Computer model helps make sense of human memory

Summary: A newly developed artificial neural network allowed researchers to insert inhibitory circuits into the model, altering the excitatory circuits and enabling them to extend associative memory.

Source: Okinawa Institute of Science and Technology

Brains are a mazy network of overlapping circuits — some pathways encourage activity while others suppress it. While earlier studies focused more on excitatory circuits, inhibitory circuits are now understood to play an equally important role in brain function. Researchers at the Okinawa Institute of Science and Technology Graduate University (OIST) and the RIKEN Center for Brain Science have created an artificial network to simulate the brain, demonstrating that tinkering with inhibitory circuits leads to extended memory.

Associative memory is the ability to connect unrelated items and store them in memory — to associate co-occurring items as a single episode. In this study, published in Physical Review Letters, the team used sequentially arranged patterns to simulate a memory, and found that a computer is able to remember patterns spanning a longer episode when the model takes inhibitory circuits into account. They go on to explain how this finding could be applied to explain our own brains.

“This simple model of processing shows us how the brain handles the pieces of information given in a serial order,” explains Professor Tomoki Fukai, head of OIST’s Neural Coding and Brain Computing Unit, who led the study with RIKEN collaborator Dr. Tatsuya Haga. “By modeling neurons using computers, we can begin to understand memory processing in our own minds.”

Lower Your Inhibitions

Thinking about the brain in terms of physical, non-biological phenomena is now a widely accepted approach in neuroscience — and many ideas lifted from physics have now been validated in animal studies. One such idea is understanding the brain’s memory system as an attractor network, a group of connected nodes that display patterns of activity and tend towards certain states. This idea of attractor networks formed the basis of this study.

A tenet of neurobiology is that “cells that fire together wire together” — neurons that are active at the same time become synchronized, which partly explains how our brains change over time. In their model, the team created excitatory circuits — patterns of neurons firing together — to replicate the brain. The model included many excitatory circuits spread across a network.

More importantly, the team inserted inhibitory circuits into the model. Different inhibitory circuits act locally on a particular circuit, or globally across the network. The circuits block unwanted signals from interfering with the excitatory circuits, which are then better able to fire and wire together. These inhibitory circuits allowed the excitatory circuits to remember a pattern representing a longer episode.

The finding matches what is currently known about the hippocampus, a brain region involved in associative memory. It is thought that a balance of excitatory and inhibitory activity is what allows new associations to form. Inhibitory activity could be regulated by a chemical called acetylcholine, which is known to play a role in memory within the hippocampus. This model is a digital representation of these processes.

A challenge to the approach, however, is the use of random sampling. The sheer number of possible outputs, or attractor states, in the network, overworks a computer’s memory capacity. The team instead had to rely on a selection of outputs, rather than a systematic review of every possible combination. This allowed them to overcome a technical difficulty without jeopardizing the model’s predictions.

This is a representation of the circuit
An excitatory circuit, μ, comprises a pattern of neurons that are firing (1) or not (0). Local and global inhibitory circuits acted on the excitatory circuit, allowing the circuit to remember a pattern for longer. This artificial network represents memory processes taking place in the hippocampus. The image is credited to OIST.

Overall, the study allowed for overarching inferences — inhibitory neurons have an important role in associative memory, and this maps to what we might expect in our own brains. Fukai says that biological studies will need to be completed to determine the exact validity of this computational work. Then, it will be possible to map the components of the simulation to their biological counterparts, building a more complete picture of the hippocampus and associative memory.

The team will next move beyond a simple model toward one with additional parameters that better represents the hippocampus, and look at the relative importance of local and global inhibitory circuits. The current model comprises neurons that are either off or on — zeros and ones. A future model will include dendrites, the branches that connect neurons in a complicated mesh. This more realistic simulation will be even better placed to make conclusions about biological brains.

About this neuroscience research article

Okinawa Institute of Science and Technology
Media Contacts:
Tomomi Okubo – Okinawa Institute of Science and Technology
Image Source:
The image is credited to OIST.

Original Research: Open access
“Extended Temporal Association Memory by Modulations of Inhibitory Circuits”. Tatsuya Haga and Tomoki Fukai.
Physical Review Letters doi:10.1103/PhysRevLett.123.078101.


Extended Temporal Association Memory by Modulations of Inhibitory Circuits1

Hebbian learning of excitatory synapses plays a central role in storing activity patterns in associative memory models. Interstimulus Hebbian learning associates multiple items by converting temporal correlation to spatial correlation between attractors. Growing evidence suggests the importance of inhibitory plasticity in memory processing, but the consequence of such regulation in associative memory has not been understood. Noting that Hebbian learning of inhibitory synapses yields an anti-Hebbian effect, we show that the combination of Hebbian and anti-Hebbian learning can significantly increase the span of temporal association between correlated attractors as well as the sensitivity of these states to external input. Furthermore, these effects are regulated by changing the ratio of local and global recurrent inhibition after learning weights for excitation-inhibition balance. Our results suggest a nontrivial role of plasticity and modulation of inhibitory circuits in associative memory.

Feel free to share this Artificial Intelligence News.
Join our Newsletter
I agree to have my personal information transferred to AWeber for Neuroscience Newsletter ( more information )
Sign up to receive our recent neuroscience headlines and summaries sent to your email once a day, totally free.
We hate spam and only use your email to contact you about newsletters. You can cancel your subscription any time.