Researchers Train Computer to Predict Human Brain

One of the biggest challenges in developing computers that can think and see in the same way as humans is to understand how human brain categorizes information through signs and language.  The researchers from the Tokyo Institute of Technology train a computer using fMRI datasets to train a computer that can predict the semantic category of an image originally viewed by five different people.

This was led by Hiroyuki Akama at the Graduate School of Decision Science and Technology, Tokyo Institute of Technology, together with co-workers in Yokohama, the USA, Italy and the UK.

In order to proceed with the study, the researchers were asked to view the pictures of animals and hand tools along with an auditory or written description.

The participants were instructed to label each pictured object with certain properties, whilst undergoing an fMRI brain scan. According to the ResearchSEA, akama and his team had nearly 240 individuals scan showing brain activity session. They analyzed the brain scans using technique known as multi voxel pattern analysis which involves using computer algorithm to identify repeating patterns of brain activity across voxels.

Based on the MVPA result the researchers tried to find out if the computer could identify whether or not the participants were viewing an animal or hand tool image based on the pattern in the sans.

The resulting scans were analyzed using algorithms that identified patterns relating to the two separate semantic groups. By using some of the auditory session data in training the algorithms, the computer identified the left out scans 80 to 90 of the time. Similar results were obtained with the orthographic session data.

A cross-modal approach was used, namely training the computer using auditory data, but testing it using orthographic. This reduced performance to 65-75 percent. Continued research in this area could lead to systems that allow people to speak through a computer simply by thinking about what they want to say.