Emergence of distributed working memory in a human brain network model

Feng, Mengli; Bandyopadhyay, Abhirup; Mejías, Jorge F.

doi:10.1101/2023.01.26.525779

Cited by 6 publications

(3 citation statements)

References 75 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This model incorporates dynamic properties of multiple postsynaptic receptors (AMPA, NMDA, and GABA-A receptors), and is tightly constrained by state-of-the-art cortical connectivity data on mouse V1 22–23 which include cell densities and laminar-specific connectivity patterns across four different cell types (pyramidal neurons and PV, SST and VIP interneurons) and five laminar modules (layers 1, 2/3, 4, 5 and 6). Our simulations show that feedforward and feedback stimuli have opposing impacts on columnar activity, leading to net columnar excitation and inhibition, respectively, in agreement with experimental and other modelling work 17,24–26 . Furthermore, our model reveals a translaminar inhibitory effect mediated by layer 6 activity, in line with a role in columnar gain control role as suggested by experiments 27,28 .…”

Section: Introductionsupporting

confidence: 89%

Cell type specific firing patterns in a V1 cortical column model depend on feedforward and feedback-driven states

Moreni,

Pennartz,

Mejias

2024

Preprint

Self Cite

View full text Add to dashboard Cite

Stimulation of specific cell groups under different network regimes (e.g., spontaneous activity or sensory-evoked activity) can provide insights into the neural dynamics of cortical columns. While these protocols are challenging to perform experimentally, modelling can serve as a powerful tool for such explorations. Using detailed electrophysiological and anatomical data from mouse V1, we modeled a spiking network model of the cortical column microcircuit. This model incorporates pyramidal cells and three distinct interneuron types (PV, SST, and VIP cells, specified per lamina), as well as the dynamic and voltage-dependent properties of AMPA, GABA, and NMDA receptors. We first demonstrate that thalamocortical feedforward (FF) and feedback (FB) stimuli arriving in the column have opposite effects, leading to net columnar excitation and inhibition respectively and revealing translaminar gain control via full-column inhibition by layer 6. We then perturb one cell group (defined by a cell type in a specific layer) at a time and observe the effects on other cell groups under distinct network states: spontaneous, feedforward-driven, feedback-driven, and a combination of feedforward and feedback. Our findings reveal that when the same group is perturbed, the columnar response may vary significantly based on its state, with strong sensory feedforward input decreasing columnar sensitivity to all perturbations and feedback input serving as modulator of intra columnar interactions. Given that activity changes within specific neuronal populations are difficult to predict a priori and considering the practical challenges of conducting experiments, our computational simulations can serve as a critical tool to predict outcomes of perturbations and assist in the design of future experimental planning.

show abstract

Section: Introductionsupporting

confidence: 89%

Cell type specific firing patterns in a V1 cortical column model depend on feedforward and feedback-driven states

Moreni,

Pennartz,

Mejias

2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…T. Markov et al, 2014), leading to a brain model constrained at both the local and large-scale levels by neuroanatomical data. Using the layer-specificity and projection-directionality provided by this dataset, and following previous work (Feng et al, 2023; Mejias and Wang, 2022), we implemented a counterstream inhibitory bias in our network, which assumes that feedforward and feedback projections along the cortical hierarchy slightly but preferentially target excitatory and inhibitory neurons, respectively. Background input to cortical areas was varied across areas to mimic the differentiated thalamocortical projections across cortex and to facilitate that all cortical areas displayed a similar spontaneous activity level.…”

Section: Resultsmentioning

confidence: 99%

Distributed evidence accumulation across macaque large-scale neocortical networks during perceptual decision making

Zou,

Palomero-Gallagher,

Zhou

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Despite the traditional view of parietal cortex as an important region for perceptual decision-making, recent evidence suggests that sensory accumulation occurs simultaneously across many cortical regions. We explored this hypothesis by integrating connectivity, cellular and receptor density datasets and building a large-scale macaque brain model able to integrate conflicting sensory signals and perform a decision-making task. Our results reveal sensory evidence accumulation supported by a distributed network of temporal, parietal and frontal regions, with flexible sequential decision pathways which depend on task difficulty. The model replicates experimental lesioning effects and reveals that the causal irrelevance of parietal areas like LIP for decision performance is explained by compensatory mechanisms within a distributed integration process. The model also reproduces observed temporal gating effects of distractor timing during and after the integration process. Overall, our work hints at perceptual integration during decision-making as a broad distributed phenomenon and provides multiple testable predictions.

show abstract

“…In addition, noise can be presented at any point in time to disturb WM (Feng et al, 2023). During a task, WM is anticipated, initiated, disturbed by noise, updated and recalled.…”

Section: Wm Tasksmentioning

confidence: 99%

Can working memory be explained by predictive coding?

Feng

2023

Preprint

View full text Add to dashboard Cite

Predictive coding (PC) is a theory in cognitive/computational neuroscience which explains cortical functions with a hierarchical process of minimising prediction errors. It provides a neuronal scheme for implementing Bayesian inference in the brain to recover the hidden state of the world from sensory input (passive inference) and to select actions to reach the goals the agent has (active inference). Since its discovery, predictive coding has been found to be a unifying theory explaining more and more cognitive functions, including perception, attention, and action planning. In this literature thesis, I review and discuss how PC can be used also as a powerful tool to understand working memory (WM), an essential function for executive control.% Giving a brief introduction to working memory and current PC frameworks, I start with an overview of how WM might fit within predictive coding frameworks.Specifically, I try to explore how PC frameworks help with explaining the following questions: 1. how is WM maintained and updated? 2. What is the relationship between attention and WM and how do they interact? 3. why does WM have limited capacity? and 4. why is WM hierarchical? By treating WM coding as part of the state inference process, we can explain WM maintenance as the stage where the state variables remain the same when there is no new evidence. WM updates, on the other hand, correspond to belief updating when new evidence arises. Since there is a trade-off between prediction complexity and accuracy during state inference, the limited capacity of WM may be an emergent property to ensure a certain level of accuracy. In a process of active inference, attention helps the agent to select actions that reduce uncertainties about the world where selected actions give rise to observations that are used to update WM. This delineates the roles of WM and attention and clarifies the mechanism of their interactions. Finally, hierarchical PC can account for the hierarchical representation of working memory in the brain where each level of WM corresponds to each level of inferred states. Based on the reviewed literature, I summarised three important ingredients for modelling WM which are temporal depth, goals and hierarchy. Future work on modelling would be to clarify whether WM is a separable component in PC, which variable WM is actually represented in PC and where in the hierarchy WM is generated and maintained. In summary, through the lens of variational Bayesian inference, WM can be assessed in the process of evidence accumulation simulated in a deep hierarchical predictive coding model. With action selection incorporated, this naturally explains WM as an emergent property of goal-directed behaviour, manifested by hierarchical inference of the brain through the minimization of expected free energy. Modelling WM in PC frameworks provides alternative explanations to some long-standing questions about WM and may help with resolving the conflicts between WM theories, for example, the ones that propose either persistent or sparse neuronal activity during WM. It may also help with developing computational tools to improve treatments for brain disorders such as schizophrenia and facilitate artificial intelligence in coping with a world full of uncertainties.

show abstract

Emergence of distributed working memory in a human brain network model

Cited by 6 publications

References 75 publications

Cell type specific firing patterns in a V1 cortical column model depend on feedforward and feedback-driven states

Cell type specific firing patterns in a V1 cortical column model depend on feedforward and feedback-driven states

Distributed evidence accumulation across macaque large-scale neocortical networks during perceptual decision making

Can working memory be explained by predictive coding?

Contact Info

Product

Resources

About