Google AI Blog: enero 2017

App Discovery with Google Play, Part 3: Machine Learning to Fight Spam and Abuse at Scale

lunes, 30 de enero de 2017

Posted by Hsu-Chieh Lee, Xing Chen, Software Engineers, and Qian An, Analyst
In Part 1 and Part 2 of this series on app discovery, we discussed using machine learning to gain a deeper understanding of the topics associated with an app, and a deep learning framework to provide personalized recommendations. In this post, we discuss a machine learning approach to fight spam and abuse on apps section of the Google Play Store, making it a safe and trusted app platform for more than a billion Android users. Identifying and blocking ‘bad’ apps from entering Google Play platformGoogle Play Developer Policy

Visualization of word embedding of samples of offensive content policy violating apps (red dots) and policy compliant apps (green dots), visualized with t-SNE (t-Distributed Stochastic Neighbor Embedding).

Preventing manipulation of app ratings and rankings adversarial problemPart 1Part 2unsupervised learningsupervised learning

A visualization of how a model trained on known bad actors (red) expands coverage to detect similar bad actors (orange) while ignoring organic users (blue).

Acknowledgements

Facilitating the discovery of public datasets

martes, 24 de enero de 2017

Posted by Natasha Noy, Google Research and Dan Brickley, Open Source Programs Office
publish datasocial sciencelife sciencehigh-energy physicsclimate sciencerelyschema.orgrecently published new guidelinesKnowledge GraphData Catalog VocabularydiscussionsTechnical Challenges

Defining more consistently what constitutes a dataset: For example, is a single table a dataset? What about a collection of related tables? What about a protein sequence? A set of images? An API that provides access to data? We hope that a better understanding of what a dataset is will emerge as we gain more experience with how data providers define, describe, and use data.

Identifying datasets: Ideally, datasets should have permanent identifiers conforming to some well known scheme that enables us to identify them uniquely, but often they don’t. Is a URL for the metadata page a good identifier? Can there be multiple identifiers? Is there a primary one?

Relating datasets to each other: When are two records describing a dataset “the same” (for instance, if one repository copies metadata from another )? What if an aggregator provides more metadata about the same dataset or cleans the data in some useful way? We are working on clarifying and defining these relationships, but it is likely that consumers of metadata will have to assume that many data providers are using these predicates imprecisely and need to be tolerant of that.

Propagating metadata between related datasets: How much of the metadata can we propagate among related datasets? For instance, we can probably propagate provenance information from a composite dataset to the datasets that it contains. But how much does the metadata “degrade” with such propagation? We expect the answer to be different depending on the application: metadata for search applications may be less precise than, say, for data integration.

Describing content of datasets: How much of the dataset content should we describe to enable support for queries similar to those used in Explore for Docs, Sheets and Slides, or other exploration and reuse of the content of the datasets (where license terms allow, of course)? How can we efficiently use content descriptions that providers already describe in a declarative way using W3C standards for describing semantics of Web resources and linked data?

otherpromisingeffortschallenging problemA Call to Action

For data providers, both individual providers and data repositories: publishing structured metadata using schema.org, DCAT, CSVW, and other community standards will make this metadata available for others to discover and use.

For data consumers (from scientists to data journalists and more): citing data properly, much as we cite scientific publications (see, for example, a recently proposed approach).

For developers: to contribute to expanding schema.org metadata for datasets, providing domain-specific vocabularies, as well as working on tools and applications that consume this rich metadata.

A Large Corpus for Supervised Word-Sense Disambiguation

miércoles, 18 de enero de 2017

Posted by Colin Evans and Dayu Yuan, Software Engineershe will receive stock in the reorganized companystockthe capital raised by a business or corporation through the issue and subscription of sharesNew Oxford American Dictionarystockgoods in a storea medieval device for punishmentAI-completeNavigli, 2009Ide and Veronis 1998Mallery 1988release of word-sense annotationsMASCSemCorEnglish WordnetSupervised Word-Sense Disambiguation[stock] in a business”[stock] in a bodegaword-sense disambiguationentity disambiguationNavigli, 2009

Facilitating the automatic construction of databases from text in order to answer questions and connect knowledge in documents. For example, understanding that a “hemi engine” is a kind of automotive machinery, and a “locomotive engine” is a kind of train, or that “Kanye West is a star” implies that he is a celebrity, but “Sirius is a star” implies that it is an astronomical object.

Disambiguating words in queries, so that results for “date palm” and “date night” or “web spam” and “spam recipe” can have distinct interpretations for different senses, and documents returned from a query have the same meaning that is implied by the query.

Manual Annotation

generalThe word is a typoNone of the aboveI can’t decideKrippendorff's alphaKrippendorff, 2004

Total

noun

verb

adjective

adverb

SemCor

115k

38k

57k

11.6k

8.6k

MASC

133k

50k

12.7k

13.6k

4.2k

Wordnet MappingsWordnetSemi-supervised Word Sense Disambiguation with Neural ModelsAcknowledgements

The Google Brain Team — Looking Back on 2016

jueves, 12 de enero de 2017

Posted by Jeff Dean, Google Senior Fellow, on behalf of the entire Google Brain TeamGoogle Brain teamResearch PublicationsICMLNIPSICLRprogram synthesisknowledge transfer from one network to anotherdistributed training of machine learning modelsgenerative models for languageunsupervised learning for roboticsautomated theorem provingbetter theoretical understanding of neural networksalgorithms for improved reinforcement learningACLCoNNLICASSPCVPRISEROSDIICLR 2017hereNatural Language UnderstandingSequence to Sequence Learning with Neural Networksgenerating captions for imagesparsing sentencessolving computational geometry problemsreplace the translation algorithms powering Google Translateresearch paperzero-shot translationresearch paperThe Great A.I. AwakeningThe NY Times MagazineRobotics[X]learn hand-eye coordinationresearch paperwe explored three possible ways for robots to learn new skillsmade multiple robotics datasets publicly availableHealthcarepaperJournal of the American Medical AssociationJAMAthat machine learning can improve the quality and efficiency of the healthcare experience for doctors and patientsMusic and Art GenerationMagentaexplore the intersection of art and machine intelligenceone-day symposiumsupported an art exhibition of machine generated artmusic generationartistic style transferour jam session demo won the Best Demo Award at NIPS 2016AI Safety and Fairnesswhite paper on Concrete Problems in AI Safetyblog post heredifferential privacy guaranteesknowledge transfer techniquesa paper on equality of opportunity in supervised learningblog post herecreated a visualization to help illustrate and interactively explore the concepts from the paperTensorFlowopen-sourced an initial version of TensorFlowthe most popular machine learning project on GitHubTensorFlow’s repository of modelsmore than 5000 TensorFlow-related repositorieswell-known research groups and large companiesDeepMindfinding sea cows Down Undersorting cucumbers in Japannumerous performance improvementsadded support for distributed trainingiOSRaspberry PiWindowsbig data infrastructureTensorBoardcomputation graphsembeddingsGoRustHaskellstate-of-the-art image classification modelsWide and DeepGitHubStackOverflowTensorFlow mailing listTensorFlow ServingGoogle Cloud Machine Learningcelebrated TensorFlow’s one year anniversary as an open-source projectpaper on the computer systems aspects of TensorFlowOSDIbackend compiler for TensorFlow called XLAadded to the open-source releaseMachine Learning Community Involvementfree online deep learning courseblog announcementTensorFlow PlaygroundGoogle Brain Residentsin seven months they have already conducted significantly original research, helping to author 21 research papersGoogle Brain team Reddit AMA (Ask Me Anything) on r/MachineLearningSpreading Machine Learning within Googlecustom machine learning accelerator ASIC that was discussed at Google I/ORankBrainNeural Machine Translation systemAlphaGo

Google Brain Residency Program - 7 months in and looking ahead

jueves, 5 de enero de 2017

Posted by Jeff Dean, Google Senior Fellow and Leslie Phillips, Google Brain Residency Program Manager“Beyond being incredibly instructive, the Google Brain Residency program has been a truly affirming experience. Working alongside people who truly love what they do--and are eager to help you develop your own passion--has vastly increased my confidence in my interests, my ability to explore them, and my plans for the near future.”Google Brain ResidencyGoogle Brain teamBayLearn ConferenceMean Field Neural NetworksRegularizing Neural Networks by Penalizing Their Output DistributionNIPS 2016 Adversarial Training workshopISMIR 2016ICLR 2017

An LSTM Cell (Left) and a state of the art RNN Cell found using a neural network (Right). This is an example of a novel architecture found using the approach presented in “Neural Architecture Search with Reinforcement Learning” (B. Zoph and Q. V. Le, submitted to ICLR 2017). This paper uses a neural network to generate novel RNN cell architectures that outperform the widely used LSTM on a variety of different tasks.

The training accuracy for neural networks, colored from black (random chance) to red (high accuracy). Overlaid in white dashed lines are the theoretical predictions showing the boundary between trainable and untrainable networks. (a) Networks with no dropout. (b)-(d) Networks with dropout rates of 0.01, 0.02, 0.06 respectively. This research explores whether theoretical calculations can replace large hyperparameter searches. For more details, read “Deep Information Propagation” (S. S. Schoenholz, J. Gilmer, S. Ganguli, J. Sohl-Dickstein, submitted to ICLR 2017).

Accepted conference papers(Google Brain Residents marked with asterisks)Unrolled Generative Adversarial NetworksLuke Metz*, Ben Poole, David Pfau, Jascha Sohl-DicksteinNIPS 2016 Adversarial Training Workshop (oral presentation)Conditional Image Synthesis with Auxiliary Classifier GANsAugustus Odena*, Chris Olah, Jon ShlensNIPS 2016 Adversarial Training Workshop (oral presentation)Regularizing Neural Networks by Penalizing Their Output DistributionGabriel Pereyra*, George Tucker, Lukasz Kaiser, Geoff HintonBayLearn 2016 (oral presentation)Mean Field Neural NetworksSamuel S. Schoenholz*, Justin Gilmer*, Jascha Sohl-DicksteinBayLearn 2016 (oral presentation)Learning to RememberAurko Roy, Ofir Nachum*, Łukasz Kaiser, Samy BengioBayLearn 2016 (poster session)Towards Generating Higher Resolution Images with Generative Adversarial NetworksAugustus Odena*, Jonathon ShlensBayLearn 2016 (poster session)Multi-Task Convolutional Music ModelsDiego Ardila, Cinjon Resnick*, Adam Roberts, Douglas EckBayLearn 2016 (poster session)Audio DeepDream: Optimizing Raw Audio With Convolutional NetworksDiego Ardila, Cinjon Resnick*, Adam Roberts, Douglas EckISMIR 2016 (poster session)Papers under review(Google Brain Residents marked with asterisks)Learning to Remember Rare Events Lukasz Kaiser, Ofir Nachum*, Aurko Roy, Samy Bengio Submitted to ICLR 2017 Neural Combinatorial Optimization with Reinforcement LearningIrwan Bello*, Hieu Pham*, Quoc V. Le, Mohammad Norouzi, Samy BengioSubmitted to ICLR 2017 HyperNetworksDavid Ha*, Andrew Dai, Quoc V. LeSubmitted to ICLR 2017 Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts LayerNoam Shazeer, Azalia Mirhoseini*, Krzysztof Maziarz, Quoc Le, Jeff DeanSubmitted to ICLR 2017 Neural Architecture Search with Reinforcement LearningBarret Zoph* and Quoc LeSubmitted to ICLR 2017 Deep Information PropagationSamuel Schoenholz*, Justin Gilmer*, Surya Ganguli, Jascha Sohl-DicksteinSubmitted to ICLR 2017 Capacity and Trainability in Recurrent Neural NetworksJasmine Collins*, Jascha Sohl-Dickstein, David SussilloSubmitted to ICLR 2017 Unrolled Generative Adversarial NetworksLuke Metz*, Ben Poole, David Pfau, Jascha Sohl-DicksteinSubmitted to ICLR 2017 Conditional Image Synthesis with Auxiliary Classifier GANsAugustus Odena*, Chris Olah, Jon ShlensSubmitted to ICLR 2017
Generating Long and Diverse Responses with Neural Conversation ModelsLouis Shao, Stephan Gouws, Denny Britz*, Anna Goldie, Brian Strope, Ray KurzweilSubmitted to ICLR 2017Intelligible Language Modeling with Input Switched Affine NetworksJakob Foerster, Justin Gilmer*, Jan Chorowski, Jascha Sohl-dickstein, David SussilloSubmitted to ICLR 2017 Regularizing Neural Networks by Penalizing Confident Output DistributionsGabriel Pereyra*, George Tucker*, Jan Chorowski, Lukasz Kaiser, Geoffrey HintonSubmitted to ICLR 2017 Unsupervised Perceptual Rewards for Imitation LearningPierre Sermanet, Kelvin Xu*, Sergey LevineSubmitted to ICLR 2017 Improving policy gradient by exploring under-appreciated rewardsOfir Nachum*, Mohammad Norouzi, Dale SchuurmansSubmitted to ICLR 2017 Protein Secondary Structure Prediction Using Deep Multi-scale Convolutional Neural Networks and Next-Step ConditioningAkosua Busia*, Jasmine Collins*, Navdeep JaitlyWe are currently accepting applications for the 2017 Google Brain Residency Programg.co/brainresidency

Blog

App Discovery with Google Play, Part 3: Machine Learning to Fight Spam and Abuse at Scale

Facilitating the discovery of public datasets

A Large Corpus for Supervised Word-Sense Disambiguation

The Google Brain Team — Looking Back on 2016

Google Brain Residency Program - 7 months in and looking ahead

Etiquetas

Archive

Feed