Learning how network structure shapes decision-making for bio-inspired computing

Schirner, Michael; Deco, Gustavo; Ritter, Petra

doi:10.1038/s41467-023-38626-y

Cited by 19 publications

(11 citation statements)

References 84 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our results align with a recent study by Schirner et al ( 2023 ), which found that participants in the Human Connectome Project who had higher Penn Matrix Reasoning scores were those who took longer on hard questions. They linked the differential time allocation to easy and hard problems to measures of functional connectivity, finding that slower solvers had higher resting state connectivity.…”

Section: Discussionsupporting

confidence: 92%

Response to Difficulty Drives Variation in IQ Test Performance

Cheyette,

Piantadosi

2024

Open Mind

View full text Add to dashboard Cite

In a large (N = 300), pre-registered experiment and data analysis model, we find that individual variation in overall performance on Raven’s Progressive Matrices is substantially driven by differential strategizing in the face of difficulty. Some participants choose to spend more time on hard problems while others choose to spend less and these differences explain about 42% of the variance in overall performance. In a data analysis jointly predicting participants’ reaction times and accuracy on each item, we find that the Raven’s task captures at most half of participants’ variation in time-controlled ability (48%) down to almost none (3%), depending on which notion of ability is assumed. Our results highlight the role that confounding factors such as motivation play in explaining individuals’ differential performance in IQ testing.

show abstract

Section: Discussionsupporting

confidence: 92%

Response to Difficulty Drives Variation in IQ Test Performance

Cheyette,

Piantadosi

2024

Open Mind

View full text Add to dashboard Cite

show abstract

“…The simulation of FC via noise integration is easier to compare with previous work as it has been extensively simulated and studied (8). Our results do not have the highest correlation in comparison to other methods, as some recent studies have correlations utilizing higher parcellations above 0.9 compared to our average correlation of about 0.7 (32). However, as illustrated in Supplemental Figure 9, the FC metric is the least specific in analyzing dynamical systems and is rather uninformative and we believe that the prediction of the actual timeseries are much more informative of the performance of the model.…”

Section: Discussioncontrasting

confidence: 88%

“…Compared to other estimates of behavioral correlates using HCP WM data, our results are quite promising as getting r-squared beyond 0.4 seems challenging (37). Another recent paper that uses network modeling to make inferences about WM using HCP data, but instead looks at average RT metrics across WM trials rather than predicting them on a per trail basis (32). Our results compared to this paper are contradictory, as it claims that there are fast and slow learners and proceeds to separate the HCP population into 5 distinct groups.…”

Section: Discussionmentioning

confidence: 99%

“…This could be due to the fact that the paper makes a lot of assumptions on the structure of what network model should be used to model task fMRI and that these results actually don’t translate to the actual data. The short comings of these findings illustrate that it is much more important to focus on the transient signals especially when using these network models that are differential equations rather than take average metrics over time (32). While we find our results promising in separating rest from task, further work needs to be done in fully understanding task activation and we hope this work will act as a catalyst and facilitate more research in this direction.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Using an ODE model to separate Rest and Task signals in fMRI

Kashyap,

Geenjaar,

Bey

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Ongoing cortical activity is shaped by the interactions between brain regions connected on a macroscopic level. While components of the activity reflect specific processes responding to a presented stimulus, the vast majority persists as a background activity that innately exists due to cortical loops present in the connectivity. Simulations of the background activity, commonly known as resting state, have been developed over recent years using sophisticated ordinary differential equations (ODE) taking into account the macroscale organization of the cortex. However, the relationship between the rest activity compared to task or stimulus driven activity remains unclear although several different models have been proposed relating them. In this study, we develop a novel method of testing the relationship between rest and task by utilizing the framework that cortical activity can be represented by an ODE. The methodology uses a well-tested data driven approach known as Sparse Identification of Nonlinear Dynamics (SINDy) to construct an ODE to represent both rest and task functional Magnetic Response Imaging (fMRI) data separately. Since this technique is still relatively novel in the context of neuroscience, we validate the algorithm in its ability to identify an ODE that correctly predicts the signal dynamics, has a structure similar to those models currently used to recapitulate rest activity, and its ability to produce an impulse response as measured in fMRI data. Then, we systematically test the relationship between the two ODE models representing rest and task in order to model the task independent network activity. The task independent component is then subsequently removed from the measured signal and we utilize behavioral measures such as reaction time measured on a trial by trial basis to test if we have successfully separated the signal, as behavioral variables would be more correlated to the stimulus dependent activity than the raw unseparated signal. Our results show evidence that the stimulus independent signal is equivalent to all the processes in Rest that are not present in Task and can be estimated by subtracting the Task model from the Rest model. This suggests that Task can be approximated as a subset of Rest. We believe that our results are pertinent to the scientific community as they represent one of the first steps on using an ODE model to understanding the relationship of stimulus response in the cortex and allow for better decoding in terms of predicting the behavioral metrics.Significance StatementThis article represents a pioneering effort in linking white matter structural connectivity, resulting in measured fMRI brain activity, and behavioral metrics within a single causal ODE model. Based solely on regression, a numerical approximation for intrinsic fMRI activity is constructed and then applied to separate aspects of the task fMRI activity based on background processes and stimulus-dependent processes. Therefore, this work represents a significant advancement towards modeling and interpreting fMRI data.

show abstract

“…Dimension 2 loadings, interpreted as processing speed, were high for tests considering reaction time. Our findings may relate to the accuracy-speed relationship described by Schirner et al (2023) using alternative estimates of g and processing speed, where accuracy is linked to high FC strength (thus FC stability), and speed with low FC strength (thus FC variability). Accuracy is facilitated by a mode of cognition where high synchrony yields longer decision -making windows for information integration, whereas speed is facilitated by a mode where low synchrony grants shorter windows for rapid decision-making.…”

Section: Discussionmentioning

confidence: 67%

Higher general intelligence is linked to stable, efficient, and typical dynamic functional brain connectivity patterns

Ng¹,

Yu²,

Feusner³

et al. 2023

Preprint

View full text Add to dashboard Cite

The capacity of the brain to adaptively and flexibly reconfigure into different network connectivity patterns may underlie general cognitive ability, or g. Network reconfiguration can be captured via assessments of dynamic connectivity (dFC), which quantifies dominant temporally recurrent connectivity patterns, or “states”, common across the population. While standard dynamic measures focus on quantifying relative time spent within states, or the probability transitioning between states, these metrics fail to capture variability between individuals present in connectivity patterns across states. Here, we provide individualized assessments of connectivity flexibility and stability over time by considering within-state pattern stability, the difference in patterns across state transitions, and how well a given state represents the “typical” state in a particular individual. We leveraged resting-state fMRI data from the large-scale Human Connectome Project and data-driven multivariate Partial Least Squares Correlation to examine emergent relationships between dynamic network properties and cognition. We found that higher g was associated with maintaining distinct states over other states, efficient reconfiguration (i.e., less pattern change during common small-magnitude state transitions such as a state to itself, and greater pattern change among rare transitions between very different states), and less reconfiguration away from population-typical patterns. These results demonstrate that higher cognitive abilities are associated with greater state-specific stability, greater connectivity differences when transitioning between distinct states, and reconfiguration into more typical, potentially optimal, connectivity patterns within states. This suggests a link between general cognition and the efficiency of reconfiguration connectivity patterns into stable, well-defined, and typical network states.Significance StatementCognitive performance can be summarized by a general capability - “g”. g has been associated with academic and professional achievement and is critical in a world where higher-order cognition is increasingly necessary to excel. Attempts to improve g are hampered by a lack of mechanistic understanding. A recent theory suggests that the capacity to reconfigure brain connectivity drives g. We investigate metrics which characterize switches in recurring dominant connectivity patterns and propose novel metrics which investigate connectivity changes during the switches and connectivity pattern deviation from the population. Our findings suggest that g is associated with the maintenance of special patterns, greater pattern changes when necessary, and having population-typical patterns. This advances our understanding of g and informs future interventions.

show abstract

Learning how network structure shapes decision-making for bio-inspired computing

Cited by 19 publications

References 84 publications

Response to Difficulty Drives Variation in IQ Test Performance

Response to Difficulty Drives Variation in IQ Test Performance

Using an ODE model to separate Rest and Task signals in fMRI

Higher general intelligence is linked to stable, efficient, and typical dynamic functional brain connectivity patterns

Contact Info

Product

Resources

About