Blog

The latest from Google Research

The Google Brain Team — Looking Back on 2017 (Part 1 of 2)

jueves, 11 de enero de 2018

Posted by Jeff Dean, Google Senior Fellow, on behalf of the entire Google Brain TeamGoogle Brain teamGoogle AIsharedGoogleAlphabetmachine learningsecond postCore ResearchAutoMLnew approaches for designing neural network architecturesstate-of-the-art results on ImageNet classification and detectionoptimization algorithmseffective activation functionsCloud AI

Convolutional architecture discovered by Neural Architecture Search

Object detection with a network discovered by AutoML

Speech Understanding and Generationdevelop a number of improvements for an end-to-end approach to speech recognition123456789

Components of the Listen-Attend-Spell end-to-end model for speech recognition

Machine Perceptionlisten for yourself

Tacotron 2’s model architecture

New Machine Learning Algorithms and Approachescapsulessparsely-gated mixtures of expertshypernetworksnew kinds of multi-modal modelsattention-based mechanismssymbolicnon-symbolicback-propagate through discrete variablesreinforcementlearningMachine Learning for Computer Systemsreinforcement learning to make placement decisions for mapping computational graphs onto a set of computational devicesThe Case for Learned Index StructuresMachine Learning for Systems and Systems for Machine Learning

Learned Models as Index Structures

Privacy and Securitya paperICLR 2017adversarial examplesdemonstrating adversarial examples in the physical worldhow to harness adversarial examples at scale during the training processUnderstanding Machine Learning Systemsdeep learningone of the best paper awards“flatness” of minima found by optimization methods is not as closely linked to good generalization as initially thoughtrandommatricesrecent study comparing many GAN approachesmethods that allow better interpretability of machine learning systemsOpenAIDeepMindYC Researchlaunch of DistillDistillmanyilluminatingarticles

Feature Visualization

How to Use t-SNE effectively

Open Datasets for Machine Learning ResearchMNISTCIFAR-10ImageNetSVHNWMT

YouTube-8M: >7 million YouTube videos annotated with 4,716 different classes

YouTube-Bounding Boxes: 5 million bounding boxes from 210,000 YouTube videos

Speech Commands Dataset: thousands of speakers saying short command words

AudioSet: 2 million 10-second YouTube clips labeled with 527 different sound events

Atomic Visual Actions (AVA): 210,000 action labels across 57,000 video clips

Open Images: 9M creative-commons licensed images labeled with 6000 classes

Open Images with Bounding Boxes: 1.2M bounding boxes for 600 classes

Examples from the YouTube-Bounding Boxes dataset: Video segments sampled at 1 frame per second, with bounding boxes successfully identified around the items of interest.

TensorFlow and Open Source Software

A map showing the broad distribution of TensorFlow users (source)

TensorFlowTensorFlow 1.0we released v1.4Eager executionXLATensorFlow Litepre-compiled TensorFlow binariessource code on GitHubTensorFlow Developer Summittalks were recordedSign up now

This rock-paper-scissors science experiment is a novel use of TensorFlow. We’ve been excited by the wide variety of uses of TensorFlow we saw in 2017, including automating cucumber sorting, finding sea cows in aerial imagery, sorting diced potatoes to make safer baby food, identifying skin cancer, helping to interpret bird call recordings in a New Zealand bird sanctuary, and identifying diseased plants in the most popular root crop on Earth in Tanzania!

one of the top five repositoriesmany companies and organizationsmore than 24,500 distinct repositories on GitHubTF-GANTensorFlow LatticeTensorFlow Object Detection APImodel repositorydeeplearn.jsopen-source hardware-accelerated implementation of deep learning APIs right in the browserTeachable MachinePerformance RNNTPUs

Cloud TPUs deliver up to 180 teraflops of machine learning acceleration

Google SearchGoogle TranslateGoogle PhotosAlphaGoLee SedolKe JiepaperISCA 2017

Cloud TPU Pods deliver up to 11.5 petaflops of machine learning acceleration

Experiments with ResNet-50 training on ImageNet show near-perfect speed-up as the number of TPU devices used increases.

second-generation TPUGoogle Cloud PlatformCloud TPUsTensorFlow Research Cloud (TFRC)presented workg.co/tpusignup(In part 2 we’ll discuss our research in the application of machine learning to domains like healthcare, robotics, different fields of science, and creativity, as well as cover our work on fairness and inclusion.)