Take the lessons of open source and apply them across your processes, not just to development.
Open source increasingly touches every corner of the technology industry. Virtually all tech companies now use open source in their products, many participate in external open source projects, and increasingly companies are starting new open source projects to serve business goals. Everywhere we look, we see more open source and deeper involvement.
We also see a lot of misconceptions and missed opportunities.
Continue reading Open source mistakes for enterprise newcomers.(image)
FM Backscatter, 20 People, Information Operations, and Techniques of Magic
Continue reading Four short links: 28 April 2017.(image)
Creative deep neural networks, AI black box, robot food delivery, and brute force productivity.
Continue reading Intelligent Bits: 28 April 2017.(image)
The O’Reilly Design Podcast: Leadership, the design of product teams, and hiring optimists.
This week, I sit down with Nate Walkingshaw, chief experience officer of Pluralsite and co-author of Product Leadership. We talk about hard and soft leadership skills, building cross-disciplinary product teams, and why it’s important to use the layover test when hiring.
Open Source Mail Delivery, Superhuman AI, Probabilistic Graphical Models, and Golden Ages
Continue reading Four short links: 27 April 2017.(image)
Achieve high scalability and performance while reducing system complexity.
The C10k problem was an area of research and optimization that tried to achieve 10,000 concurrent connections on a single commodity server. Even these days, solving this engineering task with the traditional Java toolkit is a challenge. There are many reactive approaches that easily achieve C10k, and RxJava makes them very approachable. In this chapter, we explore several implementation techniques that will improve scalability by several orders of magnitude. All of them will circle around the concept of reactive programming. If you are lucky enough to work on a greenfield project, you might consider implementing your application in a reactive manner top to bottom. Such an application should never synchronously wait for any computation or action. The architecture must be entirely event-driven and asynchronous in order to avoid blocking. We will go through several examples of a simple HTTP server and observe how it behaves with respect to design choices we made. Admittedly, performance and scalability does have a complexity price tag. But with RxJava the additional complexity will be reduced significantly.
The classic thread per connection model struggles to solve the C10k problem. With 10,000 threads we do the following:
Continue reading Designing a reactive HTTP server with RxJava.(image)
Learn to make your sites and apps accessible to all users with this Learning Path at Fluent 2017.
“Building a better web” is the theme of Fluent this year, and hopefully it’s the goal of everyone who attends. But how do you define “a better web”? Some of the building blocks are no-brainers to most of us: the web should be fast, available, and secure. But if we build websites that are literally unusable for a huge number of people, then speed, availability, and security don’t mean much.
Continue reading Putting web accessibility front and center.(image)
Learn computer worm malware inside and out by building your own.
Continue reading How do I build my own computer worm for penetration testing?.(image)
Learn how you can use OpenVAS to scan your network for hosts and fingerprint their listening services to obtain access.
Continue reading How can I conduct a vulnerability scan of my network using OpenVAS?.(image)
Learn how you can use Nmap to scan your network to find out which services and hosts are listening and may be vulnerable to compromise.
Continue reading How can I scan my network using Nmap?.(image)
The O’Reilly Security Podcast: Scaling machine learning for security, the evolving nature of security data, and how adversaries can use machine learning against us.
In this special episode of the Security Podcast, O’Reilly’s Ben Lorica talks with Parvez Ahammad, who leads the data science and machine learning efforts at Instart Logic. He has applied machine learning in a variety of domains, most recently to computational neuroscience and security. Lorica and Ahammad discuss the challenges of using machine learning in information security.
Continue reading Parvez Ahammad on applying machine learning to security.(image)
Diogo Almeida examines the capabilities and challenges in deep learning.
Continue reading Deep learning: Modular in theory, inflexible in practice.(image)
Information Asymmetry, Startup Simulator, HTTP Filter, and Inside Juicero's Hardware
Continue reading Four short links: 26 April 2017.(image)
Building transparency and individual choice into IoT security.
For IoT security to be successful, there needs to be an effective way to reason about how humanity can trust the security, safety, and privacy of this massive transformation of the world. Most importantly, “ordinary people,” whether they are consumers or workers, must be able to safely, reliably, and intuitively interact with vast, complex, interconnected systems of IoT devices. It can be overwhelming to think about all the ways individuals and society can be damaged by the haphazard engineering of systems that merge the physical and digital worlds. Technologists have done a terrible job with security technology so far, yet now we are about to impose those failures onto the physical world on a scale that only ubiquitous, pervasive, even invasive computing and connectivity can accomplish. Continuing the status quo is unsustainable.
The IoT can be thought of as a hyper-connected, hyper-distributed collection of resources. The complex ecosystem surrounding IoT devices means trusting them will not be intuitive. These connected devices can potentially be controlled and observed by others anywhere on the planet. For example, before the IoT, it was always easy to physically check the locks on your doors and decide to trust those who had the keys. Now with Internet connected “smartlocks,” you can check or alter their state from anywhere. How can an “ordinary person” track who has the electronic key and discern that the software controlling the lock is secure and resistant to hacker attacks? A February 2017 survey of IoT consumers showed that 72% were not sure how to check if their devices had been compromised.
Continue reading A human-centric trust model for the Internet of Things.(image)
Citizen Neuroscience, Counter-Drone Techniques, Cloud Vision Illusions, and Advanced R
Continue reading Four short links: 25 April 2017.(image)
Learn how to perform a security assessment on a MySQL database with InSpec.
Learn how to perform security assessments with InSpec over SSH.
Learn how to integrate InSpec and detect weaknesses in your Docker container.
Python cheat sheet, open source DL guide, Keen IO, and digital signal processing.
Introduction to Artificial Neural Networks and Deep Learning: A Practical Guide with Applications in Python. If you're wondering what deep learning is all about, this open source guide by Sebastian Raschka (@rasbt) hits all the right notes (TensorFlow, RNN, etc). It's still in the early stages (mostly just appendices about math), but if the chapters are close to this quality, it's going to be a great book.
Analyzing your Keen IO data with Jupyter Notebooks. Keen IO is a handy analytics package that you can embed in your website or mobile application to track event data. In addition to the easy-to-use APIs, it also has a nice front-end app where you can do queries and get essential information. We've used it almost since it came out at O'Reilly, and it just keeps getting better and better. However, while the dashboards are nice and keep the folks back home happy, sometimes you just want to throw that stuff into Pandas and go deeper. This refreshingly concise article by Joanne Cheng (@joannecheng) shows you how, and has a nice companion notebook with details.
Cacophony for the whole family. Everything Allen Downey (@AllenDowney) writes is a model for how to explain a complex subject in clear, compelling language. But, he's also really funny. This excerpt from Think DSP, which shows how to use digital signal processing to simulate an elementary school band, resonated with me because I've been listening to my kids practice on their recorders. This article nails the experience. Plus, it's a cool example of how to generate audio files with python in the notebook. (Be sure to keep your volume low, though!)
Continue reading Jupyter Digest: 24 April 2017.(image)
Sports Analytics, Book Recommendations, Breakthrough Tech, and Engineering Leadership
Continue reading Four short links: 24 April 2017.(image)
Glare-Free, People Suck, Robots Swing, and This Cute Hack Wants You to Be Happy
Continue reading Four short links: 21 April 2017.(image)
This excerpt from Jake VanderPlas' Python Data Science Handbook
This excerpt from Jake VanderPlas' Python Data Science Handbook
There are several Python libraries which provide solid implementations of a range of machine learning algorithms. One of the best known is Scikit-Learn, a package that provides efficient versions of a large number of common algorithms. Scikit-Learn is characterized by a clean, uniform, and streamlined API, as well as by very useful and complete online documentation. A benefit of this uniformity is that once you understand the basic use and syntax of Scikit-Learn for one type of model, switching to a new model or algorithm is very straightforward.
This section provides an overview of the Scikit-Learn API; a solid understanding of these API elements will form the foundation for understanding the deeper practical discussion of machine learning algorithms and approaches in the following chapters.
Continue reading Introduction to scikit-learn.(image)
Kenny Daniel on implementing neural networks in production.
Continue reading Practicalities of employing deep learning at scale.(image)
The O’Reilly Data Show Podcast: Reza Zadeh on deep learning, hardware/software interfaces, and why computer vision is so exciting.
In this episode of the Data Show, I spoke with Reza Zadeh, adjunct professor at Stanford University, co-organizer of ScaledML, and co-founder of Matroid, a startup focused on commercial applications of deep learning and computer vision. Zadeh also is the co-author of the forthcoming book TensorFlow for Deep Learning (now in early release). Our conversation took place on the eve of the recent ScaledML conference, and much of our conversation was focused on practical and real-world strategies for scaling machine learning. In particular, we spoke about the rise of deep learning, hardware/software interfaces for machine learning, and the many commercial applications of computer vision.
Continue reading Scaling machine learning.(image)
Phillip Hunter discusses the reasons why voice-driven experiences are now prevalent.
Continue reading Why Alexa and why now.(image)
2017-04-20T10:45:00ZThree models for how automakers could partner with fleet operating companies to provide autonomous vehicles for on-demand mobility.In my book, The Big Data Opportunity in Our Driverless Future, I make two arguments: 1) that societal and urban challenges are accelerating the adoption of on-demand mobility, and 2) technology advances, including big data and machine intelligence, are making Autonomous Connected and Electrified (ACE) vehicles a reality. ACE vehicles and on-demand mobility will cause three major shifts that can lead to the disruption of the automotive and transportation industries: a consumer shift, an automotive industry shift, and a mobility services shift. In this post, I examine what is causing these shifts, the value chain that is emerging as a result of these shifts, big data’s key role in the value chain, and the models being created around this value chain. Shifts in personal mobility and technology open the door to ACE vehicles Changes in personal mobility Let’s begin by reviewing the most important challenges contributing to changes in personal mobility. One key challenge is the fact that urbanization is increasing, and more megacities are being created. According to a UN report in 2014, 54% of the population lived in urban areas, and by 2050, an additional 2.5 billion people will be added to these areas. Another challenge impacting personal mobility is traffic congestion. Particularly in megacities, congestion is severely impacting individual productivity because the transportation infrastructure has reached, or is reaching, capacity. We spend too much time commuting to work or home, and when we arrive at our destinations, we aren’t productive. Pollution and climate change are also impacting the quality of our lives, particularly in cities (and here). Transportation contributes 28% of greenhouse gases (50% of that coming from passenger cars and light-duty vehicles). Transportation now regularly emits more earth-warming gases into the atmosphere than any other sector, according to the federal Energy Information Administration. Last year, transportation surpassed the electric power sector for the first time since the late 1970s in terms of polluting culprits. Traffic-related pollution is negatively impacting the quality of life in megacities, as cities in Asia and Europe are finding out. Population is another factor affecting personal mobility. The population of many developed countries is aging fast. These populations will require constant assistance in various forms, including transportation assistance, in order to continue functioning properly. Lastly, the socioeconomic conditions of certain population segments lead them to adopt the sharing economy to address many of their needs, including their transportation needs. Millennials are leading the way in this adoption. On-demand mobility services—specifically services such as ride-hailing, ridesharing, bike-sharing, car-sharing, and various forms of car rental—are seen as a particularly promising way of addressing these challenges. Changes in technology At the same time, five technological advances are leading to the development of Autonomous, Connected and Electrified (ACE) vehicles. The first is the easy collection and management of data from sensors that are incorporated into the vehicle, and from specialized [...]
Account Takeover, Trello Inspiration, Forced Obsolescence, and Technology Caution
Continue reading Four short links: 20 April 2017.(image)
Learn what Terraform does, and how it compares to Chef, Puppet, Ansible, CloudFormation, and other tools.
Software is not done when the code is working on your computer. It’s not done when the tests pass. And it’s not done when someone gives you a "ship it" on a code review. Software isn’t done until you deliver it to the user.
Software delivery consists of all the work you need to do to make the code available to a customer, such as running that code on production servers, making the code resilient to outages and traffic spikes, and protecting the code from attackers. Before you dive into the details of Terraform, it’s worth taking a step back to see where Terraform fits into the bigger picture of software delivery.
Continue reading Why use Terraform?.(image)
How service workers, HTTPS, and other techniques can help you achieve security and speed.
Security and Performance are both considered to be two separate goals, as one can often negatively impact the other. We see various security solutions — whether implemented at the frontend or backend — slowing down the performance of sites. And vice versa: we see certain performance techniques leaving sites more vulnerable to security threats. In the face of these challenges we need to make sure that our sites are adapting to the needs of the customers.
Every year, we see an increase in security attacks in many different areas (for example, in finance and healthcare) and we can only expect this rise in threats to continue with the Internet of Things (IoT). This increase is can be attributed to an escalation in web traffic, with resulting high volume/high load situations which motivate us to more closely focus on performance. It is also due to the growth in the number of third party resources being leveraged on sites today — not only do they introduce potential security holes such as delivering malware and/or malicious code from third party attackers, but they can slow down the site and delay content parsing/loading.
Continue reading Fast and safe frontend fixes.(image)
Most video game designers dream of designing a blockbuster hit, but what makes a game break out?
The video game industry is broadly regarded as “hit driven” and this defines it not just for investors and companies, but for developers downstream as well. (By “video games”, I mean interactive games for any platform: mobile phone, PC, console, etc.) Engineers, artists, and designers working on games are always trying to pick winning development teams — and it’s amazing how much of the development process we just can’t control.
In seventeen years making games (and 2 years lecturing and teaching about them), I’ve seen a few misconceptions about creating blockbuster video games pop up repeatedly regarding what they must contain, and almost as importantly, what they should not contain. In this article, I’ll explain seven necessary ingredients for a blockbuster video game, divided into controllable and uncontrollable variables. Understanding these variables will help you focus on the ones you can control and let go of the ones you can't. And who knows — after reading this list, you may decide you're not interested in trying to create a blockbuster because doing so would require too many compromises1. Either way, you’ll go into the design process with your eyes open.
Continue reading 7 secrets of blockbuster video games.(image)
Allen Downey takes on elementary school band camp in this fun application of Digital Signal Processing.
Allen Downey takes on elementary school band camp in this fun application of Digital Signal Processing.
This is an example that demonstrates some of the features in the Think DSP library.
Continue reading Cacophony for the whole family.(image)
SRE calls for a unique blend of skills, which makes team building and hiring difficult. Learn how LinkedIn addressed these problems with their own SRE team.
Continue reading 3 things that helped LinkedIn grow their SRE team.(image)
Learn how to access a server from any web browser using Windows 2016 Server Management Tools.
Continue reading How can I effectively use the Windows 2016 Server Management Tools?.(image)
Learn how to perform hot-add/remove in Windows Server 2016 Hyper-V and how it can help you achieve more granular control over memory.
Learn how to deploy Storage Spaces Direct to enable local storage as clustered storage.
Continue reading How can I create a Storage Spaces Direct instance?.(image)
2017-04-19T11:15:00ZMiroculus democratizes early cancer detection with an open research database and a digital microfluidic platform. Early detection can completely change the trajectory of a cancer diagnosis. We all know by now that the earlier cancer is found, the better. Jorge Soto, the CTO of Miroculus, said in his 2014 TED talk, “Catching cancer early is the closest thing we have to a silver bullet cure against it.” When Soto formed a highly interdisciplinary team at Singularity University, they founded their company around this goal. As cancer develops, it leaves a breadcrumb trail of mutation or disruption. If you are looking for a cancer diagnostic, there are a plethora of options to choose from. You can look at protein expression, DNA mutations, or different types of RNA (noncoding RNA, cRNA, etc.). As it develops, this cancer trail leaks into the bloodstream or urine, where it is easier to sample. When searching for the right diagnostic, you are typically looking for three attributes: Is it easy to get? Is it easy to detect? Is it reliable? Many researchers and companies, including Miroculus, believe that extracellular microRNAs (miRNAs) fit the bill. miRNAs are class of noncoding RNAs that regulate various cellular processes (like differentiation and replication). They make for an attractive diagnostic because they are small, stable, and present in blood and plasma. The promise of miRNAs for detection is controversial, however. While the best biomarker would be present (or absent) in a disease state compared to normal, miRNAs are instead diagnostic in their concentration. For example, a particular miRNA may be below a certain level normally, but exceed that threshold if the person has a certain type of cancer. To be proven reliable, all other factors that could lead to a significant change in concentration without the disease present need to be eliminated (for example, diet, exercise, or other medications). Also, it’s been hard to compare miRNA concentrations using different platforms, or even the same platform with different vendors’ products. It’s difficult to determine exactly how many miRNAs have been approved to date for cancer diagnostics, but to put it simply, few to none Still, this technique holds incredible promise. With over 2,000 miRNAs in the human genome, it could be possible to develop an miRNA “fingerprint” for any number of diseases (not just cancer). So Miroculus built Loom, an miRNA data engine, so that its technology can evolve alongside the highly dynamic miRNA research field. Loom is an interactive site (updated monthly) that summarizes literature by miRNA, gene, or condition (Figure 1-1). They’ve chosen to make this tool open-access for the community, and it plays a key role for the company as well. Instead of funneling resources into a secret discovery effort, Miroculus can use Loom to identify key players in the field for collaboration in addition to its own R&D. Since cancer is one of the most highly funded areas of research, there are plenty of publicly available findings to draw upon. It’s important to them that this tool gives back [...]
Featured Snippets, Text Summarization, VC's Business Model, and Aphyr's Tools
Continue reading Four short links: 19 April 2017.(image)
Learn how to tokenize, breaking a sentence down into its words and punctuation, using NLTK and spaCy.
In this video, Jonathan Mugan will show you two different ways to tokenize a sentence using two popular Python libraries. First, you will learn how to use the word_tokenize function from the NLTK library. Next, Jonathan explains how to accomplish the same tokenization using the spaCy library. This lesson is designed for Python programmers that need to know how to tokenize data for natural language text processing purposes.
Continue reading How can I tokenize a sentence with Python?.(image)
Learn how to use the gensim Python library to determine the similarity between two or more documents.
In this training video, Jonathan Mugan will introduce you to the gensim library; and how you can use it to find the similarities between documents. You will go through the process, step by step, starting with tokenizing a list of documents to create a dictionary, and use that to create a corpus and then a tf-idf (term frequency-inverse document frequency) object. You will also examine a couple of techniques to compare documents, and finish up by working on exercises based on what you have learned. This video is designed for programmers familiar with Python who would like to learn more about using gensim for natural language text processing.
Continue reading How do I compare document similarity using Python?.(image)
Find out what pays and what doesn't for software engineers, developers, and other programming professionals.
This report explores the landscape of the professionals working in all facets of software development, including details about the relationship between roles, location, company size, industry, and earnings. The results are based on more than 6,800 responses collected via an online survey. We paid special attention to the variables that correlate with salary, but it’s not just about money: we also analyzed what tools, tasks, and organizational processes respondents most commonly use.
In this, our second annual Software Development Salary Survey, we find some consistency in what matters to software developers. Much like last year, our results show that the better-paying jobs tend to concentrate in tech centers, that experience matters more than age, and that knowing more tools, working with more people in a wider variety of roles, and working for larger organizations all correlate with higher wages. And, the data shows that knowing when to hold ’em and when to fold ’em (i.e., self-reported good negotiating skills) might be a key to higher salaries.
Continue reading 2017 Software Development Salary Survey.(image)
Word Processors, Open Data, Robot-Proof Jobs, and New Coder Con
Continue reading Four short links: April 18.(image)
Learn how to use spaCy to parse a sentence to return the parts of speech (noun, verb, etc.) and dependencies.
Jonathan Mugan shows you how to use the spaCy Python library to parse text; and show the structure and dependencies of a sentence. You will also learn how to retrieve the most important word (or the "head") of a sentence using the get_head_of_sentence function. This tutorial is designed for programmers familiar with Python, that would like to learn more about using the spaCy library for natural language text processing.
Continue reading How do you find the parts of speech in a sentence using Python?.(image)
Reproducibility, TensorFlow examples, the new NBA, and 30,699 Kobe Bryant shots.
Continue reading Jupyter Digest: 17 April 2017.(image)
2017-04-17T10:30:00ZProduct management is the connective tissue for identifying, building, and shipping products customers want.A great product manager can help align customer needs with business goals, inspire developers and designers, and make critical connections across functions and silos. But the actual work that product managers do is harder to pin down than that of their counterparts in engineering and design. Given the ambiguity and variability around the role, what can organizations do to hire great product managers, and set them up for success? Understand your company’s product management needs Before posting a job listing or looking for internal candidates, take the time to think through why you need product managers in the first place. Do you feel your current product direction is not achieving its intended business goals? Are you struggling to articulate those goals in the first place? Do you feel the efficiency of your design and engineering team could be improved? In some cases, this exercise may lead you to realize that what you need is not actually a product manager at all. If, for example, you feel your team is unequivocally building the right things but needs to work faster, hiring another developer might increase your overall speed more than hiring a product manager would. Broadly speaking, if you simply want to improve the performance or output of a single role (designer, developer, researcher), a product manager might not be a good fit. If, however, you want to improve the way these roles align and collaborate, a product manager would likely be a critical addition to your team. To learn more about the impact and benefits of product management, read Why is Product Management so Relevant Today? To understand the role of product manager and how product management can ensure product market fit, read Chapter 1 of Making it Right. Identify candidates with strong connective skills Some organizations are well known for favoring a certain “profile” for product management candidates. Amazon, for example, looks for MBAs. Google, on the other hand, prefers candidates with a computer science degree from a prestigious university. Generally speaking, the “classic” profile for a product manager is either a technical person with some business savvy, or a business-savvy person who will be able to earn the trust and respect of developers. Recruiting against these profiles can seem like the easiest and safest approach. After all, how wrong can Google and Amazon be about hiring for a role that these organizations helped define? But the pool of candidates that fits this “classic” profile can be frustratingly small—and while they might fit the profile on paper, they sometimes struggle with the actual day-to-day demands of the job.Some product managers may have design, engineering, or business backgrounds, but the best product managers possess the [...]
Emoji Detail, Macintosh Emulation, In-Browser Editors, and Furless Tickle Me Elmo
Continue reading Four short links: 17 April 2017.(image)
Visual Effects, Political Economics, Building Software, and Low-Level Programming
Continue reading Four short links: 14 April 2017.(image)
June Andrews talks about simple, cost-effective algorithmic computing at scale.
Continue reading How Pinterest uses humans in the loop to analyze clusters.(image)
The O’Reilly Design Podcast: Asking the right questions, conducting research in an agile environment, and conscious confidence.
In this week’s Design Podcast, I sit down with David Farkas, associate director of user experience at EPAM and co-author of the book UX Research. We talk about his book, why everyone should learn to conduct research, and how to open up your mind to ask the right questions.
Farkas and his co-author Brad Nunnally also are teaching a series of online courses:
Continue reading David Farkas on how to approach user research.(image)
Song Han on compression techniques and inference engines to optimize deep learning in production.
Continue reading Shrinking and accelerating deep neural networks.(image)
Messaging as the operating system for the enterprise.
In this episode of the Bots Podcast, we peer into the giant companies that are beginning to adopt messaging and bots. My guest is Tom Hadfield, founder of Message.io, a service that syndicates bots across many different messaging platforms.
Continue reading Tom Hadfield on bots in the enterprise.(image)
Make Your Own Phone, Awful Onboarding UX, Institutionalizing New Systems, and WebGL for Visualization
Continue reading Four short links: 13 April 2017.(image)
Get practical knowledge on the advantages microservices can bring to your project.
Continue reading Why would you use microservices?.(image)
The O’Reilly Security Podcast: The five stages of vulnerability disclosure grief, hacking the government, and the pros and cons of bug bounty programs.
In this episode, I talk with Katie Moussouris, founder and CEO of Luta Security. We discuss the five stages of vulnerability disclosure grief, hacking the government, and the pros and cons of bug bounty programs.
Continue reading Katie Moussouris on procuring and processing bug reports.(image)
Kurt Brown discusses services in use, such as Genie, Metacat, Charlotte, and Microbots.
Continue reading Stitching data processing engines at Netflix.(image)
Everyday citizens are becoming empowered to contribute to modern medical science.
From the Mylan EpiPen pricing scandal, to the whistleblower story that crashed the blood-testing startup Theranos, among many Americans, there is a growing public distrust in governance over the biomedical enterprise and there are questions being raised about who gets access to cutting-edge sophisticated drugs and therapies.
At the same time, there’s a parallel story brewing about citizens who decide not to wait to shape their own medical future. One of them is Tal Golesworthy, a bright and resolved engineer who, suffering from a genetic disease that damages his heart, designed a surgical device that would save him and other patients from a more risky procedure. Dana Lewis, a digital communication specialist suffering from Type 1 diabetes, created an artificial pancreas based on an algorithm that calculates the need for insulin based on a patient’s blood sugar levels. And to find a cure for their daughters suffering of the rare Batten disease, a couple raised millions on a crowdfunding platform to hire their own research team. While these individuals and other communities are reshaping their involvement in health research and practice, they are raising new ethical, safety, and governance issues for policymakers, practitioners, and patients.
Continue reading Citizen health innovators: Exploring stories of modern health.(image)
2017-04-12T11:00:00ZA closer look at the reasoning inside your deep networks.This article is a gentle introduction to attentional and memory-based interfaces in deep neural architectures, using TensorFlow. Incorporating attention mechanisms is very simple and can offer transparency and interpretability to our complex models. We conclude with extensions and caveats of the interfaces. As you read the article, please access all of the code on GitHub and view the IPython notebook here; all code is compatible with TensorFlow version 1.0. The intended audience for this notebook are developers and researchers who have some basic understanding of TensorFlow and fundamental deep learning concepts. Check out this post for a nice introduction to TensorFlow. Applying selective attention to deep learning Attentional interfaces in deep neural networks are loosely based on visual attention mechanisms, found in many animals. These mechanisms allow the organisms to dynamically focus on pertinent parts of a visual input and respond accordingly. This basic idea of selective attention has been carried over to deep learning, where it is being used in image analysis, translation, question/answering, speech, and a variety of other tasks. Figure 1. Attention on an image for a specific (underlined) word in the caption. Credit: Kelvin Xu, Jimmy Ba, Ryan Kiros, Kyunghyun Cho, Aaron Courville, Ruslan Salakhutdinov, Richard Zemel: “Show, Attend and Tell: Neural Image Caption Generation with Visual Attention”. Used with Permission. These interfaces also offer much needed model interpretability, by allowing us to see which parts of the input are attended to at any point in time. A common disadvantage with deep neural architectures is the lack of interpretability and the associated "black box" stigma. Implementation of these interfaces has not only been shown to increase model performance, but also offer more transparent and sensible results. And as you will see in our implementation below, they produce some interesting visualizations that are consistent with how we would attend to the inputs. Research on attentional interfaces is very popular these days because they offer multiple benefits. There are increasingly complex variants to the attention mechanisms, but the overall foundation remains the same. Figure 2. Attention heatmaps offering interpretability on where the model is looking to respond. Credit: Karl Moritz Hermann, Tomáš Kočiský, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman: “Teaching Machines to Read and Comprehend”. Used with Permi[...]
2017-04-12T10:10:00Z3D-Printed Titanium, Searching Cellphones, Augmented Reality, and Educational Virtual Robots Boeing To Use 3D-Printed Titanium Parts on the Dreamliner -- Norsk expects the U.S. regulatory agency will approve the material properties and production process for printed parts later this year. That will "open up the floodgates," Yates said, by allowing Norsk to print thousands of other parts for each Dreamliner, without each part requiring separate FAA approval, resulting in millions in expecting savings per plane. [...] General Electric Co is already printing metal fuel nozzles for aircraft engines. But Norsk and Boeing said the titanium parts are the first printed structural components designed to bear the stress of an airframe in flight. (via Slashdot) How the Denver Police Crack and Search Cellphones (Vice) -- DPD's Investigative Technology Bureau has plans for just about any eventuality in fact, with rules ordering the battery be removed to guard against remote data deletion, all chargers and wires be seized with the phone, and perhaps the most important of all: don't start digging through the phone on your own because it is a) illegal and b) you are gonna destroy evidence. The First Decade of Augmented Reality (Benedict Evans) -- the more you think about AR as placing objects and data into the world around you, the more this becomes an AI question as much as a physical interface question. What should I see as I walk up to you in particular? LinkedIn or Tinder? When should I see that new message—should it be shown to me now or later? Do I stand outside a restaurant and say 'Hey Foursquare, is this any good?' or does the device's OS do that automatically? How is this brokered—by the OS, the services that you've added, or by a single 'Google Brain' in the cloud? Google, Apple, Microsoft, and Magic Leap might all have different philosophical attitudes to this. Robotopia -- Introducing kids to coding with tiny virtual robots. Continue reading Four short links: 12 April 2017.[...]
Dudley Storey and Sarah Drasner explain why SVG has become a fundamental technology.
Leading up to the release of their books, I asked two of our authors, Dudley Storey (Using SVG with CSS3 and HTML5) and Sarah Drasner (SVG Animations) a few questions about SVG (Scalable Vector Graphics). SVG is a fundamental technology that has far more depth than most people realize, and the scope of their answers is revealing.
Sarah Drasner is an award-winning Speaker, Consultant, and Staff Writer at CSS-Tricks. She is also the co-founder of Web Animation Workshops, with Val Head.Sarah is formerly Manager of UX Design & Engineering at Trulia (Zillow). You can find Sarah on Twitter @Sara_Edo.
Continue reading SVG: 5 questions, 10 answers.(image)
Learn how uncoupling development from security using AWS Identity and Access Management can enhance security.
Continue reading How can I retrieve application secrets securely with AWS Lambda? .(image)
Learn how to call one Lambda from another AWS Lambda function in the AWS management console.
Continue reading How do I invoke a Lambda from another Lambda in AWS?.(image)