Machine Learning Achieves Humanitarian Ends

Machine Learning Achieves Humanitarian Ends

  • Comments 1
  • Likes

Posted by Rob Knies

Infer.NET logo

In February 2012, in response to an initiative from the administration of U.S. President Obama to harness technology and innovation to encourage development for longtime scientific challenges such as health, food security, and climate change, the U.S. Patent and Trademark Office launched Patents for Humanity, a program to recognize those who use patented technology to aid the less fortunate.

The inaugural winners are in, and prominent among them is Infer.NET, a Microsoft Research Cambridge library for machine learning, which won one of the contest’s four categories, information technology.

The awards were presented April 11 in the Dirksen Senate Office Building on Capitol Hill in Washington, D.C. Teresa Stanek Rea, acting undersecretary of Commerce for Intellectual Property and acting director of the U.S. Patent and Trademark Office, made the presentations, and U.S. Sen. Patrick Leahy, chairman of the Senate Judiciary Committee, who introduced last year’s Patents for Humanity Program Improvement Act, spoke during the event. Fred Humphries, Microsoft vice president for U.S. Government Affairs, accepted the award.

John Winn, along with Tom Minka one of the inventors of Infer.NET, didn’t exactly see such an honor coming.

Infer.NET teamWinn’s reaction upon hearing of the award: “Quite surprised!”

“Our goal in creating Infer.NET was to democratize machine learning—to make a tool that brings advanced machine-learning techniques to a broader audience,” Winn says. “As part of doing this, we wanted to test out Infer.NET in a wide variety of areas.

“I have been particularly interested in medical and biological applications, but others have used Infer.NET to solve a dizzying range of problems, including modeling climate change.”

Winn and Minka, both senior researchers in the Machine Learning and Perception group at Microsoft Research Cambridge, devised Infer.NET, which is available for download, as a framework for running Bayesian inference in graphical models to accomplish machine learning in a simplified manner. It uses state-of-the-art algorithms to draw probabilistic inferences from data, and the result is a simplified, flexible tool that can scale for large amounts of data and various types of data sets and problems. It can be used to solve many machine-learning problems, from classification or clustering to customized solutions to domain-specific problems.

“Infer.NET brings the power of advanced machine-learning methods to many more people,” Winn says. “Rather than write machine-learning code, an Infer.NET user writes a description of the process behind their data—a model of the data. Infer.NET then automatically generates a customized machine-learning algorithm for analyzing that data.”

The Patents for Humanity program, conducted with support from the Ewing Marion Kauffman Foundation, is a voluntary competition for patent owners and businesses. It encourages businesses to apply their patented technology to address humanitarian challenges. The effort advances the president’s global development agenda by rewarding companies that bring life-saving technologies to underserved people of the world while showing how patents are an integral part of tackling the world’s challenges.

Entrants compete in four categories: medical technology, food and nutrition, clean technology, and IT.

Microsoft’s Legal and Corporate Affairs team submitted the Infer.NET entry by focusing on three specific ways in which the tool has helped humanity:

  • Asthma epidemiology: Microsoft and the University of Manchester are using Infer.NET to model genetic and environmental factors jointly, in contrast to prior research that approached them separately. The immediate goal is to study the development of asthma, but a successful outcome could highlight the benefits of a model-based approach to the analysis of clinical data, which could lead to much broader applicability.
  • Gene analysis: The Wellcome Trust Sanger Institute is analyzing key parts of the DNA sequence of an individual. Infer.NET has helped to identify correlations between our genetics and the activity of genes in different tissues and the symptoms or characteristics of the individuals.
  • Global warming and deforestation: Forest-dynamics research at Microsoft Research Cambridge uses Infer.NET to examine how key drivers of such dynamics, such as the growth and mortality rates of different-sized trees in different kinds of forests, vary across geographic space and from year to year.

Rea acknowledged in an award-notification email that Infer.NET richly warranted the recognition.

“Congratulations on your impressive accomplishment!” she wrote. “Your efforts to improve the lives of the less fortunate with game-changing technology are vital to overcoming the global challenges facing humanity.”

Work on Infer.NET continues, award or no award.

“Infer.NET aims to encapsulate some of the latest machine-learning techniques,” Winn says. “In the course of developing Infer.NET, we’ve also had to develop new ways to automate parts of the machine-learning process that were previously done by hand. But there are still a lot of unsolved problems in machine learning, and researchers are constantly making progress on these.

“Our aim is to bring the most valuable new techniques into Infer.NET—along with some of our own—so that it can be used to solve even harder problems.”

For a moment, though, such challenges take a back seat. Life’s sweet moments are meant to be savored.

“The fact that Infer.NET can be applied to such important problems,” Winn smiles, “is a source of great satisfaction and very motivating!”

Your comment has been posted.   Close
Thank you, your comment requires moderation so it may take a while to appear.   Close
Leave a Comment
  • Yes absolutely Correct,Machines is anything which reduces the human effort.