Traditionally, when we think of algorithms learning how to learn, the flow of insight is very much from humans to the machine. Research from the University of California, Irvine School of Biological Sciences suggests that the reverse could also be true, with insights into how algorithms learn helping us to understand how the brain absorbs new information.

The researchers examined artificial neural networks, which are designed to emulate the way neurons in the brain function by absorbing and classifying huge quantities of information. Unlike the human brain, however, these neural networks usually forget what they know, especially when information is thrown at them very quickly.

How to learn

It has long been believed that we learn new things via an interplay between the neocortex and hippocampus parts of the brain. The latter captures new information, whereas the former meshes the new information with existing knowledge.

Recently this assumption has been called into question, not least due to the amount of time it would take the brain to successfully sort the vast quantity of information we gather throughout our lifetime. The researchers believe this could help to explain why neural networks struggle to retain long-term knowledge.

A common approach in deep machine learning involves retraining the network on an entire dataset from the past, regardless of whether this data is related to any new information the neural network might encounter. Suffice to say, this is extremely time-consuming, but the researchers found a better way.

“We found that when ANNs interleaved a much smaller subset of old information, including mainly items that were similar to the new knowledge they were acquiring, they learned it without forgetting what they already knew,” the researchers explain.

“It allowed ANNs to take in fresh information very efficiently, without having to review everything they had previously acquired,” they continue. “These findings suggest a brain mechanism for why experts at something can learn new things in that area much faster than non-experts. If the brain already has a cognitive framework related to the new information, the new material can be absorbed more quickly because changes are only needed in the part of brain’s network that encodes the expert knowledge.”

The researchers believe that their insights could be crucial to tackling various cognitive issues. For instance, understanding the various mechanisms that underpin learning is crucial to any progress we might make.

“It gives us insights into what’s going on when brains don’t work the way they are supposed to,” they continue. “We could develop training strategies for people with memory problems from aging or those with brain damage. It could also lead to the ability to manipulate brain circuits so people can overcome these deficits.”

The findings offer possibilities as well for making algorithms in machines such as medical diagnostic equipment, autonomous cars and many others more precise and efficient.