💉Using AI to help increase blood donations

This is a post describing how #Facebook uses #ML and #AI to detect intent or interest in blood donation and to match these people with local hospitals, increasing probability of people actually donating blood. Sounds like an awesome and inspiring application for AI.

Link: https://tech.fb.com/using-ai-to-help-increase-blood-donations/

IPython notebooks and git

#IPython or #Jupyter is one of the most popular tools in Data Science. It usage may questionable, but it is optimal for beginners and people who are making thier first steps. This article covers rare theme — keeping notebooks in git repository and optimizing collaboration using them. Main problem lies in technical information (like cell execution count), which is redundant and can be omitted, but still gets written to git in the default scenario.

Link: https://pascalbugnion.net/blog/ipython-notebooks-and-git.html
Code: https://gist.github.com/pbugnion/ea2797393033b54674af

#datascience #practicalML

​​PI-REC: Progressive Image Reconstruction Network With Edge and Color Domain

Paper on how image can be reconstracted from doodle and color palette.

Link: https://arxiv.org/pdf/1903.10146.pdf

#ImageReconstruction #CV #DL

​​Fast video object segmentation with Spatio-Temporal GANs

Spatio-Temporal GANs to the Video Object Segmentation task, allowing to run at 32 FPS without fine-tuning.

#FaSTGAN #GAN #Segmentation #videomining #CV #DL

​​Step Change Improvement in Molecular Property Prediction with PotentialNet

Paper on a significant improvement in ability to predict molecular properties in drug design. #ML algorithms are getting better and better than classical methods.

Link: https://medium.com/@pandelab/step-change-improvement-in-molecular-property-prediction-with-potentialnet-f431ffa32a2c

#drugsdesign #biolearning #healthcare

​​Open-sourcing PyTorch-BigGraph for faster embeddings of extremely large graphs

PyTorch-BigGraph— a tool that for faster and easier producing graph embeddings for extremely large graphs. Outputs high-quality embeddings without specialized computing resources like GPUs or huge amounts of memory.

Link: https://ai.facebook.com/blog/open-sourcing-pytorch-biggraph-for-faster-embeddings-of-extremely-large-graphs/
Github: https://github.com/facebookresearch/PyTorch-BigGraph

#PyTorch #Facebook #OpenSourceRelease #Embeddings #GraphLearning

Note on moving from being frequntist to being bayesian

Beautiful article about core differences and personal experience.

Link: http://www.fharrell.com/post/journey/

#InBayesWeTrust #BvsF

​​AI talent flow map

Very interesting map of the flow of students between countries on the 2019 Global AI Talent report.

Link: https://jfgagne.ai/talent-2019/

​​Deep Neural Networks Improve Radiologists’ Performance in Breast Cancer Screening

Paper claims model to perform better than human.

ArXiV: https://arxiv.org/abs/1903.08297
GitHub (code & models): https://github.com/nyukat/breast_cancer_classifier

#BreastCaner #healthcare #DL #CV

​​🔥DeepMind released dataset for training algebraic perception

Mathematics Dataset is a large-scale extendable dataset of mathematical questions, for training (and evaluating the abilities of) neural models that can solve algebraic problems (reason algebraically).

ArXiV: https://arxiv.org/abs/1904.01557
GitHub: https://github.com/deepmind/mathematics_dataset

​​🤖Handl: New dataset labeling tool release

Handl is a tool to label and manage data for machine learning. It employs 25k qualified crowdworkers who help tech companies to deal with data preparation and get paid for it. Consensus algorithm ensures the quality of labeling for any type of data — images, texts, and sounds.
#Handl was released today at Product Hunt, so developers might benefit from community upvotes, please consider supporting such useful tool on Product Hunt.


Link: https://handl.ai
Product Hunt url: https://www.producthunt.com/posts/handl-3

#handl #machinelearning #ai #data #datalabeling

Scalable Muscle-actuated Human Simulation and Control

Spectacular work from Seoul National University on motion-mimicing DeepRL with a comprehensive full-body musculoskeletal model with 346 muscles, including predictive simulations of post-operative gaits.

YouTube: https://www.youtube.com/watch?v=a3jfyJ9JVeM
Project link: http://mrl.snu.ac.kr/research/ProjectScalable/Page.htm

#SIGGRAPH2019 #RL

🎉🔥15000!!! Thank you all, dear scientists!

Make Trump Sing Again

Generated by a Trump TTS model trained based off the paper "Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis", where given a reference audio the model will try to replicate that style.

ArXiV: https://arxiv.org/pdf/1803.09017.pdf
Youtube: https://youtu.be/3rgAVT8b4fw

#tts #song #speech #DL

DGC-Net: Dense Geometric Correspondence network

Paper addresses the challenge of dense pixel correspondence estimation between two images. Practically, this means that it is about comparing different views of one object, which is very important to make #CV more robust.

ArXiV: https://arxiv.org/abs/1810.08393
Github: https://github.com/AaltoVision/DGC-Net
YouTube: https://www.youtube.com/watch?v=xnQMEr4FbHE
Project page: https://aaltovision.github.io/dgc-net-site/

#CNN #DL

Automatic feature selection:

EGU is still ongoing, but I am burning to share with you some of my findings already!

Research group in University of Lausanne developed a pretty promising algorithm for automatic feature selection based on General Regression Neural Network (GRNN, also known as Nadaraya-Watson Estimator). The idea behind is pretty simple and therefor powerful - why won't we build the simplest network that can train really fast and brute force all possible combination of features to check how they affect accuracy, learning rate etc and than select the best performing once.

Here is Python implementation on GitHub: https://github.com/federhub/pyGRNN
Also, check their poster: https://github.com/federhub/pyGRNN/blob/master/EGU2019_FS_using_simple_and_efficient_ML_models.pdf

Stay tuned, subscribe and share!
xoxo

​​Autodesk claims to use GANs to design a chair.

First-of-its-kind chair from Philippe Starck and Kartell. Imagined by a human and cocreated with intelligent generative design

Link: https://adsknews.autodesk.com/news/starck-intelligent-generative-design
Generative design explanation link: https://www.autodesk.com/solutions/generative-design

#Autodesk #generativedesign

​​Google’s progress on AutoML

Hint: it’s beating some old competition solutions.

Link: https://cloud.google.com/blog/products/ai-machine-learning/expanding-google-cloud-ai-to-make-it-easier-for-developers-to-build-and-deploy-ai

#AutoML #Kaggle

​​How is Uber predicting demand, surge and where will be high demand area.

One more post from brilliant #Uber engineering team, sharing their approach and general experience about forecasting.

Link: https://eng.uber.com/forecasting-introduction/

#ts #timeseries #arima #demandprediction #ml

General talk on who makes the choice now: machine or human

Discussion based on the book «A Human’s Guide to Machine Intelligence: How Algorithms Are Shaping Our Lives and How We Can Stay in Control».

Link: https://knowledge.wharton.upenn.edu/article/algorithms-decision-making/

#podcast #general #meta #publicml