Entries from September 2017 ↓

• New Theory Cracks Open the Black Box of Deep Learning | Quanta Magazine
• Twitter
  RT @BarristerSecret: I predict that this decision will be a watershed in demonstrating how few people know how to use licence and licens…
• Twitter
  RT @catherine_mayer: James Watson’s autobiog: Rosalind Franklin dowdy, "had to go or be put in her place" but "a good brain" if emotions…
• Twitter
  RT @TheJihyeLee: The Korean original statement said "늙다리 미치광이," which means old beast lunatic — which was translated into "dotard."
• Twitter
  RT @sarahljaffe: noting once again: this plan is designed to punish states *that provided healthcare for people*
• Twitter
  RT @m_older: also the entire concept of society is that we do better together than apart. Especially true for unpredictable shoc…
• Twitter
  RT @TimNoakes: “I’ve been preambulating the liminal spaces of the submerged memory-city” – Psychogeographic Pigeon, London, 2017
• Russian hacking election: States switch to paper ballots
• Gene Luen Yang wins a Macarthur "genius" prize! / Boing Boing
• Soton Archaeology on Twitter: "#Archaeology from another world! 3D point cloud from AUV survey of #Roman #Shipwreck lying 2km deep, via #blackseamap for @unisouthampton https://t.co/Wz9WdAPBY8"
• Twitter
  RT @LibyaLiberty: My favorite exchange on Twitter today.
• My Life in Domestic Goddesses
• Best Blue Cheese in the World – Organic Cheese
• The Dinner Party | The New Yorker
• 35 Best Books of 2017 (So Far) – Best New Books of 2017
• Penguin Random House on Twitter: "The Best Books of 2017 (So Far): https://t.co/kYAJynvBNY @esquire https://t.co/cJO8ogXyRG"
• Graham-Cassidy Health Care Bill: Even Insurers Oppose It | Fortune.com
• Twitter
  RT @LailaLalami: Must-read. "This past year should have worn away our capacity to say demonstrations of white supremacy and racial v…
• Was Charlottesville the Exception or the Rule? – The New York Times
• Twitter
  RT @brianschatz: This is why they are in a rush. The more we know, the worse it gets.
• Twitter
  RT @SirWilliamD: The oldest known photograph of #London – Whitehall from Trafalgar Square c.1839. A Daguerreotype by M. de St. Croix.
• Twitter
  RT @ASlavitt: BREAKING: The bipartisan Medicaid Directors from all 50 states just issued this negative statement about Graham-Cas…
• Testosterone Rex triumphs as Royal Society science book of the year | Books | The Guardian
• Twitter
  RT @samswey: The 15 most deadly police departments in America. St. Louis is off the charts.
• Police Accountability Tool — Mapping Police Violence
• Twitter
  RT @SanhoTree: Please make #PuertoRico trend. They’re 100% without power and can’t tweet it themselves. Unprecedented levels of as…
• Syria Holds the Secret to Our Species’ Survival – Mother Jones
• Twitter
  RT @kibblesmith: When you get a work e-mail on the weekend.
• Snopes.com and the Search for Facts in a Post-Fact World | WIRED

Digest powered by RSS Digest

• PipelineAI – Features
• Why is Python Growing So Quickly? – Stack Overflow Blog
• Where to Donate to Mexico Earthquake Victims – The New York Times
• Twitter
  RT @douglasmack: See also: residents of the *U.S.* Virgin Islands.
• How to get the Kubernetes help you need | ZDNet
• San Francisco, Oakland sue major oil companies over rising seas – SFGate.com
• The Political Awakening of Silicon Valley — The California Sunday Magazine
• Bruce Arthur on Twitter: "Jimmy Kimmel’s monologue on health care tonight was angry, funny, smart and incredible. https://t.co/P2vTrQmAgs"
• SEC reveals it was hacked, information may have been used for illegal stock trades – The Washington Post
• Twitter
  RT @ericgeller: Almost 3.5 million Americans will be without power for four to six months.
• Manafort Working on Kurdish Referendum Opposed by U.S. – NYTimes.com
• Channel 4 News on Twitter: "60 perfectly preserved shipwrecks over 2,400 years old have been found in the Black Sea. The last time anyone saw them, Socrates was alive. https://t.co/bUBPJnGcHb"
• Manafort offered to give Russian billionaire ‘private briefings’ on 2016 campaign – The Washington Post
• World Without Mind: The Existential Threat of Big Tech
• Thanks, Obama: My Hopey, Changey White House Years
• “Devastating:" Authors retract paper in Nature journal upon discovering error – Retraction Watch at Retraction Watch
• Twitter
  RT @edd: Happy news: I’ve joined Google, to work on making TensorFlow an even more successful open source project…
• Twitter
  RT @EricaJoy: I appreciate James Damore continuing to provide evidence that the bar at Google was most certainly lowered in a spe…
• Twitter
  RT @ClimateCentral: The scariest thing about 2017’s hurricanes: They got really bad really fast via…
• What’s scary about 2017’s hurricanes isn’t just their strength. It’s how fast they’re achieving it – The Washington Post
  RT @ClimateCentral: The scariest thing about 2017’s hurricanes: They got really bad really fast via…
• Research Blog: Revisiting the Unreasonable Effectiveness of Data
• PipelineAI – About Us
• Azure/Machine-Learning-Operationalization: Deploying machine learning models to Azure
  Deploying machine learning models to Azure

  Operationalization is the process of publishing models and code as web services and the consumption of these services to produce business results.

  Azure Machine Learning Operationalization is a component of the Azure CLI that enables operationalization of models that you create through Vienna that use the CNTK, SPARK, and Python machine learning platforms.

• terinjokes/StickerConstructorSpec: The Sticker Standard – hexagon sticker spec
  This specifications defines the standard sizes and orientation to allow consistent sticker sizes between vendors. Allowing maximum enjoyment from receivers of stickers, and allow sticker companies to develop tools and interfaces for creating compliant stickers. This specification is also a definitive place to know what dimensions a sticker should be.

Digest powered by RSS Digest

• Pilot Vanishing Point Fountain Pens – JetPens.com
• Using Vault on Compute Engine for Secret Management  |  Solutions  |  Google Cloud Platform
• [1704.05021] Sparse Communication for Distributed Gradient Descent
  We make distributed stochastic gradient descent faster by exchanging sparse updates instead of dense updates. Gradient updates are positively skewed as most updates are near zero, so we map the 99% smallest updates (by absolute value) to zero then exchange sparse matrices. This method can be combined with quantization to further improve the compression. We explore different configurations and apply them to neural machine translation and MNIST image classification tasks. Most configurations work on MNIST, whereas different configurations reduce convergence rate on the more complex translation task. Our experiments show that we can achieve up to 49% speed up on MNIST and 22% on NMT without damaging the final accuracy or BLEU.

  ..goog has unpub’d paper on doing the same..

• Twitter
  RT @mattyglesias: Hard to imagine the guy who did shady insider trading as a House member would get mixed up in corruption.
• The free knowledge-sharing platform for technology
  Tech.io is a collaborative platform to discover and share unique perspectives on any technology through open-source playgrounds. Empower others to learn by crafting hands-on tutorials on topics that matter to you.
• Tech Workers Join Calls For Amazon to Improve Working Conditions for Security Guards – Slog – The Stranger
• Google Sued by 3 Female Ex-Employees Who Say It Pays Women Less Than Men – The New York Times
• The Cassini spacecraft crashed into Saturn, ending a successful 20-year mission – The Washington Post
• Social influence and political mobilization: Further evidence from a randomized experiment in the 2012 U.S. presidential election
• Google Now Giving Female Employees Free Day Each Week To Work On Lawsuits – The Onion – America’s Finest News Source
• Twitter
  RT @upulie: I’m hoping that I can successfully replicate my experiment and thus create reliable nanoparticles that can deliver…
• Introducing Kaleidoscope for iPad – BPXL Craft – Medium
• Astronomers find a giant, black, glowing exoplanet – Axios
• The Graham-Cassidy Bill Is a Potential Disaster | The New Yorker
• Twitter
  RT @jonathanchait: Bill is moving so fast, and without analysis, so sponsors can tell different untruths to different constituencies.
• Scripps says climate change may represent "existential" threat to humanity – The San Diego Union-Tribune
• Credible sources of accurate information about AI · fast.ai
• Twitter
  RT @TopherSpiro: Zero press coverage
• Jeff Lawson on Twitter: "Introducing Studio: the Visual Builder for Twilio https://t.co/9Hf8rgLGRz https://t.co/A8OMdMUbcO"
• Google Cloud Platform Blog: Introducing managed SSL for Google App Engine
• Matt Baldwin on Twitter: "We’ve combined @kubernetesio & @IstioMesh with our managed, turn-key offering @StackPointCloud https://t.co/U0gASrBAmC"
• StackPointCloud Announces Managed Istio – Blog | StackPointCloud
• Introducing: Unity Machine Learning Agents – Unity Blog
  RT @fchollet: Unity machine learning agents:
• Time-aligned spectrograms with CHAT files in R | Kaggle
• Tyke
• Genetic Engineering Holds the Power to Save Humanity or Kill It – MIT Technology Review
• Undercover With the Alt-Right – NYTimes.com
• nvALT 2.2.8, because I got High (Sierra) – BrettTerpstra.com
• [1611.04558v1] Google’s Multilingual Neural Machine Translation System: Enabling Zero-Shot Translation
  We propose a simple, elegant solution to use a single Neural Machine Translation (NMT) model to translate between multiple languages. Our solution requires no change in the model architecture from our base system but instead introduces an artificial token at the beginning of the input sentence to specify the required target language. The rest of the model, which includes encoder, decoder and attention, remains unchanged and is shared across all languages. Using a shared wordpiece vocabulary, our approach enables Multilingual NMT using a single model without any increase in parameters, which is significantly simpler than previous proposals for Multilingual NMT. Our method often improves the translation quality of all involved language pairs, even while keeping the total number of model parameters constant. On the WMT’14 benchmarks, a single multilingual model achieves comparable performance for English→French and surpasses state-of-the-art results for English→German. Similarly, a single multilingual model surpasses state-of-the-art results for French→English and German→English on WMT’14 and WMT’15 benchmarks respectively. On production corpora, multilingual models of up to twelve language pairs allow for better translation of many individual pairs. In addition to improving the translation quality of language pairs that the model was trained with, our models can also learn to perform implicit bridging between language pairs never seen explicitly during training, showing that transfer learning and zero-shot translation is possible for neural translation. Finally, we show analyses that hints at a universal interlingua representation in our models and show some interesting examples when mixing languages.

Digest powered by RSS Digest

• Satya Nadella Rewrites Microsoft’s Code
• How to Work From Home – NYTimes.com
• What do you think of PyTorch compared to…4 answers – Quora
• Should I go for TensorFlow or PyTorch? – Quora
• PyTorch — Dynamic Batching – Illia Polosukhin – Medium
  If you have been reading my blog, you may have seen that I was a TensorFlow contributor and built a lot of high-level APIs there.
  In Feb 2017 though, I have left Google and co-founded my own company — NEAR.ai. Where we are teaching machines to write code from natural language.
  As part of this work, we are building Deep Learning models that are reading or writing code in a tree format. After trying to manage this complexity in TensorFlow, I’ve decided to give a try to PyTorch.
• Facebook and Microsoft introduce new open ecosystem for interchangeable AI frameworks – Facebook Research
  Facebook and Microsoft are today introducing Open Neural Network Exchange (ONNX) format, a standard for representing deep learning models that enables models to be transferred between frameworks. ONNX is the first step toward an open ecosystem where AI developers can easily move between state-of-the-art tools and choose the combination that is best for them.
• Identity Theft, Credit Reports, and You | Kalzumeus Software
• 2017 Finalists: fiction | Kirkus Reviews
• Top 10 twists in fiction | Books | The Guardian
• Trump Administration Rejects Study Showing Positive Impact of Refugees – NYTimes.com
• Open Source is Missing Out on Microservices – Kubernetes is the Solution
• Equifax Suffered a Hack Almost Five Months Earlier Than the Date It Disclosed – Bloomberg
• Ann Beattie Rearranges the House at Night
• Dave Weigel on Twitter: "CBO’s quick estimate of #GrahamCassidy won’t cover "effects on the deficit, health insurance coverage, or premiums.” https://t.co/LUy2OSmhXk"
• How to Read Bob Mueller’s Hand – POLITICO Magazine
• Agony Auntie: Running Into a One-Night Stand at the Grocery Store
  Samantha irby
• Cholera vaccination campaign BEGINS in north-eastern Nigeria | WHO | Regional Office for Africa
• Twitter
  RT @yufengg: Thanks to all who attended our session at #TheAIConf this morning, it was a blast! Updated slides here:…

Digest powered by RSS Digest

• Is Trump a White Supremacist? – The New York Times
• Generative Adversarial Networks – tyler schoenrock – Medium
• Computers increasingly fool people by generating realistic videos and images.
• What’s a Generative Adversarial Network? A Google Researcher Explains | NVIDIA Blog
• Whitesplaining Ta-Nehisi Coates | HuffPost
• Twitter
  soooo cool
• Twitter
  RT @Mikel_Jollett: Every time Trump sends a tweet about how "important" his day is going to be, he’s golfing.
• Twitter
  RT @JYSexton: Never, ever underestimate how badly Republicans want people to lose their health insurance.
• Amber Tamblyn: I’m Done With Not Being Believed – NYTimes.com
• Twitter
  RT @CCriadoPerez: I unironically love this so sue me
• Rohingya crisis in Myanmar: Refugees recount horror of ‘slaughter,’ then a perilous journey to Bangladesh – The Washington Post
• TV show triggers little-known phobia

Digest powered by RSS Digest

• Amy Unruh talks, workshops & event profile · Eventil
  RT @eventilhq: #upcoming Getting Started With Tensorflow · @amygdala at @strangeloop_stl 2017 (Sep 28-30)
• MassBailFund on Twitter: "We are closed indefinitely. The need for our help far surpasses our resources and we are shut down. https://t.co/d5nsM2V0nY"
• The Week My Husband Left And My House Was Burgled I Secured A Grant To Begin The Project That Became BRCA1 | HuffPost UK
• devopsdays Minneapolis 2017 – Aneel Lakhani – A Field Guide to Observability – YouTube
• The reMarkable tablet wants to replace all your paper notebooks | WIRED
• Turn Bluetooth Off When You’re Not Using It | WIRED
• I trust Apple’s security team to have implemented this in a way that preserves s… | Hacker News
• The Corpus Christi Prime – Friendly Fields and Open Maps
• Twitter
  RT @ianbremmer: US. Only households that have made any real progress in the last 17 years are those in the top 20%.

  (WaPo)

• [1708.04202] Learning to Plan Chemical Syntheses
  From medicines to materials, small organic molecules are indispensable for human well-being. To plan their syntheses, chemists employ a problem solving technique called retrosynthesis. In retrosynthesis, target molecules are recursively transformed into increasingly simpler precursor compounds until a set of readily available starting materials is obtained. Computer-aided retrosynthesis would be a highly valuable tool, however, past approaches were slow and provided results of unsatisfactory quality. Here, we employ Monte Carlo Tree Search (MCTS) to efficiently discover retrosynthetic routes. MCTS was combined with an expansion policy network that guides the search, and an "in-scope" filter network to pre-select the most promising retrosynthetic steps. These deep neural networks were trained on 12 million reactions, which represents essentially all reactions ever published in organic chemistry. Our system solves almost twice as many molecules and is 30 times faster in comparison to the traditional search method based on extracted rules and hand-coded heuristics. Finally after a 60 year history of computer-aided synthesis planning, chemists can no longer distinguish between routes generated by a computer system and real routes taken from the scientific literature. We anticipate that our method will accelerate drug and materials discovery by assisting chemists to plan better syntheses faster, and by enabling fully automated robot synthesis.
• Facebook announces major AI commitment to CIFAR : CIFAR
  Facebook announced a major investment with CIFAR today, a result of the Institute’s leadership in the field of artificial intelligence (AI). The US$2.625 million investment over five years will continue Facebook’s support of CIFAR’s Learning in Machines & Brains program, and will also fund a Facebook-CIFAR Chair in Artificial Intelligence at the Montreal Institute for Learning Algorithms (MILA).

  Facebook made the announcement Friday at a ceremony at McGill University in Montreal, attended by Prime Minister Justin Trudeau and CIFAR President & CEO Alan Bernstein. Facebook also announced funding for a Facebook AI Research (FAIR) Lab to be headed by Joëlle Pineau, a CIFAR Senior Fellow in the Learning in Machines & Brains program, and an artificial intelligence researcher at McGill. Pineau will be joined at FAIR by CIFAR Associate Fellow Pascal Vincent, an associate professor at the University of Montreal.

• Mercedes-Benz introduces the world’s first in-car 3 word address voice navigation system | what3words
• [1603.08155] Perceptual Losses for Real-Time Style Transfer and Super-Resolution
  We consider image transformation problems, where an input image is transformed into an output image. Recent methods for such problems typically train feed-forward convolutional neural networks using a \emph{per-pixel} loss between the output and ground-truth images. Parallel work has shown that high-quality images can be generated by defining and optimizing \emph{perceptual} loss functions based on high-level features extracted from pretrained networks. We combine the benefits of both approaches, and propose the use of perceptual loss functions for training feed-forward networks for image transformation tasks. We show results on image style transfer, where a feed-forward network is trained to solve the optimization problem proposed by Gatys et al in real-time. Compared to the optimization-based method, our network gives similar qualitative results but is three orders of magnitude faster. We also experiment with single-image super-resolution, where replacing a per-pixel loss with a perceptual loss gives visually pleasing results.

  (re: Prisma) . Fei-Fei

• [1603.03417] Texture Networks: Feed-forward Synthesis of Textures and Stylized Images
  Gatys et al. recently demonstrated that deep networks can generate beautiful textures and stylized images from a single texture example. However, their methods requires a slow and memory-consuming optimization process. We propose here an alternative approach that moves the computational burden to a learning stage. Given a single example of a texture, our approach trains compact feed-forward convolutional networks to generate multiple samples of the same texture of arbitrary size and to transfer artistic style from a given image to any other image. The resulting networks are remarkably light-weight and can generate textures of quality comparable to Gatys~et~al., but hundreds of times faster. More generally, our approach highlights the power and flexibility of generative feed-forward models trained with complex and expressive loss functions.

  (re: Prisma)

• [1604.04382] Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks
  This paper proposes Markovian Generative Adversarial Networks (MGANs), a method for training generative neural networks for efficient texture synthesis. While deep neural network approaches have recently demonstrated remarkable results in terms of synthesis quality, they still come at considerable computational costs (minutes of run-time for low-res images). Our paper addresses this efficiency issue. Instead of a numerical deconvolution in previous work, we precompute a feed-forward, strided convolutional network that captures the feature statistics of Markovian patches and is able to directly generate outputs of arbitrary dimensions. Such network can directly decode brown noise to realistic texture, or photos to artistic paintings. With adversarial training, we obtain quality comparable to recent neural texture synthesis methods. As no optimization is required any longer at generation time, our run-time performance (0.25M pixel images at 25Hz) surpasses previous neural texture synthesizers by a significant margin (at least 500 times faster). We apply this idea to texture synthesis, style transfer, and video stylization.

  (re Prisma?)

• Is Apple’s iPhone Upgrade Plan Right for You? | WIRED
• GitHub – kubernetes/client-go: Go client for Kubernetes.
• The Realities of Being a FOSS Maintainer – Site Feedback – Caddy Community
• Blue/Green Deployments on Kubernetes – Ian Lewis
• GitHub – taki0112/SENet-Tensorflow: Simple Tensorflow implementation of Squeeze Excitation Networks using Cifar10 (ResNeXt, Inception-v4, Inception-resnet-v2)
• Twitter
  RT @hhavrilesky: "is actually interesting"

  uh

  why

  wouldn’t

  it

  be

  ?

Digest powered by RSS Digest

• Why Hillary Clinton’s Book Is Worth Reading – The Atlantic
• The Birth of Hollywood’s Brat Pack — New York Magazine
• How The Kardashians Reflect Changing Ideas Around Plastic Surgery
• Cheat Sheets for AI, Neural Networks, Machine Learning, Deep Learning & Big Data
• Reading "Playboy: The Complete Centerfolds, 1953-2016" cover to cover
• How to Clean Up Safely After a Hurricane or Flood – NYTimes.com
• Twitter
  (whoa)
• Artificial intelligence pioneer says we need to start over – Axios
  Hinton
• Brutally Honest Emmy Ballot #2 2017: Aziz Ansari Is "Unbearable" and ‘Better Call Saul’ Is "A Small Show" | Hollywood Reporter
• [1709.04511] An Empirical Study of AI Population Dynamics with Million-agent Reinforcement Learning
  In this paper, we conduct an empirical study on discovering the ordered collective dynamics obtained by a population of artificial intelligence (AI) agents. Our intention is to put AI agents into a simulated natural context, and then to understand their induced dynamics at the population level. In particular, we aim to verify if the principles developed in the real world could also be used in understanding an artificially-created intelligent population. To achieve this, we simulate a large-scale predator-prey world, where the laws of the world are designed by only the findings or logical equivalence that have been discovered in nature. We endow the agents with the intelligence based on deep reinforcement learning, and scale the population size up to millions. Our results show that the population dynamics of AI agents, driven only by each agent’s individual self interest, reveals an ordered pattern that is similar to the Lotka-Volterra model studied in population biology. We further discover the emergent behaviors of collective adaptations in studying how the agents’ grouping behaviors will change with the environmental resources. Both of the two findings could be explained by the self-organization theory in nature.
• Joëlle Pineau to head new Facebook AI (FAIR) lab in Montreal : McGill Reporter
• Though She Be But Little – Uncanny Magazine
• Black Death skeletons unearthed by Crossrail project – BBC News
• Stainless Steel Magnetic Eyeglass Holder | ReadeREST
  badge holder
• Untitled (https://www.panthera.org/blog/2017/09/12/how-photograph-snow-leopard)
  RT @preetalina: Oh just a stunning #cameratrap Snow Leopard photo by Sebastian Kennerknecht! via…
• Twitter
  RT @preetalina: Oh just a stunning #cameratrap Snow Leopard photo by Sebastian Kennerknecht! via…

Digest powered by RSS Digest

• VGG-16 pre-trained model for Keras
  ##VGG16 model for Keras

  This is the Keras model of the 16-layer network used by the VGG team in the ILSVRC-2014 competition.

  It has been obtained by directly converting the Caffe model provived by the authors.

  Details about the network architecture can be found in the following arXiv paper:

  Very Deep Convolutional Networks for Large-Scale Image Recognition
  K. Simonyan, A. Zisserman
  arXiv:1409.1556
  In the paper, the VGG-16 model is denoted as configuration D. It achieves 7.5% top-5 error on ILSVRC-2012-val, 7.4% top-5 error on ILSVRC-2012-test.

  Please cite the paper if you use the models.

• The most popular baby names in every state – Holy Kaw!
• See the Cassini Probe’s Grand Finale at Saturn
• OK Computer at 20: an oral history of Radiohead’s iconic 1997 album
• Amber Tamblyn Pens Open Letter to James Woods | Teen Vogue
• The fanged, faceless sea creature that washed ashore during Harvey has been identified – The Washington Post
• CBO: Trump is making Obamacare premiums more expensive – Vox
• America’s vitriol towards Clinton reveals a nation mired in misogyny | Hadley Freeman | Opinion | The Guardian
• The Sorrow and the Shame of the Accidental Killer | The New Yorker
• Google Sets Limits on Addiction Treatment Ads, Citing Safety – NYTimes.com
• NASA is crashing the $4 billion Cassini spacecraft into Saturn – Vox
• Scripps says climate change may represent "existential" threat to humanity – The San Diego Union-Tribune
• ‘Th’ sound to vanish from English language by 2066 because of multiculturalism, say linguists
• Nic◾⛓Carter on Twitter: "Just read the most fascinating paper by Princeton researchers (BlockSci). on blockchain analysis. Link: https://t.co/DD8Q4oxG8h"
• Silicon Valley’s People Problem, Explained – CityLab
• Twitter
  RT @BadAstronomer: One of the very last images of Enceladus, partially lit by the Sun, partially by Saturn, setting behind the planet.
• Facebook’s New Lab Bolsters Montreal’s Bragging Rights As An AI Hub
• Patent EP0802493A2 – Process flow design technique – Google Patents
• Biologist reaches into electric eel tank, comes out with equation to measure shocks | Research News @ Vanderbilt | Vanderbilt University
• Reconstruction of a Train Wreck: How Priming Research Went off the Rails | Replicability-Index
• You Are Already Living Inside a Computer – The Atlantic
• Patent EP0802493A3 – Process flow design technique – Google Patents
• Harrison Ford on ‘Star Wars’, ‘Blade Runner’, and Punching Ryan Gosling in the Face | GQ
• Twitter
  RT @mcnees: These are the last images from a loyal space robot hurtling towards a fiery doom, this account should have 20 milli…
• Research Blog: The Google Brain Team’s Approach to Research
• Twitter
  RT @KuraFire: I strongly applaud @justkelly_ok for her (& Pease + Wisuri’s) bravery in their effort to address workplace & salary…
• Google ‘segregates’ women into lower-paying jobs, stifling careers, lawsuit says | Technology | The Guardian
• A VGG-like CNN in keras for Fashion-MNIST with 94% accuracy | SAS Programming for Data Mining
• HazyResearch/snorkel: A training data creation and management system focused on information extraction
  A training data creation and management system focused on information extraction http://snorkel.stanford.edu

  Snorkel is a system for rapidly creating, modeling, and managing training data, currently focused on accelerating the development of structured or "dark" data extraction applications for domains in which large labeled training sets are not available or easy to obtain.

  Today’s state-of-the-art machine learning models require massive labeled training sets–which usually do not exist for real-world applications. Instead, Snorkel is based around the new data programming paradigm, in which the developer focuses on writing a set of labeling functions, which are just scripts that programmatically label data. The resulting labels are noisy, but Snorkel automatically models this process—learning, essentially, which labeling functions are more accurate than others—and then uses this to train an end model (for example, a deep neural network in TensorFlow).

  Surprisingly, by modeling a noisy training set creation process in this way, we can take potentially low-quality labeling functions from the user, and use these to train high-quality end models. We see Snorkel as providing a general framework for many weak supervision techniques, and as defining a new programming model for weakly-supervised machine learning systems.

Digest powered by RSS Digest

• Heptio raises $25M Series B funding round as Kubernetes takes over the world – GeekWire
• Macbook Pro 13" Late 2016 has issues connecting… | Official Apple Support Communities
• ‘An insult’: Chesterfield’s Princess Diana floral tribute ridiculed | UK news | The Guardian
• Haunted By Data
• What Ta-Nehisi Coates gets wrong about leftists
• Tesla Remotely Extended The Range Of Drivers In Florida For Free… And That’s NOT A Good Thing | Techdirt
• Ed Piotrowski on Twitter: "Most of Florida lost power in #Irma. Here’s what it looked like from space. https://t.co/wi6KHjavd1 https://t.co/KwydjER3a9"
• Equifax’s Maddening Unaccountability – The New York Times
• Tech Developers: Google is in town on the 28th of September
  Capetown
• The Comey Letter Probably Cost Clinton The Election | FiveThirtyEight
• (503) http://a.co/dTjWLyR
  RT @akafrancie: Took network of people (+moved mountains) but Amazon List for Anguilla Relief is live +WE HAVE A PLANE to deliver ->
• francie on Twitter: "Took network of people (+moved mountains) but Amazon List for Anguilla Relief is live +WE HAVE A PLANE to deliver -> https://t.co/ACVHwTJIyZ"
• The Perfect Wife | The New Yorker
  Edie Windsor
• Twitter
  "We call successful women unlikable because they’re rupturing the rules of femininity."
• This is why we hate successful women so much | Metro News
• Fire Management May Date Back 40,000 Years – Archaeology Magazine
• Twitter
  RT @mustafasuleymn: No women??? Seriously?!?!
• London Review of Books

  You Are the Product

• Facebook just doesn’t care. And yet we’ve handed it the keys to our civilisation – One Man & His Blog
• Equifax’s Instructions Are Confusing. Here’s What to Do Now. – The New York Times
• ML Engine BigQuery: Request had insufficient authentication scopes – Stack Overflow
• Google Cloud Summit in Seattle
• Performance RNN: Generating Music with Expressive Timing and Dynamics
• workshops/07_structured_data.ipynb at master · tensorflow/workshops
  Josh’s datasets example
• Misc/Blog_Estimators_DataSet.py at master · mhyttsten/Misc
  datasets/estimators blog post — https://developers.googleblog.com/2017/09/introducing-tensorflow-datasets.html
• models/detection_model_zoo.md at master · tensorflow/models
  We provide a collection of detection models pre-trained on the COCO dataset. These models can be useful for out-of-the-box inference if you are interested in categories already in COCO (e.g., humans, cars, etc). They are also useful for initializing your models when training on novel datasets.
• Google Developers Blog: Introduction to TensorFlow Datasets and Estimators

Digest powered by RSS Digest

• tensorflow-mnist-tutorial/README_BATCHNORM.md at a14af720f9631d638f3f2521b153b43bafcbf1a4 · martin-gorner/tensorflow-mnist-tutorial
• FarMcKon/gitmarks_2: Gitmarks 2.0. is a peer to peer bookmarking tool.
  Gitmarks 2.0. is a peer to peer bookmarking tool. http://farmckon.github.com/gitmarks_2
• Twitter
  RT @EdFelten: The White House Office of Science and Technology Policy home page is currently blank. Entire OSTP site is nearly em…
• Jennie Darlington, participant in groundbreaking Antarctic exploration, dies at 93 – The Washington Post
• Chatbot lets you sue Equifax for up to $25,000 without a lawyer | The Verge
• Equifax’s Maddening Unaccountability – The New York Times
• Hillary Clinton Opens Up About ‘What Happened,’ With Candor, Defiance and Dark Humor – The New York Times
• Gabrielle Deydier: what it’s like to be fat in France | Society | The Guardian
• Words in Air: The Complete Correspondence Between Elizabeth Bishop and Robert Lowell:Amazon:Books
• Twitter
  RT @shane_bauer: Aside from two paragraphs in a Washington Post article, there’s still been zero national coverage of this.
• The Spencerfield Spirit Co Edinburgh Gin’s Rhubarb and Ginger Liqueur 500ml | KWMWine.com
• The Falling Man
• My friend David Bowie by Hanif Kureishi | Books | The Guardian
• Uranus Opens And Closes Every Day To Let Out Planet’s Solar Wind, Study Finds
• Individuality drives collective behavior of schooling fish – Scienmag: Latest Science and Health News
• hunkim/DeepLearningZeroToAll: TensorFlow Basic Tutorial Labs
  TensorFlow Basic Tutorial Labs https://www.youtube.com/user/hunkims
• Performing prediction with TensorFlow object detection models on Google Cloud Machine Learning Engine | Google Cloud Big Data and Machine Learning Blog  |  Google Cloud Platform
  “Performing prediction with TensorFlow object detection models on Google Cloud Machine Learning Engine”—
• Why developer evangelism is the secret to the success of Kubernetes – TechRepublic
• Twitter
  RT @sarahljaffe: fast food workers in the first ever Fight for $15 strike told me they were radicalized by having to work despite Hu…
• The Wall Street Journal’s Trump problem | Media | The Guardian
  RT @ritholtz: WSJ losing reporters as current + ex-staffers show outrage over pressure from management to normalize Trump

Digest powered by RSS Digest