- Hidden Technical Debt in Machine Learning Systems (PDF) — We explore several ML-specific risk factors to account for in system design. These include boundary erosion, entanglement, hidden feedback loops, undeclared consumers, data dependencies, configuration issues, changes in the external world, and a variety of system-level anti-patterns.
- Large-Scale Content-Based Matching of Midi and Audio Files — We present a system that can efficiently match and align MIDI files to entries in a large corpus of audio content based solely on content, i.e., without using any metadata.
- Critical Social Research on Self-Tracking — I am currently working on an article that is a comprehensive review of both literatures, in the attempt to outline what each can contribute to understanding self-tracking as an ethos and a practice, and its wider sociocultural implications. Here is a reading list of the work from critical social researchers that I am aware of. Trigger warning: phrases like “The discursive construction of student subjectivities.”
- Warp-CTC — Baidu’s open source deep learning code. Connectionist Temporal Classification is a loss function useful for performing supervised learning on sequence data, without needing an alignment between input data and labels.
"audio" entries
Four short links: 18 January 2016
Machine Learning Technical Debt, Audio Matching, Self-Tracking Research, and Baidu's Open Source Deep Learning Code
Learning the Web
Finding a gentle entry to a big space
The Web welcomes, but it’s awfully big. While HTML, CSS, and JavaScript may all be appropriate entry points for newcomers who want to create, finding a solid starting point can be complicated. Social media has minimized the level of HTML and Web knowledge people need to start contributing, but when it’s time to make the jump, the Web offers perhaps too many options.
Part of the challenge is that HTML, CSS, and JavaScript may be the marquee technologies, but they’re not actually what hosts a website or app. Setting up a site requires an additional set of technologies, from domain names to hosting to web server choices. Setting up a site – long before you get to packaging an app! – requires mastering an additional technical toolset and vocabulary that will help you navigate where you need to put your projects. Our free report, Getting Started with the Web, provides the core foundations beginners need.
Those aren’t the only barriers, though. Read more…
Add depth to your project with practical web audio
Enhance the user experience with the thoughtful use of sound.
There is little debate that Web Audio is cool. Take for example Stepkit by Brent Jackson (embedded below).
It’s definitely a fun toy to play with, but most of us probably couldn’t think of how this might be relevant to our jobs. When I presented 8-bit game music with the Web Audio API at last year’s Fluent Conference, I readily admitted that it was intended to be purely fun rather than practical.
Recently I explored the idea of adding audio to web apps, but I think the big problem isn’t that web developers were unsure how to add audio to their app, but that they don’t think they should add audio to web apps. In this article, I’d like to make the case that you should be considering audio when designing your web application user interface.
Four short links: 19 August 2013
Aural Viz, SPOF ID, Information Asymmetry, and Support IA
- choir.io explained (Alex Dong) — Sound is the perfect medium for wearable computers to talk back to us. Sound has a dozen of properties that we can tune to convey different level of emotions and intrusivenesses. Different sound packs would fit into various contexts.
- Identity Single Point of Failure (Tim Bray) — continuing his excellent series on federated identity. There’s this guy here at Google, Eric Sachs, who’s been doing Identity stuff in the white-hot center of the Internet universe for a lot of years. One of his mantras is “If you’re typing a password into something, unless they have 100+ full-time engineers working on security and abuse and fraud, you should be nervous.” I think he’s right.
- What Does It Really Matter If Companies Are Tracking Us Online? (The Atlantic) — Rather, the failures will come in the form of consumers being systematically charged more than they would have been had less information about that particular consumer. Sometimes, that will mean exploiting people who are not of a particular class, say upcharging men for flowers if a computer recognizes that that he’s looking for flowers the day after his anniversary. A summary of Ryan Calo’s paper. (via Slashdot)
- Life Inside Brewster’s Magnificent Contraption (Jason Scott) — I’ve been really busy. Checking my upload statistics, here’s what I’ve added to the Internet Archive: Over 169,000 individual objects, totaling 245 terabytes. You should subscribe and keep them in business. I did.
Four short links: 15 August 2013
Audio Visualization, 3D Printed Toys, Data Center Computing, and Downloding Not Yet Beaten
- github realtime activity — audio triggered by github activity, built with choir.io.
- Makies Hit Shelves at Selfridges — 3d printing business gaining mainstream distribution. Win!
- The Datacenter as Computer — we must treat the datacenter itself as one massive warehouse-scale computer (WSC). We describe the architecture of WSCs, the main factors influencing their design, operation, and cost structure, and the characteristics of their software base. We hope it will be useful to architects and programmers of today’s WSCs, as well as those of future many-core platforms which may one day implement the equivalent of today’s WSCs on a single board. (via Mike Loukides)
- Illegal Downloads Not Erased By Simultaneous Release — Data gathered by TorrentFreak throughout the day reveals that most early downloaders, a massive 16.1%, come from Australia. Down Under the show aired on the pay TV network Foxtel, but it appears that many Aussies prefer to download a copy instead. The same is true for the United States and Canada, with 16% and 9.6% of the total downloads respectively, despite the legal offerings. Unclear whether this represents greater or less downloading than would have happened without simultaneous release.
Four short links: 12 June 2013
Geodata DVCS, Monitoring Stack, Robotic Roaches, and Audio Destress
- geogit — opengeo project exploring the use of distributed management of spatial data. […] adapts [git’s] core concepts to handle versioning of geospatial data. Shapefiles, PostGIS or SpatiaLite data stored in a change-tracking repository, with all the fun gut features for branching history, merging, remote/local repos, etc. BSD-licensed. First sound attempt at open source data management.
- Introducing Loupe — Etsy’s monitoring stack. It consists of two parts: Skyline and Oculus. We first use Skyline to detect anomalous metrics. Then, we search for that metric in Oculus, to see if any other metrics look similar. At that point, we can make an informed diagnosis and hopefully fix the problem.
- Bluetooth-Controlled Robotic Cockroach (Kickstarter) — ’nuff said. (via BoingBoing)
- Nature Sounds of New Zealand — if all the surveillance roboroach anomaly detection drone printing stories get to you, put this on headphones and recharge. (caution: contains nature)
Four short links: 16 March 2012
Squirrel Targeting with Computer Vision, Audio Recognition, Single Page Apps, and Persisting at Failing
- Militarizing Your Backyard With Python and Computer Vision (video) — using a water cannon, computer video, Arduino, and Python to keep marauding squirrel hordes under control. See the finished result for Yakkity Saxed moist rodent goodness.
- Soundbite — dialogue search for Apple’s Final Cut Pro and Adobe Premiere Pro. Boris Soundbite quickly and accurately finds any word or phrase spoken in recorded media. Shoot squirrels with computer vision, search audio with computer hearing. We live in the future, people. (via Andy Baio)
- Single Page Apps with Backbone.js — interesting and detailed dissection of how one site did it. Single page apps are where the server sends back one HTML file which changes (via Javascript) in response to the user’s activity, possibly with API calls happening in the background, but where the browser is very definitely not requesting more full HTML pages from the server. The idea is to have speed (pull less across the wire each time the page changes) and also to use the language you already know to build the web page (Javascript).
- Why Finish Books? (NY Review of Books) — the more bad books you finish, the fewer good ones you”ll have time to start. Applying this to the rest of life is left as an exercise for the reader.
Four short links: 3 November 2011
Getting Feedback, Colour Design, Discovering Musicians, Weather Prediction App
- Feedback Without Frustration (YouTube) — Scott Berkun at the HIVE conference talks about how feedback fails, and how to get it successfully. He is so good.
- Americhrome — history of the official palette of the United States of America.
- Discovering Talented Musicians with Musical Analysis (Google Research blgo) — very clever, they do acoustical analysis and then train up a machine learning engine by asking humans to rate some tracks. Then they set it loose on YouTube and it finds people who are good but not yet popular. My favourite: I’ll Follow You Into The Dark by a gentleman with a wonderful voice.
- Dark Sky (Kickstarter) — hyperlocal hyper-realtime weather prediction. Uses radar imagery to figure out what’s going on around you, then tells you what the weather will be like for the next 30-60 minutes. Clever use of data plus software.
Four short links: 31 October 2011
Solitude and Leadership, Data Repository, Copyright History, and Open Source Audio
- Solitude and Leadership — an amazing essay on the value of managing one’s information diet. Far more than yet another Carr/Morozov “the Internet is making us dumb!!” hate on short-form content, this is an eloquent exposition of the need for long-form thoughts. I find for myself that my first thought is never my best thought. My first thought is always someone else’s; it’s always what I’ve already heard about the subject, always the conventional wisdom. It’s only by concentrating, sticking to the question, being patient, letting all the parts of my mind come into play, that I arrive at an original idea. By giving my brain a chance to make associations, draw connections, take me by surprise. And often even that idea doesn’t turn out to be very good. I need time to think about it, too, to make mistakes and recognize them, to make false starts and correct them, to outlast my impulses, to defeat my desire to declare the job done and move on to the next thing. (via Best American Nonrequired Reading 2011)
- Building The Perfect Data Repository (Cameron Neylon) — in which Cameron talks about solving problems for the people with the data. One of the problems with many efforts in this space is how they are conceived and sold as the user. “Making it easy to put your data on the web” and “helping others to find your data” solve problems that most researchers don’t think they have. […] A successful data repository system will start by solving a different problem, a problem that all researchers recognize they have”
- Macaulay on Copyright — periodically someone rediscovers how the the 1841 debate on copyright mirrors our own, but that it was discovered before does not mean it is not worth reading again. At present the holder of copyright has the public feeling on his side. Those who invade copyright are regarded as knaves who take the bread out of the mouths of deserving men.[…] Pass this law: and that feeling is at an end. Men very different from the present race of piratical booksellers will soon infringe this intolerable monopoly. Great masses of capital will be constantly employed in the violation of the law. Every art will be employed to evade legal pursuit; and the whole nation will be in the plot.
- ALAC — Apple Lossless Audio Codec is now open source by Apple.
Four short links: 12 April 2011
Email Game, Faster B Trees, RFID+Projectors, and Airport Express Broken
- The Email Game — game mechanics to get you answering email more efficiently. Can’t wait to hear that conversation with corporate IT. “You want us to install what on the Exchange server?” (via Demo Day Wrapup)
- Stratified B-trees and versioning dictionaries — A classic versioned data structure in storage and computer science is the copy-on-write (CoW) B-tree — it underlies many of today’s file systems and databases, including WAFL, ZFS, Btrfs and more. Unfortunately, it doesn’t inherit the B-tree’s optimality properties; it has poor space utilization, cannot offer fast updates, and relies on random IO to scale. Yet, nothing better has been developed since. We describe the `stratified B-tree’, which beats all known semi-external memory versioned B-trees, including the CoW B-tree. In particular, it is the first versioned dictionary to achieve optimal tradeoffs between space, query and update performance. (via Bob Ippolito)
- DisplayCabinet (Ben Bashford) — We embedded a group of inanimate ornamental objects with RFID tags. Totems or avatars that represent either people, products or services. We also added RFID tags to a set of house keys and a wallet. Functional things that you carry with you. This group of objects combine with a set of shelves containing a hidden projector and RFID reader to become DisplayCabinet. (via Chris Heathcote)
- shairport — Aussie pulled the encryption keys from an Airport Express device, so now you can have software pretend to be an Airport Express.