"machine learning" entries

Four short links: 24 March 2014

Four short links: 24 March 2014

Google Flu, Embeddable JS, Data Analysis, and Belief in the Browser

  1. The Parable of Google Flu (PDF) — We explore two
    issues that contributed to [Google Flu Trends]’s mistakes—big data hubris and algorithm dynamics—and offer lessons for moving forward in the big data age.
    Overtrained and underfed?
  2. Duktape — a lightweight embeddable Javascript engine. Because an app without an API is like a lightbulb without an IP address: retro but not cool.
  3. Principles of Good Data Analysis (Greg Reda) — Once you’ve settled on your approach and data sources, you need to make sure you understand how the data was generated or captured, especially if you are using your own company’s data. Treble so if you are using data you snaffled off the net, riddled with collection bias and untold omissions. (via Stijn Debrouwere)
  4. Deep Belief Networks in Javascript — just object recognition in the browser. The code relies on GPU shaders to perform calculations on over 60 million neural connections in real time. From the ever-more-awesome Pete Warden.
Comment

Crowdsourcing Feature discovery

More than algorithms, companies gain access to models that incorporate ideas generated by teams of data scientists

Data scientists were among the earliest and most enthusiastic users of crowdsourcing services. Lukas Biewald noted in a recent talk that one of the reasons he started CrowdFlower was that as a data scientist he got frustrated with having to create training sets for many of the problems he faced. More recently, companies have been experimenting with active learning (humans1 take care of uncertain cases, models handle the routine ones). Along those lines, Adam Marcus described in detail how Locu uses Crowdsourcing services to perform structured extraction (converting semi/unstructured data into structured data).

Another area where crowdsourcing is popping up is feature engineering and feature discovery. Experienced data scientists will attest that generating features is as (if not more) important than choice of algorithm. Startup CrowdAnalytix uses public/open data sets to help companies enhance their analytic models. The company has access to several thousand data scientists spread across 50 countries and counts a major social network among its customers. Its current focus is on providing “enterprise risk quantification services to Fortune 1000 companies”.

CrowdAnalytix breaks up projects in two phases: feature engineering and modeling. During the feature engineering phase, data scientists are presented with a problem (independent variable(s)) and are asked to propose features (predictors) and brief explanations for why they might prove useful. A panel of judges evaluate2 features based on the accompanying evidence and explanations. Typically 100+ teams enter this phase of the project, and 30+ teams propose reasonable features.

Read more…

Comments: 4
Four short links: 14 March 2014

Four short links: 14 March 2014

Facebook Criticism, New Games, Face Recognition, and Public Uber

  1. The Facebook experiment has failed. Let’s go backFacebook gets worse the more you use it. The innovation within Facebook happens within a framework that’s taken as given. This essay questions that frame, well.
  2. Meet the People Making New Games for Old Hardware“We’re all fighting for the same goal,” Cobb says. “There’s something artistic, and disciplined, about creating games for machines with limited hardware. You can’t pass off bloat as content, and you can’t drop in a licensed album in place of a hand-crafted digital soundtrack. To make something great you have to work hard, and straight from the heart. That’s what a lot of gamers still wish to see. And we’re happy to provide it for them.”
  3. DeepFace: Closing the Gap to Human-Level Performance in Face Verification — Facebook research into using deep neural networks for face recognition. Our method reaches an accuracy of 97.25% on the Labeled Faces in the Wild (LFW) dataset, reducing the error of the current state of the art by more than 25%, closely approaching human-level performance. “The best minds of my generation are thinking about how to make people click ads.” —Jeff Hammerbacher.
  4. Helsinki Does Uber for BusesHelsinki’s Kutsuplus lets you select your pick-up and drop-off locations and times, using a phone app, and then sends out a bus to take you exactly where you need to go.
Comment

An Invitation to Practical Machine Learning

PracticalMachineLearning_covDoes it make sense for me to have a car? If so, which one is the best choice for my needs: a gasoline, hybrid, or electric?  And should I buy or lease?

In order to make an effective decision, I need to understand key issues about the design, performance, and cost of cars, regardless of whether or not I actually know how to build one myself. The same is true for people deciding if machine learning is a good choice for their business goals or project.  Will the payoff be worth the effort?  What machine learning approach is most likely to produce valuable results for your particular situation? What size team with what expertise is necessary to be able to develop, deploy, and maintain your machine learning system?

Given the complex and previously esoteric nature of machine learning as a field – the sometimes daunting array of learning algorithms and the math needed to understand and employ them – many people feel the topic is one best left only to the few.

Read more…

Comments: 2

Bridging the gap between research and implementation

Hardcore Data Science speakers provided many practical suggestions and tips

One of the most popular offerings at Strata Santa Clara was Hardcore Data Science day. Over the next few weeks we hope to profile some of the speakers who presented, and make the video of the talks available as a bundle. In the meantime here are some notes and highlights from a day packed with great talks.

Data Structures
We’ve come to think of analytics as being comprised primarily of data and algorithms. Once data has been collected, “wrangled”, and stored, algorithms are unleashed to unlock its value. Longtime machine-learning researcher Alice Zheng of GraphLab, reminded attendees that data structures are critical to scaling machine-learning algorithms. Unfortunately there is a disconnect between machine-learning research and implementation (so much so, that some recent advances in large-scale ML are “rediscoveries” of known data structures):

Data and Algorithms: The Disconnect

While there are many data structures that arise in computer science, Alice devoted her talk to two data structures1 that are widely used in machine-learning:

Read more…

Comments: 2
Four short links: 6 February 2014

Four short links: 6 February 2014

Emotions Wanted, Future's So Bright, Machine Learning for Security, and Medieval Unicode Fonts

  1. What Machines Can’t Do (NY Times) — In the 1950s, the bureaucracy was the computer. People were organized into technocratic systems in order to perform routinized information processing. But now the computer is the computer. The role of the human is not to be dispassionate, depersonalized or neutral. It is precisely the emotive traits that are rewarded: the voracious lust for understanding, the enthusiasm for work, the ability to grasp the gist, the empathetic sensitivity to what will attract attention and linger in the mind. Cf the fantastic The Most Human Human. (via Jim Stogdill)
  2. The Technium: A Conversation with Kevin Kelly (Edge) — If we were sent back with a time machine, even 20 years, and reported to people what we have right now and describe what we were going to get in this device in our pocket—we’d have this free encyclopedia, and we’d have street maps to most of the cities of the world, and we’d have box scores in real time and stock quotes and weather reports, PDFs for every manual in the world—we’d make this very, very, very long list of things that we would say we would have and we get on this device in our pocket, and then we would tell them that most of this content was free. You would simply be declared insane. They would say there is no economic model to make this. What is the economics of this? It doesn’t make any sense, and it seems far-fetched and nearly impossible. But the next twenty years are going to make this last twenty years just pale. (via Sara Winge)
  3. Applying Machine Learning to Network Security Monitoring (Slideshare) — interesting deck on big data + machine learning as applied to netsec. See also their ML Sec Project. (via Anton Chuvakin)
  4. Medieval Unicode Font Initiative — code points for medieval markup. I would have put money on Ogonek being a fantasy warrior race. Go figure.
Comment: 1

Business analysts want access to advanced analytics

Business users are starting to tackle problems that require machine-learning and statistics

I talk with many new companies who build tools for business analysts and other non-technical users. These new tools streamline and simplify important data tasks including interactive analysis (e.g., pivot tables and cohort analysis), interactive visual analysis (as popularized by Tableau and Qlikview), and more recently data preparation. Some of the newer tools scale to large data sets, while others explicitly target small to medium-sized data.

As I noted in a recent post, companies are beginning to build data analysis tools1 that target non-experts. Companies are betting that as business users start interacting with data, they will want to tackle some problems that require advanced analytics. With business analysts far outnumbering data scientists, it makes sense to offload some problems to non-experts2.

Moreover data seems to support the notion that business users are interested in more complex problems. I recently looked at data3 from 11 large Meetups (in NYC and the SF Bay Area) that target business analysts and business intelligence users. Altogether these Meetups had close to 5,000 active4 members. As you can see in the chart below, business users are interested in topics like machine learning (1 in 5), predictive analytics (1 in 4), and data mining (1 in 4):

Key topics of interest: Active members of SF & NYC meetups for business analysts

Read more…

Comments: 2
Four short links: 28 January 2014

Four short links: 28 January 2014

Client-Server, Total Information Awareness, MSFT Joins OCP, and Tissue Modelling

  1. Intel On-Device Voice Recognition (Quartz) — interesting because the tension between client-side and server-side functionality is still alive and well. Features migrate from core to edge and back again as cycles, data, algorithms, and responsiveness expectations change.
  2. Meet Microsoft’s Personal Assistant (Bloomberg) — total information awareness assistant. By Seeing, Hearing, and Knowing All, in the future even elevators will be trying to read our minds. (via The Next Web)
  3. Microsoft Contributes Cloud Server Designs to Open Compute ProjectAs part of this effort, Microsoft Open Technologies Inc. is open sourcing the software code we created for the management of hardware operations, such as server diagnostics, power supply and fan control. We would like to help build an open source software community within OCP as well. (via Data Center Knowledge)
  4. Open Tissue Wiki — open source (ZLib license) generic algorithms and data structures for rapid development of interactive modeling and simulation.
Comment: 1
Four short links: 22 January 2014

Four short links: 22 January 2014

Mating Math, Precogs Are Coming, Tor Bad Guys, and Mind Maps

  1. How a Math Genius Hacked OkCupid to Find True Love (Wired) — if he doesn’t end up working for OK Cupid, productising this as a new service, something is wrong with the world.
  2. Humin: The App That Uses Context to Enable Better Human Connections (WaPo) — Humin is part of a growing trend of apps and services attempting to use context and anticipation to better serve users. The precogs are coming. I knew it.
  3. Spoiled Onions — analysis identifying bad actors in the Tor network, Since September 2013, we discovered several malicious or misconfigured exit relays[…]. These exit relays engaged in various attacks such as SSH and HTTPS MitM, HTML injection, and SSL stripping. We also found exit relays which were unintentionally interfering with network traffic because they were subject to DNS censorship.
  4. My Mind (Github) — a web application for creating and managing Mind maps. It is free to use and you can fork its source code. It is distributed under the terms of the MIT license.
Comment

The democratization of medical science

An interview with Ash Damle of Lumiata on the role of data in healthcare.

Vinod Khosla has stirred up some controversy in the healthcare community over the last several years by suggesting that computers might be able to provide better care than doctors. This includes remarks he made at Strata Rx in 2012, including that, “We need to move from the practice of medicine to the science of medicine. And the science of medicine is way too complex for human beings to do.”

So when I saw the news that Khosla Ventures has just invested $4M in Series A funding into Lumiata (formerly MEDgle), a company that specializes in healthcare data analytics, I was very curious to hear more about that company’s vision. Ash Damle is the CEO at Lumiata. We recently spoke by phone to discuss how data can improve access to care and help level the playing field of care quality.

Tell me a little about Lumiata: what it is and what it does.

Lumiata network graph of diagnosis interrelation.

A Lumiata network graph of diagnosis interrelation.

Ash Damle: We’re bringing together the best of medical science and graph analytics to provide the best prescriptive analysis to those providing care. We data-mine all the publicly available data sources, such as journals, de-identified records, etc. We analyze the data to make sure we’re learning the right things and, most importantly, what the relationships are among the data. We have fundamentally delved into looking at that whole graph, the way Google does to provide you with relevant search results. We curate those relationships to make sure they’re sensible, and take into account behavioral and social factors.

Read more…

Comment