27.23TB of research data in torrents! Includes dataset such as:
- Breast Cancer Cell Segmentation
- Liver Tumor Segmentation
- MRI Lesion Segmentation in Multiple Sclerosis
- Electron Microscopy, Hippocampus
- Digital Surface & Digital Terrain Model

And courses recordings, including:
- Introduction to Computer Science [CS50x] [Harvard] [2018]
- Artificial Intelligence(EDX)
- Richard Feynman's Lectures on Physics (The Messenger Lectures) (🔥)
- [Coursera] Machine Learning (Stanford University) (ml)
- [Coursera] Natural Language Processing (Stanford University) (nlp)
- [Coursera] Neural Networks for Machine Learning (University of Toronto) (neuralnets)

http://academictorrents.com/

#course #torrent #dataset

Kaggle kernel on classifing Russian Troll tweets.

https://www.kaggle.com/kmader/from-hate-speech-to-russian-bot-tweets

#kaggle #kernel #nlp

Teams at #DeepMind and #Moorfields have developed AI technology that can detect eye disease and prioritise patients. 'Clinically applicable deep learning for diagnosis and referral in retinal OCT' has been published online in #NatureMedicine today:
https://www.nature.com/articles/s41591-018-0107-6

#cv #dl

Fast image-to-image translation in the browser. With 3 trained models introduced. Plus a processed dataset of 1000 images for edges2cats translation.

Demo: https://zaidalyafeai.github.io/pix2pix/cats.html
Code: https://github.com/zaidalyafeai/zaidalyafeai.github.io/tree/master/pix2pix

#tf #tensorflow #tfjs #pix2pix #cv

«IEEE’s camera identification challenge — different approach to teaming up»

https://hackernoon.com/ieees-camera-identification-challenge-different-approach-to-teaming-up-28da44dfe635

#cv #dbrain

Neural nets are terrible at arithmetic & counting. If you train one in 1 to 10, it will do okay on 3 + 5 but fail miserably for 1000 + 3000. Resolving this, «Neural Arithmetic Logic Units» can track time, do arithmetic on images of numbers, & extrapolate, providing better results than other architectures.

https://arxiv.org/pdf/1808.00508.pdf

#nn #architecture #concept #deepmind #arithmetic

The Conversational Intelligence Challenge 2 (ConvAI second part) got announced today.

The aim of our competition is to establish a concrete scenario for testing chatbots that aim to engage humans, and become a standard evaluation tool in order to make such systems directly comparable.

The most promising team attending will receive a travel grant to attend #NIPS2018

#nlp #dl #dialoguesystem #competition

Stanford engineers have combined two types of computers to create a faster and less energy-intensive image processor for use in autonomous vehicles, security cameras and medical devices.

https://news.stanford.edu/2018/08/17/new-ai-camera-revolutionize-autonomous-vehicles/

Recent Advances for a Better Understanding of Deep Learning − Part I

https://towardsdatascience.com/recent-advances-for-a-better-understanding-of-deep-learning-part-i-5ce34d1cc914

#dl #theory

What is a Generative Adversarial Network?

Another article about how GANs work.

http://hunterheidenreich.com/blog/what-is-a-gan/

#gan #theory #dl

Curious About How To Be A Data Scientist? Hear From A Netflix Data Scientist

Article about how data science #production works. How problem should be defined and how project should be maintained and run.

https://towardsdatascience.com/a-peek-into-a-netflix-data-scientists-day-66bf3dacabb9

First 1e6 integers, represented as binary vectors indicating their prime factors, and laid out using the sparse matrix support in leland_mcinnes's UMAP dimensionality reduction algorithm. This is from a 1000000x78628 (!) binary matrix. Very pretty structure emerges.

"Learning Hierarchical Semantic Image Manipulation through Structured Representations":

https://arxiv.org/abs/1808.07535

#cv #dl

Online ad demand prediction #kaggle competition 1st place summary:
https://www.kaggle.com/c/avito-demand-prediction/discussion/59880

Winner explains how to combine categorical, numerical, image and text features into a single #NN that gets you into top 10 without stacking.

#PyTorch tensor basics: https://www.kdnuggets.com/2018/05/pytorch-tensor-basics.html

New model to naturally detect depression in conversations

Link: http://news.mit.edu/2018/neural-network-model-detect-depression-conversations-0830

#nlp #audio #dl

Tickets for one of the key conferences #NIPS2018 were sold out in 15 minutes.

NIPS Conference Registrations 2002 thru 2019.
[2018] War erupts for tickets
[2019] AI researchers discover time travel.

#NIPS2018

1st place solution in the recent Home Credit Default Risk #Kaggle competition

- extensive feature engineering, with ~700 of features total used
- XGBoost, LightGBM, CatBoost, FastRGF, DAE+NN, Lin Reg
- 3-level ensembling (stacking x2 + blending)

Link: https://www.kaggle.com/c/home-credit-default-risk/discussion/64821