Does the Brain Learn in the Same Manner that Machines Learn?

Identifying how neural activity changes with learning is anything but black and white. Recently, some have presumed that learning in the brain, or biological learning, may be visualized in terms of optimization, which is how learning happens in artificial networks like computers or robots.

A new approaches piece co-authored by Carnegie Mellon University and University of Pittsburgh researchers connects machine learning to biological learning, revealing that the two approaches are not compatible yet can be harnessed to provide valuable insights into how the brain runs.

Brain vs Machine

“How we measure the modifications we see in the brain, and a subject’s conduct during learning is ever-evolving,” claims Byron Yu, professor of biomedical engineering and electrical and computer engineering.

It turns out that in artificial intelligence and machine learning, there is a well-developed framework in which an element learns, named optimization.

We and others in the field have been thinking about how the brain learns in contrast to this framework, which was created to teach artificial agents to learn.

The optimization point of view suggests that activity in the brain ought to modify during learning in a mathematically prescribed way, similar to how the task of artificial neurons changes in a specific way when they are educated to drive a robot or play chess.

“One thing we have an interest in understanding is how the learning process unfolds with time, not merely considering a photo of before and after learning happens,” explains Jay Hennig, a recent Ph.D. graduate in neural computation and machine learning at Carnegie Mellon.

“In this approaches piece, we provide three main takeaways that would be essential for people to consider in the context of thinking about why neural activity could alter throughout learning that can not be readily described in regards to optimization.”

Unconventional ways of learning

The takeaways include the inflexibility of neural variability throughout learning, the usage of numerous learning processes even during basic tasks, and the existence of huge task-nonspecific activity modifications.

“It is attractive to draw from successful examples of artificial learning agents and suppose the brain has to do whatever they do,” suggests Aaron Batista, professor of bioengineering at the University of Pittsburgh. Yet, one specific distinction between artificial and biological learning systems is the artificial system normally does just a single thing and does it absolutely well.

Activity in the brain is quite distinct, with lots of processes occurring at the same time. We and others have observed that things occur in the brain that machine learning models can not yet explain.”

Steve Chase, professor of biomedical engineering at Carnegie Mellon and the Neuroscience Institute, says, “We see a theme constructing and a way for the future. By underlining these areas where neuroscience can inform machine learning and vice versa, we intend to link them to the optimization sight to eventually comprehend, on a deeper level, how learning develops in the brain.”

This job is co-authored with Emily Oby, research study faculty in bioengineering at the University of Pittsburgh, and Darby Losey, Ph.D. student in neural computation and machine learning at CMU.

The team’s work is ongoing and done in cooperation with the Center for Neural Basis of Cognition, a cross-university research study and educational program between Carnegie Mellon and the University of Pittsburgh that takes advantage of each institution’s strengths to explore the cognitive and neural mechanisms that trigger biological intelligence and habits.

Read the original article on Medical Xpress.

Related “Scientists Target Protein to Lower the Risk of Prostate Cancer Spread”