​​Deep Reinforcement Learning for de-novo Drug Design

- 2 networks trained separately:
- generative produces chemically feasible molecule reps
- predictive forecasts desired properties.
- then both trained jointly with the #RL

Link: https://github.com/isayev/ReLeaSE

#PyTorch

#StyleGAN trained on art perfomance.

#GAN

​​Large-Scale Object Mining for Object Discovery from Unlabeled Video

Paper about process of object discovery.

Link: https://arxiv.org/abs/1903.00362

#Video #DL #CV

Really short and brief, yet useful #tutorial on #PyTorch #transferlearning

Transfer learning is a concept of using network trained for a certain task for another one.

Link: https://medium.com/@iamHarin17/transfer-learning-in-pytorch-f7736598b1ed

#DL #novice #entrylevel #beginner

Time series basics

Time series — data, with points having timestamps. Some might think that #timeseries are mostly used in algorithmic trading, but they often used in malware detection, network data analysis or any other field, dealing with some flow of time-labeled data. These two resources provide deep and easy #introduction into #TS analysis.

Github: https://github.com/akshaykapoor347/Time-series-modeling-basics
Data Camp presentation: https://s3.amazonaws.com/assets.datacamp.com/production/course_5702/slides/chapter3.pdf

#beginner #novice #python #entrylevel

​​Open AI releasing MMO.

Spoiler: it is not MMORPG. It is Massively Multiagent Mame environment for reinforcement learning agents. It will allow to develop something what for #trueAI will be like an amoeba to human. But it’s live now.

Link: https://blog.openai.com/neural-mmo/
Github: https://github.com/openai/neural-mmo
3DClient github: https://github.com/jsuarez5341/neural-mmo-client

#OpenAI

Modern Deep Learning Techniques Applied to Natural Language Processing

Online collaborative book

Link: https://nlpoverview.com/index.html
Github: https://github.com/omarsar/nlp_overview
ArXiV: https://arxiv.org/abs/1708.02709

#NLP #beginner #novice #entrylevel #DL

Call for applications for Machine Learning Summer School

The Machine Learning Summer School will take place at #Skoltech, Moscow (August 26 - September 6 2019) with leading figures such as Yarin Gal, Arthur Gretton, Shimon Whiteson, Mark Girolami, Justin Solomon & others.
Applications open until the 6th of May.

Link: http://mlss2019.skoltech.ru
From: @powerofdata

#deeplearning #summerschool #education #mlss #DL

🇫🇷From the subscribers living (or being) in Paris:

On Saturday, 9th of March, in Paris, come to weekly DS Breakfast at 10h30 at Malongo Cafe, 50 Rue Saint-André des Arts, 75006 Paris.

​​Exploring Neural Networks with Activation Atlases

Amazing interactive article on feature visualizations, letting us see through the eyes of the neural network. The hidden layers of neural networks are quite fun to inspect.

Interactive website: https://distill.pub/2019/activation-atlas/

#CV #DL #visualization

Introducing Activation Atlases by #OpenAI

OpenAI in collaboration with #Google created activation atlases, a new technique for visualizing what interactions between neurons can represent.

Link: https://blog.openai.com/introducing-activation-atlases/
Direct demo link: https://distill.pub/2019/activation-atlas/app.html
Github: https://github.com/tensorflow/lucid/#activation-atlas-notebooks

#Google has open-sourced #FederatedLearning code

Step-by-step #tutorial showing how to perform Federated Learning using the same infrastructure Google
uses on 10s of millions of smartphones.

Link: https://medium.com/tensorflow/introducing-tensorflow-federated-a4147aa20041

​​Brilliant thread on ROC curve usage

ROC curve is used to measure how good is the discrimination between two distributions. This is a nice thread to refresh your memory, or to finally understand how ROCAUC works.

Link: https://threadreaderapp.com/thread/1104134423673479169.html
An introduction to prediction research: http://www.cecilejanssens.org/wp-content/uploads/2018/01/PredictionManual2.0.pdf

#ROC #AUC

Lessons learned building natural language processing systems in health care

It’s suprising, but #NLP in #healthcare doesn’t work the same way, as a researcher might expect, because of semantic.
Lesson #1: Off-the-shelf NLP models don’t work
Lesson #2: Build trainable NLP pipelines
Lesson #3: Start with labeling ground truth

Link: https://www.oreilly.com/ideas/lessons-learned-building-natural-language-processing-systems-in-health-care

​​Website using Deep Learning to colorize pictures.

Link: https://colourise.sg/#colorize

#DL #CV #demo

Another major release of #SpaCy

👽 ULMFit/BERT/ELMo-like language model pretraining
📐 Better matching & rule-based NER
🧮 Platform-independent matrix multiplication
🚀 2-3 times faster tokenization
⏳ Installation without local compilers

Link: https://explosion.ai/blog/spacy-v2-1

#NLP #python

​​A deep learning framework for nucleus segmentation using image style transfer

One of the challenges in applying DL to tissue and cell analysis (which can be used to, but not which usage is not limited to, cancer diagnostics) is boosting annoted training sets. This paper may help with the matter.

Link: https://www.biorxiv.org/content/10.1101/580605v1
#deeplearning #microscopy

​​Neural network that turns sketches into realistic photo.

Paper is called «Semantic Image Synthesis with Spatially-Adaptive Normalization».

#CVPR19 oral paper on a new conditional normalization layer for semantic image synthesis #SPADE and its demo app #GauGAN

ArXiV: https://arxiv.org/abs/1903.07291
Website: https://nvlabs.github.io/SPADE/

#GAN #CV #DL

​​Important article in Nature about statistical significance

Scientists rise up against statistical significance — about motion to move from widely using and quoting statistical significance to confindence intervals.

Link: https://www.nature.com/articles/d41586-019-00857-9

#statistics #statsignificance #nature #science

Next-level learning approach: using MRI to peek into baby brains to improve CV

MRI Scanning 17 babies for 26 hours to see how face-recognizing brain regions mature. When just 4-6 months, babies prefer to look at faces & socially relevant things. This means face recognition is learned via evolution: data-hungry & sample-inefficient.

Link: https://www.nature.com/articles/ncomms13995

#nature