New paper from the Tavoni Lab offers insights into the memory capacity of brain-inspired networks.
Gaia Tavoni, PhD, an assistant professor of neuroscience, and Kaining Zhang, PhD, a postdoctoral research associate in the Tavoni Lab, have uncovered a key principle that helps brainlike networks store more memories and make them more resilient to damage. Their findings, published June 20 in PRX Life, shed light on how the brain’s natural diversity may support a powerful and flexible memory system.
Traditional models of memory, such as canonical Hopfield networks, assume all neurons behave equally. But real brains are more complex — some neurons are more active, and some have more connections than others. The study extends classical theory to account for such biological variability. The researchers found that although diversity in neuron activity and connectivity typically reduces memory capacity, this loss can be avoided if these two types of variability are properly matched. As a result, networks that mirror the brain’s diversity can perform as well as idealized models while being more realistic. This principle applies to both independent memories and memories organized around shared themes.
The study also examines how these findings apply to the hippocampus, a brain region crucial for memory. Tavoni and Zang discovered that a wiring strategy called “quasi-indexing” can boost memory capacity in the hippocampus and protect stored information even when some neurons are lost, offering new insights into how the brain encodes and safeguards memories.