Deep learning, artificial intelligence leading the way to smart houses

Dr. Liang Dong is leading deep learning research. Here, he shows off a digital communication board. Baylee VerSteeg | Multimedia Journalist

By Samantha Amaro | Reporter

Plans for smart houses in the future are slowly becoming more and more plausible. A house that does all the manual labor for the occupants, where dinner is ready on the kitchen table and all the amenities in a house are included in these plans. Thanks to Baylor University’s School of Electrical and Computer Engineering and deep learning research, a future with smart houses is getting closer.

Listed among the research opportunities in the School of Electrical and Computer Engineering is deep learning. This research is helping Artificial Intelligence (AI) to develop into what is presented in science fiction novels and television shows.

Dr. Liang Dong, an associate professor of electrical and computer engineering, is leading this research. Dong has been with deep learning for about three years, but only brought this research into Baylor two years ago. The research is being funded not only by Baylor, but by Intel and a prospective new funder, the United States’ Department of Defense (DOD).

Intel is interested in the AI research going on in Dong’s deep learning research, while the DOD is interested in applying Deep Learning in combat.

“The computer teaches itself,” Dong said. “Deep learning is more to mimic a human brain.”

Through the use of algorithms and data, computers are able to compare results against many other previous studies. So far, the deep learning project is being tailored for the specific use of analyzing medical images like from positron emission tomography (PET) scans and computed tomography (CT) scans in hospitals. This would help to more accurately catch the development of cancer and other diseases. The research — conducted at the Baylor Research and Innovation Collaborative (BRIC) — is essentially split up into two categories.

The theoretical research is composed of distributed deep learning and energy-efficient deep learning. Distributed deep learning deals with investigating how to use several local machines to compute different parts of the main neural network. It solves the problem of the large amount of time it takes to train a deep neural network in a single machine. Energy-efficient deep learning focuses on the problem of being able to provide a constant source of energy for necessary continuous projects.

There are many possible and practical applications for deep learning, such as in building smart hospitals or for the production of smart homes. The idea for utilizing deep learning in hospitals is to take an x-ray using less radiation and power, then comparing the results to other x-rays that have been taken in the past. The computer would compare a single x-ray to hundreds, if not thousands, of previous results and point out which areas seem uncommon for a healthy human system, or which areas look like they’re developing a kind of disease by comparing a person’s x-ray to those of others who already have the disease.

Dong is already working with a hospital in Dallas, the University of Texas Southwestern Medical Center, to test if this research is efficient and plausible enough for doctors to implement in other hospitals.

There is also the smart home project — to see if a house can measure a person’s overall health. By applying sensors throughout the house, or even by sticking to a certain area of the house like the kitchen, sensors would read a person’s biorhythm and send alerts if something looks wrong. It will alert the occupants of the house that there is a sign of lower heart rate or of a drop in heart rate — things that can flag some serious health issues in the near future.

Several Ph.D and masters students are involved with this research, helping Dong in the labs at the BRIC. Among these is Hailing District, China, doctoral student Yuchen Qian. He is working on his own project as well as being involved with Dong’s Deep Learning research. The research process, he explained, is rather simple.

“When we have ideas we talk about them to him in his office. Then we program the codes and we run the coding,” Qian said. “Since we can to the experiments, we can also do the experimenting.”

All in all, there are still many things expected to generate from Dong’s research in deep learning.