#Deeplearning #Automation #Scheduling

A recent success of AI & Deep learning for multi-machine/robot scheduling problems!

Arxiv link https://arxiv.org/abs/1905.12204

Three issues are particularly important in this context: quality of the resulting decisions, scalability, and transferability.

Please check out the recent research which addressed those challenges! 96% optimality, transferable only with 1% loss in performance.

Visual exploration of neural nets - https://towardsdatascience.com/neural-networks-why-do-they-work-so-well-part-i-22f0d3690511

XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks by Mohammad Rastegari


https://videolectures.net/eccv2016_rastegari_neural_networks/?q=eccv%202016

Google AI - Release of Handbook Tutorials on Learning Keras and OpenCV

Hi everyone. I'm happy to let people know that we (Developer Relations at Google AI) are releasing handbooks and accompany presentations/code labs for learning Keras/OpenCV. The material is written for software engineers whom want a 'straight path with no math' to learning machine learning. The handbooks and code samples are free to download (licensed under CC-BY and Apache 2.0).

https://github.com/GoogleCloudPlatform/keras-idiomatic-programmer

This is a PyTorch implementation of "MixHop: Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing" (ICML 2019) that I made. MixHop has a state-of-the-art performance on several node classification benchmarks. In addition, the approximate version is pretty scalable. Enjoy!

https://github.com/benedekrozemberczki/MixHop-and-N-GCN

Nice new work from FAIR on assessing bias in datasets.

Turns out image recognition systems will recognize everyday objects more reliably if the picture was shot in developed countries than if it was shot in the developing world.

A critique of early approaches to visual object recognition was that they could not take context into account to help with the recognition. Now they do.....a bit too much.


https://ai.facebook.com/blog/new-way-to-assess-ai-bias-in-object-recognition-systems/

Search engine for computer vision datasets

[https://www.visualdata.io/](https://www.visualdata.io/)

DeepMDP: Learning Continuous Latent Space Models for Representation Learning
Gelada et al.: https://arxiv.org/abs/1906.02736
#ArtificialIntelligence #DeepLearning #ReinforcementLearning

1000x Faster Data Augmentation: New paper and code from Berkeley AI research.

From the announcement:

"Population Based Augmentation (PBA), an algorithm that quickly and efficiently learns a state-of-the-art approach to augmenting data for neural network training. PBA matches the previous best result on CIFAR and SVHN but uses one thousand times less compute, enabling researchers and practitioners to effectively learn new augmentation policies using a single workstation GPU. You can use PBA broadly to improve deep learning performance on image recognition tasks."

Announcement: https://lnkd.in/frFCS8R
Arxiv: https://lnkd.in/f58piHN
Code: https://lnkd.in/fQfWpkR

#neuralnetworks #deeplearning #artificialintelligence #machinelearning

Provably efficient reinforcement learning with rich observations.

https://www.microsoft.com/en-us/research/blog/provably-efficient-reinforcement-learning-with-rich-observations/

This is probably the best #PyTorch Deep Learning course I have encountered.
https://fleuret.org/dlc/

andrew ng : ML+ radiologist outperforms a human radiologist alone at detecting cerebral aneurysms.

Deep Learning–Assisted Diagnosis of Cerebral Aneurysms https://news.stanford.edu/2019/06/07/ai-tool-helps-radiologists-detect-brain-aneurysms/

Visualizing and Measuring the Geometry of BERT
Coenen, Reif, Yuan et al.: https://arxiv.org/pdf/1906.02715.pdf
#ArtificialIntelligence #DeepLearning #BERT #NLP

Language, trees, and geometry in neural networks
code https://pair-code.github.io/interpretability/bert-tree/
paper https://arxiv.org/pdf/1906.02715.pdf

Population-based Augmentation
1000x Faster Data Augmentation
Daniel Ho, Eric Liang, Richard Liaw Jun 7, 2019
https://bair.berkeley.edu/blog/2019/06/07/data_aug/
paper https://arxiv.org/pdf/1905.05393.pdf

Material used for Deep Learning related workshops for Machine Learning Tokyo
Implementation and Cheat Sheet: https://github.com/Machine-Learning-Tokyo/DL-workshop-series
#artificialintelligence #deeplearning #machinelearning

Residual Flows for Invertible Generative Modeling
Chen et al.: https://arxiv.org/abs/1906.02735
#artificialintelligence #deeplearning #generativemodels

Hot Papers from Google Brain, DeepMind and Facebook AI




https://www.google.com/amp/s/syncedreview.com/2019/06/02/hot-papers-from-google-brain-deepmind-and-facebook-ai/amp/

A machine-learning model from MIT researchers computationally breaks down how segments of amino acid chains determine a protein’s function, which could help researchers design and test new proteins for drug development or biological research.

https://news.mit.edu/2019/machine-learning-amino-acids-protein-function-0322

Yann LeCun et al. publishing evolutionary algorithm tools. Welcoming the era of deep neuroevolution indeed! (https://eng.uber.com/deep-neuroevolution) Great to see the traditional ML community adopt these tools in the cases when they are useful.