Getting More Information Out of Information

Utah State University professor Adele Cutler’s passion for statistics has taken her to the boundaries where the discipline converges with computer science and electrical engineering. Her work has led her to research projects using such tools as bioinformatics, archetypal analysis, and machine learning. She’s applied these methods to diverse fields ranging from genetics, medicine, and astronomy, to banking, air traffic control and national security.

 

“An advantage of statistics is that you can participate in exciting research in a lot of different disciplines without restricting yourself,” says Cutler, a faculty member in the College of Science’s Department of Mathematics and Statistics. “As statisticians, what we’re really trying to do is think of better ways to get information out of data.”

 

Of particular significance to Cutler is her ongoing work with Random Forests™, a trademarked statistical classifier developed by the late Leo Breiman, her mentor and longtime colleague.

 

Breiman, professor emeritus of statistics at the University of California-Berkley, died July 7, 2005 at the age of 77. Renowned for his work with statistical computation, Breiman was elected a member of the National Academy of Sciences and of the American Academy of Arts and Sciences.

 

“Random Forests was really a work of a lifetime,” says Cutler, who collaborated with Breiman for more than 20 years. “It’s a powerful, versatile tool that outperforms traditional statistical tools.”

 

Many data sets encountered in today’s scientific fields are much bigger and complex “than anything we’ve dealt with before,” she says. “Random Forests allows us to interpret data and gain insights in ways other tools can’t. We can explore, for example, why a ‘yes’ is a ‘yes.”

 

Each of us encounters applications using Random Forests, says Cutler, though we may not even realize it. Did you look up anything on Amazon.com or another online retailer today? You may not have noticed, but the site automatically logged your interests and, like an attentive salesperson, offered up a slew of suggestions for you.

 

Or perhaps you had a non-virtual shopping experience and handed your keys, with a colorful, dangling array of bar-coded, frequent shopper cards, over to a human checkout clerk. “Retailers collect an amazing amount of information about our preferences,” says Cutler.

 

In the life sciences, where recent developments in genomics have created floods of information, she says, Random Forests provides researchers with the ability to distill critical insights from huge data sets.

 

Cutler’s fascination with statistics never wanes. “If I get a day when I can do anything I want, I’ll sit at the computer and work on Random Forests,” she says.

 

Her goal is to continue Breiman’s work and complete a book on the subject. He even chose the cover art for the book – a work by Cutler’s young son, Phil.

 

For a presentation at a conference, Cutler selected a photo of a forest showing bare branches shrouded in fog. “I thought it was really pretty, but Leo (Breiman) said, ‘Too gloomy.’

 

So Cutler commissioned her son Phil, then seven years of age, to come up with a drawing. “Leo loved Phil’s crayon drawing. He said, ‘It’s bright, cheerful and, most importantly, shows the simplicity of the method.’”

 

In a world obscured by mystery, complexity and reams of data, says Cutler, statistics provides a light at the end of the tunnel. “Statistical tools give scientists that moment of clarity, where it all becomes clear,” she says.