6月 102016
 

Learn more about machine learning"Shall we play a game?"

If you’re a child of the ’80s like me, you might recognize this famous line from the movie WarGames. This innocent-sounding question comes not from one of the movie’s human stars, but from a military super-computer named Joshua, after a bored high school student, played by Matthew Broderick, accesses the computer’s hard drive.

Thinking he’s hacked into a video game company, Broderick’s character accepts Joshua’s challenge and chooses the most intriguing game he can find: global thermonuclear war. To Joshua, though, it’s not just a game. Joshua is an intelligent computer programmed to learn through simulations like the one Broderick’s character initiates. And because the computer actually does control the arsenal of U.S. nuclear weapons, it’s a “game” that puts the planet on the brink of World War III.

If you want to see how the movie ends, I’d encourage you to check it out.

The reason I mention the movie is because, in addition to scaring the heck out of me, WarGames was my first exposure to machine learning, the idea that computers can learn and adapt based on the data they collect. Of course, machine learning has changed a lot since my 1983 Hollywood introduction.

Today, progressive organizations in all industries and disciplines are using the technology – which draws from various fields of study like artificial intelligence, data mining, statistics and optimization – to transform data into business value.

Machine learning defined

The concept is pretty straightforward. Computers “learn” from patterns in data to build a model, that when exposed to new data, can independently adapt without human intervention. The process repeats itself over and over again, with each iteration building on previous information and computations to create more accurate results or better predictions of future outcomes.

Machine learning is revolutionary for many analytically-mature organizations. Big data means even bigger possibilities. Consider what organizations like Amazon and Netflix are doing with recommendation systems based largely on machine learning principles and millions of data points.

Facebook is another great example. I’m sure you’ve noticed that when you upload a photo with people in it, Facebook will suggest that you tag certain individuals it recognizes as Facebook users, even pre-populating names to make it easier for you. That’s all done by a computer using facial recognition and an algorithm that analyzes an individual’s facial features to match faces with names.

My favorite machine learning resources

I won’t pretend to be an expert, but fascinating applications of machine learning like these push me to learn more about the science. It’s why I’m writing this post. If you’re interested in the concepts around machine learning, SAS has some excellent resources that have helped guide me in my discovery process. I believe you would find them helpful as well.

  1. For starters, SAS has an awesome web page on machine learning that includes a discussion on what it is, why it matters, who’s using it and how it works. You’ll also find a 60-second overview video and another video demonstrating how SAS Enterprise Miner uses pattern recognition to accurately recognize pictures and deliver that information to a user without human intervention. The best thing about the site, though, is that it serves as an aggregator for a number of other articles and resources to help you learn more. It probably is the best place to begin.
  2. Another excellent resource is a new O’Reilly white paper, The Evolution of Analytics: Opportunities and Challenges for Machine Learning in Business. SAS Data Scientist Patrick Hall, Solution Architect Wen Phan and Senior Marketing Specialist Katie Whitson give a really good overview of what machine learning is and what it is not. They also outline opportunities for businesses to effectively use machine learning and how to overcome some of the challenges they might encounter trying to weave it into their existing analytics strategy. This includes excellent discussions around talent scarcity, data and infrastructure challenges and the change management required to execute an analytics strategy that includes machine learning.
    Perhaps the best part of the report is the section highlighting some of the modern applications for machine learning – recommendation systems, streaming analytics, deep learning and cognitive computing. The authors highlight two companies – Geneia in the healthcare industry and Equifax in financial services – that are effectively using machine learning, walking readers through the journey these companies have taken to get to where they are today. Really interesting stuff.
  3. To discuss machine learning with your peers I recommend the SAS Data Mining Community. Conversations around machine learning are happening in this space all the time. With nearly 100,000 members from across all fields and disciplines, and thousands logged in at any given time, SAS Communities are a great place to ask and answer questions, or just follow along with the conversation. One discussion thread you should definitely check out is titled Recently Published Machine Learning Resources. Hall, one of the author mentioned above, highlights a couple of useful articles, the white paper and other resources. He also references a GitHub repo that includes a ton of good resources, including several machine learning quick reference tables outlining best practices.

So there you have it, a few good places to start if you’re looking to build your knowledge of machine learning. The possibilities machine learning presents to the world are very exciting.

Just please to be careful if that system you build asks you to play a game.

 

tags: machine learning, SAS

3 good resources for humans who want to learn more about machine learning was published on SAS Users.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

(required)

(required)