Neuron-Astrocyte Associative Memory

14 Nov 2023  ·  Leo Kozachkov, Jean-Jacques Slotine, Dmitry Krotov ·

Astrocytes, a unique type of glial cell, are thought to play a significant role in memory due to their involvement in modulating synaptic plasticity. Nonetheless, no existing theories explain how neurons, synapses, and astrocytes could collectively contribute to memory function. To address this, we propose a biophysical model of neuron-astrocyte interactions that unifies various viewpoints on astrocyte function in a principled, biologically-grounded framework. A key aspect of the model is that astrocytes mediate long-range interactions between distant tripartite synapses. This effectively creates ``multi-neuron synapses" where more than two neurons interact at the same synapse. Such multi-neuron synapses are ubiquitous in models of Dense Associative Memory (also known as Modern Hopfield Networks) and are known to lead to superlinear memory storage capacity, which is a desirable computational feature. We establish a theoretical relationship between neuron-astrocyte networks and Dense Associative Memories and demonstrate that neuron-astrocyte networks have a larger memory storage capacity per compute unit compared to previously published biological implementations of Dense Associative Memories. This theoretical correspondence suggests the exciting hypothesis that memories could be stored, at least partially, within astrocytes instead of in the synaptic weights between neurons. Importantly, the many-neuron synapses can be influenced by feedforward signals into the astrocytes, such as neuromodulators, potentially originating from distant neurons.

PDF Abstract
No code implementations yet. Submit your code now

Tasks


Datasets


Results from the Paper


  Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

Methods


No methods listed for this paper. Add relevant methods here