​​Dimensionality reduction for visualizing single-cell data using UMAP

UMAP is an t-SNE replacement for #visualization.

UMAP is being increasingly accepted as a powerful tool for visualizing single cell datasets. This paper compares UMAP to #TSNE

While UMAP is unquestionably better than default t-SNE in preserving global structure, it's worth mentioning that (very recently) it was shown that this limitation of t-SNE appears to be addressable with better parameters/initialization.

Article link: https://www.nature.com/articles/nbt.4314

​​Papers from #DeepMind panel at #NIPS2018

Work on radiotherapy planning: https://arxiv.org/abs/1809.04430
Triaging eye diseases: https://www.nature.com/articles/s41591-018-0107-6
Probabilistic U-net: https://arxiv.org/abs/1806.05034

#segmentation #CV #Unet

FAIR turns five: What we’ve accomplished and where we’re headed

#Facebook AI Research group report on the work done.

Link: https://code.fb.com/ai-research/fair-fifth-anniversary/

​​Probabilistic Image Segmentation

Re-implementation of the model described in «A Probabilistic U-Net for Segmentation of Ambiguous Images».

ArXiV: https://arxiv.org/abs/1806.05034
Github: https://github.com/SimonKohl/probabilistic_unet

Imitation by watching YouTube

Learning features from YouTube videos through self-supervision allows us to solve hard exploration games in Atari.

Paper: https://papers.nips.cc/paper/7557-playing-hard-exploration-games-by-watching-youtube.pdf
Youtube: https://www.youtube.com/watch?v=s8ZSVfYmtpc&feature=youtu.be

#RL #YouTube

Forensic Deep Learning: Kaggle Camera Model Identification Challenge

Report on Kaggle solution on camera model identification.

Link: https://towardsdatascience.com/forensic-deep-learning-kaggle-camera-model-identification-challenge-f6a3892561bd

#CV #Kaggle

​​Visualizing the Loss Landscape of Neural Nets

Github: https://github.com/tomgoldstein/loss-landscape

#NN #loss #vizualization #DL

Do Better ImageNet Models Transfer Better?

Finding: better ImageNet architectures tend to work better on other datasets too. Surprise: pretraining on ImageNet dataset sometimes doesn't help very much.

ArXiV: https://arxiv.org/abs/1805.08974

#ImageNet #finetuning #transferlearning

How Many Samples are Needed to Learn a Convolutional Neural Network

Article questioning fact that CNNs use a more compact representation than the Fully-connected Neural Network (FNN) and thus require fewer training samples to accurately estimate their parameters.

ArXiV: https://arxiv.org/abs/1805.07883

#CNN #nn

​​Deep learning for chest X-rays

Important work on chest X-Ray analysis.

ArXiV: https://arxiv.org/abs/1711.05225

#DL #medical #bioinformatics

Critics on the last article, suggesting that data was not so good and the advance not that significant as claimed.

Link: https://lukeoakdenrayner.wordpress.com/2018/01/24/chexnet-an-in-depth-review/

​​UberAI introduces a new approach for making Neural Networks process images faster & more accurately with jpeg representations.

Link: https://eng.uber.com/neural-networks-jpeg/
Paper: https://papers.nips.cc/paper/7649-faster-neural-networks-straight-from-jpeg

#nn #CV #Uber

​​Stunning face generation results in “A Style-Based Generator Architecture for Generative Adversarial Networks”

ArXiV: https://arxiv.org/pdf/1812.04948.pdf

#CV #nn #GAN

Yeah, these are not real people.

New book by Andrew Ng

Drawn from his experience leading Google Brain, Baidu's AI Group, and Landing AI, this 5-step Playbook provides a roadmap for your company to transform into a great AI company.

Site: https://landing.ai/ai-transformation-playbook
Direct link: https://d6hi0znd7umn4.cloudfront.net/content/uploads/2018/12/AI-Transformation-Playbook.pdf

#book

​​Facebook has released #PyText — new framework on top of #PyTorch.

This framework is build to make it easier for developers to build #NLP models.

Link: https://code.fb.com/ai-research/pytext-open-source-nlp-framework/

One of the four best #NIPS2018 papers on new #ODEnet architecture, which can be used in #healthcare for predicting patient health.

Link: https://www.technologyreview.com/s/612561/a-radical-new-neural-network-design-could-overcome-big-challenges-in-ai/