can improve high-performance computing for solving partial differential equations, with potential applications that range from modeling #climatechange to simulating fusion reactions. Learn all about it here

Link: https://ai.googleblog.com/2019/07/learning-better-simulation-methods-for.html

#PDE #DE #GoogleAI

On the concept of 'intellectual debt'

There is technical debt — when you know you should rewrite some stuff, or implement some features, but they don't seem critical at the moment. So article introduces a concept of 'intellectual debt', which resies with more broad and common use of #MachineLearning and #DeepLearning (specially, the latter). What happens when AI gives us seemingly correct answers that we wouldn't have thought of ourselves, without any theory to explain them?

Link: https://www.newyorker.com/tech/annals-of-technology/the-hidden-costs-of-automated-thinking

#Meta #common #lyrics

​​New dataset with adversarial examples

Natural Adversarial Examples are real-world and unmodified examples which cause classifiers to be consistently confused. The new dataset has 7,500 images, which we personally labeled over several months.

ArXiV: https://arxiv.org/abs/1907.07174
Dataset and code: https://github.com/hendrycks/natural-adv-examples

#Dataset #Adversarial

​​Release of 27 pretrained models for NLP / NLU for PyTorch

Hugging Face open sources a new library that contains up to 27 pretrained models to conduct state-of-the-art NLP/NLU tasks.

Link: https://medium.com/dair-ai/pytorch-transformers-for-state-of-the-art-nlp-3348911ffa5b

#SOTA #NLP #NLU #PyTorch #opensource

ODS breakfast in Paris! See you this Saturday at 10:30 at Malongo Café, 50 Rue Saint-André des Arts.

Filter autoselect in VSCO by Google

#VSCO used #TensorFlow Lite to develop the 'For This Photo' feature, which uses on-device ML to suggest photo filter presets from a curated list.

YouTube: https://www.youtube.com/watch?v=fHbjfeitIvE
Link: https://medium.com/tensorflow/suggesting-presets-for-images-building-for-this-photo-at-vsco-9b94041c4ba4

#mobile #device #cv #dl

Baidu's recent paper: Hubless Nearest Neighbor Search

Hubless Nearest Neighbor Search, a new method for Bilingual Lexicon Induction, improves retrieval accuracy significantly. Empirical results show HNN outperforms NN, ISF and other state-of-the-art.

Github: https://github.com/baidu-research/HNN
Paper: https://github.com/baidu-research/HNN/blob/master/doc/HNN.pdf

#ACL2019 #NLP #NLU

​​Plato Research Dialogue System: A Flexible Conversational AI Platform

The Plato Research Dialogue System is a platform #Uber developed to enable experts and non-experts alike to quickly build, train, and deploy conversational AI agents.

Link: https://eng.uber.com/plato-research-dialogue-system/

#ConversationalAI #converstaion #NLP #NLU

​​Model for tweaking graph visualization layout parameters

New #MachineLearning model builds a WYSIWYG interface to intuitively produce a layout you want!

Demo: http://kwonoh.net/dgl
Paper: http://arxiv.org/abs/1904.12225

#Visualization #ML

​​GSCNN: video segmetation architecture

Semantic segmentation GSCNN significantly outperforms DeepLabV3+ on Cityscapes benchmark.

Paper: https://arxiv.org/abs/1907.05740
Github (Project): https://github.com/nv-tlabs/GSCNN

#DL #CV #NVidiaAI #Nvidia #autonomous #selfdriving #car #RL #segmentation

​​Large Scale Adversarial Representation Learning

DeepMind shows that GANs can be harnessed for unsupervised representation learning, with state-of-the-art results on ImageNet. Reconstructions, as shown in paper, tend to emphasise high-level semantics over pixel-level details.

Link: https://arxiv.org/abs/1907.02544

#DeepMind #GAN #CV #DL #SOTA

Overview of 10 Stanford's Data Science courses

A survivor’s guide to Artificial Intelligence courses at Stanford

Link: https://huyenchip.com/2018/03/30/guide-to-Artificial-Intelligence-Stanford.html

#Staford #MOOC #course #free #rating #learning

Top 8 trends from ICLR 2019

Overview of trends on #ICLR2019:
1. Inclusivity
2. Unsupervised representation learning & transfer learning
3. Retro ML
4. RNN is losing its luster with researchers
5. GANs are still going on strong
6. The lack of biologically inspired deep learning
7. Reinforcement learning is still the most popular topic by submissions
8. Most accepted papers will be quickly forgotten

Link: https://huyenchip.com/2019/05/12/top-8-trends-from-iclr-2019.html

#ICLR #overview