SVSU students to test several artificial intelligence methods for creativity program

January 28, 2008 —

by Christina Dillbeck

Students Joshua Urbain (left) and Joshua Cieszlak (right) are working together to research artificial intelligence methods. They plan to continue their research until the end of the 2008 year.

A pair of SVSU students is looking for ways to make certain artificial intelligence methods faster while making research methods less expensive.

Joshua Urbain, 23, and Joshua Cieszlak, 21, are testing several AI methods as part of the Student Research Creativity Institute program.

They hope to find ways to separate the methods among many computers instead of one supercomputer.

Urbain says avoiding the need to rely on one supercomputer for these methods would yield faster speeds.

This would eliminate the need for a costly supercomputer, he said, and allow a group of regular computers to perform the same task.

"By doing this, we're giving the group of computers smaller tasks instead of relying on one supercomputer to perform the entire job," he said.

The two students started their research at the beginning of the month and plan to continue it through December.

They plan to study several expert systems to see if separating them among several computers is possible.

Expert systems are forms of artificial intelligence that contain the knowledge and analytical skills of one or more experts in a given field.

The most common expert system is a computer program that troubleshoots a set of problems.

These programs usually require the user to input the general problem the user is experiencing. The program then analyzes the information and offers a set of possible solutions.

Urbain and Cieszlak chose a course advisor expert system that Urbain created as a jumping off point for their research.

The expert system requires users to input the courses they have already taken, according to Urbain, and the system tells them the next class they need to take.

Urbain said the program as it was would not benefit from multiple computers.

"For the expert systems to benefit from multiple computers, the results need to keep building off each other," he said. "If this system did that, it would not only say you need to take Class B because you've taken Class A. It would also say that you need to take Class C after you take Class B."

Urbain said he and Cieszlak believe they can find a way for the system to benefit from multiple computers.

The duo also will test other artificial intelligence methods such as neural networks, fuzzy systems, path-findings, and data mining.

A neural network, according to Urbain, is a method that simulates the brain by having neurons firing off to teach it.

Fuzzy systems add a gradient between settings. Urbain used a water faucet to exemplify how the system works.

"Say a faucet has a hot and cold setting," he said. "A fuzzy system will use a gradient to add another setting, like lukewarm water."

Path-finding systems are utilized for Web sites such as Mapquest.

Data mining utilizes user searches to suggest other pages to the user.

Amazon.com, the site Urbain and Cieszlak will use for the data mining portion of their research, uses the system to make product suggestions based on the user's searches and purchases.