A model of working memory for encoding multiple items and ordered sequences exploiting the theta-gamma code

Abstract

Recent experimental evidence suggests that oscillatory activity plays a pivotal role in the maintenance of information in working memory, both in rodents and humans. In particular, cross-frequency coupling between theta and gamma oscillations has been suggested as a core mechanism for multi-item memory. The aim of this work is to present an original neural network model, based on oscillating neural masses, to investigate mechanisms at the basis of working memory in different conditions. We show that this model, with different synapse values, can be used to address different problems, such as the reconstruction of an item from partial information, the maintenance of multiple items simultaneously in memory, without any sequential order, and the reconstruction of an ordered sequence starting from an initial cue. The model consists of four interconnected layers; synapses are trained using Hebbian and anti-Hebbian mechanisms, in order to synchronize features in the same items, and desynchronize features in different items. Simulations show that the trained network is able to desynchronize up to nine items without a fixed order using the gamma rhythm. Moreover, the network can replicate a sequence of items using a gamma rhythm nested inside a theta rhythm. The reduction in some parameters, mainly concerning the strength of GABAergic synapses, induce memory alterations which mimic neurological deficits. Finally, the network, isolated from the external environment (“imagination phase”) and stimulated with high uniform noise, can randomly recover sequences previously learned, and link them together by exploiting the similarity among items.

Theta and alpha power track the acquisition and reversal of threat predictions and correlate with skin conductance response

“…Further validation of this functional interpretation of brain rhythms may be achieved using mechanistic neural network models of oscillating brain populations (Ursino et al, 2010), reciprocally interconnected according to a biologically inspired connectivity. Indeed, such models have been recently used to analyze essential phenomena, such as the genesis of brain rhythms during sleep (Cona et al, 2014), the function of theta‐gamma coupling in the hippocampus (Cona & Ursino, 2013, 2015), the role of rhythms in working memory (Ursino et al, 2022). Hence, they represent a valid instrument to test the present interpretation on the role of each rhythm in the acquisition and reversal of threat predictions.…”

Robust working memory in a two-dimensional continuous attractor network

“…Indeed, the observation of stimulus-selective oscillatory dynamics in WM tasks has led to the hypothesis that neural oscillations in different frequency bands play an important role in the maintenance of information (Roux and Uhlhaas, 2014). Computational studies investigating this hypothesis typically use neural mass models with no inherent spatial structure (Ursino et al, 2023;Pina et al, 2018). Ensemble activity of cortical microcircuits comprising distinct cell types shows stable oscillations.…”

Rhythms of human attention and memory: An embedded process perspective

“…Key to our understanding of the specific neural network properties that may give rise to theta-gamma coupling dynamics (e.g., based on specific properties of cortical pyramidal cells; Lundqvist et al, 2011 ; Hummos and Nair, 2017 ) and how these dynamics may implement complex cognitive functions in hippocampal networks ( Ursino et al, 2022 ), are computational models (see Chadwick et al, 2015 , for a sequencing model based on fewer assumptions). So called attractor network models have also contributed to the alternative perspective that theta-gamma dynamics may arise as an epiphenomenon of the specific neurophysiology of neural networks, putting into debate their mechanistic function (e.g., Lundqvist et al, 2010 ; Kovach et al, 2018 ; Sheremet et al, 2019 ), discussed in more detail below.…”

“…A computational model of how the theta-gamma code may support the restoration and sequence coding of several learned items in WM, has recently been developed based on hippocampal connectivity and Hebbian learning principles ( Ursino et al, 2022 ). This model provides specific support for the idea that the theta-gamma code indeed supports those functions, derived from specific network characteristics.…”