"Quand la machine apprend"
Un livre dont je suis l'auteur qui sort aujourd'hui en librairie.
Edité par Odile Jacob.

Broché: https://livre.fnac.com/a13700561/Yann-Le-Cun-Quand-la-machine-apprend

Broché: https://www.amazon.fr/gp/aw/d/2738149316/ref=ox_sc_act_image_1

En PDF et ePub: https://www.odilejacob.fr/catalogue/sciences/informatique/quand-la-machine-apprend_9782738149312.php

Format Kindle:
https://www.amazon.fr/dp/B07Z5N3LQM/ref=cm_sw_r_tw_awdb_c_x_wIUPDbTFT7Q07

Unsupervised Word Embeddings Capture Latent Knowledge from Materials Science Literature,"
https://go.nature.com/32dCEfi

An algorithm with no training in materials science can scan the text of millions of papers and uncover new scientific knowledge.

3.3 million abstracts of published materials science papers were fed into an algorithm called Word2vec.

By analyzing relationships between words the algorithm was able to predict discoveries of new thermoelectric materials years in advance and suggest as-yet unknown materials as candidates for thermoelectric materials.

https://towardsdatascience.com/using-unsupervised-machine-learning-to-uncover-hidden-scientific-knowledge-6a3689e1c78d

Online evaluation of machine learning models
https://conferences.oreilly.com/artificial-intelligence/ai-eu/public/schedule/detail/78170

U of T, Vector Institute create new deep learning faculty positions in honour of Geoffrey Hinton


https://www.utoronto.ca/news/u-t-vector-institute-create-new-deep-learning-faculty-positions-honour-geoffrey-hinton https://t.iss.one/ArtificialIntelligenceArticles

High energy: Facebook's AI guru LeCun imagines AI's next frontier
https://www.zdnet.com/article/high-energy-facebooks-ai-guru-lecun-imagines-ais-next-frontier/

Emergent properties of the local geometry of neural loss landscapes
Stanislav Fort and Surya Ganguli : https://arxiv.org/abs/1910.05929
#MachineLearning #NeuralNetwork #DeepLearning

Video Architecture Search
Blog by Michael S. Ryoo and AJ Piergiovanni : https://ai.googleblog.com/2019/10/video-architecture-search.html
#ArtificialIntelligence #DeepLearning #NeuralArchitectureSearch

Kimera: an Open-Source Library for Real-Time Metric-Semantic Localization and Mapping
Video: https://www.youtube.com/watch?v=-5XxXRABXJs&feature=youtu.be
Code: https://github.com/MIT-SPARK/Kimera
Paper: https://arxiv.org/abs/1910.02490

From Facebook AI:

ReAgent, the most comprehensive and modular open source platform for creating AI-based reasoning systems, and the first to include policy evaluation to incorporate offline feedback to improve models.
https://ai.facebook.com/blog/open-sourcing-reagent-a-platform-for-reasoning-systems/

Quantized Reinforcement Learning (QUARL)
Srivatsan Krishnan, Sharad Chitlangia, Maximilian Lam, Zishen Wan, Aleksandra Faust, Vijay Janapa Reddi : https://arxiv.org/abs/1910.01055
Code: https://github.com/harvard-edge/quarl
#DeepLearning #ReinforcementLearning #Quantization #Robotics

Minimal implementation of a Neural Architecture Search system
https://github.com/nicklashansen/minimal-nas

Active Learning for Graph Neural Networks via Node Feature Propagation
Wu et al.: https://arxiv.org/abs/1910.07567
#ArtificialIntelligence #GraphNeuralNetworks #MachineLearning

Deep Reinforcement Learning meets Graph Neural Networks: An optical network routing use case
Almasan et al.: https://arxiv.org/abs/1910.07421
#DeepLearning #ReinforcementLearning #GraphNeuralNetworks

Machine learning has the potential to improve different steps of the radiology workflow including order scheduling and triage, clinical decision support systems, detection and interpretation of findings, postprocessing and dose estimation, examination quality control, and radiology reporting.
https://pubs.rsna.org/doi/10.1148/radiol.2018171820

Bias-Resilient Neural Network
Adeli et al.: https://arxiv.org/abs/1910.03676
#ArtificialIntelligence #DeepLearning #NeuralNetworks

Precise measurement of quantum observables with neural-network estimators
Torlai et al.: https://arxiv.org/abs/1910.07596
#ArtificialIntelligence #QuantumPhysics #NeuralNetworks

"Restoring ancient text using deep learning: a case study on Greek epigraphy"
Assael et al.: https://arxiv.org/abs/1910.06262
Code: https://github.com/sommerschield/ancient-text-restoration
#ArtificialIntelligence #DeepLearning #NeuralNetworks

DeepGCNs: Making GCNs Go as Deep as CNNs

https://deepai.org/publication/deepgcns-making-gcns-go-as-deep-as-cnns?fbclid=IwAR2edqmWo5uKSGybcgRWW43ov-03resk_as2EoJ52nzeaF_3jSnnV3bxH1o

#DeepAI #neuralnetworks #CNNs ##GCNs

Deep networks work by learning complex, often hierarchical internal representations of input data. These form a kind of functional language the network uses to describe the data.

Language can emerge from tasks like object recognition: has pointy ears, whiskers, tail => cat.

This relates to Wittgenstein’s "language-game" in Philosophical Investigations, where a functional language emerge from simple tasks before defining a vocabulary.

The visual vocabulary of a convolutional neural network seems to emerge from low level features such as edges and orientations, and builds up textures, patterns and composites, … and builds up even further into complete objects: houses, dogs, etc.

Source: NeurIPS 2018“Unsupervised Deep Learning” Tutorial – Part 1 by Alex Graves - https://media.neurips.cc/Conferences/NIPS2018/Slides/Deep_Unsupervised_Learning.pdf

#artificialintelligence #deeplearning #language #machinelearning