Michael W. Cole
@mwcole.bsky.social
Professor, director of neuroscience lab at Rutgers University – neuroimaging, cognitive control, network neuroscience
Writing book “Brain Flows: How Network Dynamics Generate The Human Mind” for Princeton University Press
https://www.colelab.org
Writing book “Brain Flows: How Network Dynamics Generate The Human Mind” for Princeton University Press
https://www.colelab.org
And regularization improved prediction of individual differences in demographics (age) and behavior/cognition (general intelligence) relative to standard partial correlation. The glasso results were more interpretable than pairwise correlation (fewer false connections) 10/n
September 14, 2025 at 9:34 PM
And regularization improved prediction of individual differences in demographics (age) and behavior/cognition (general intelligence) relative to standard partial correlation. The glasso results were more interpretable than pairwise correlation (fewer false connections) 10/n
Also empirical, prediction of task-evoked activity (via activity flow modeling) was better with regularized partial correlation 9/n
September 14, 2025 at 9:34 PM
Also empirical, prediction of task-evoked activity (via activity flow modeling) was better with regularized partial correlation 9/n
First empirical validation: regularized partial correlation was much closer to structural connectivity, which doesn’t have the causal confounding problem (despite other issues) 7/n
September 14, 2025 at 9:34 PM
First empirical validation: regularized partial correlation was much closer to structural connectivity, which doesn’t have the causal confounding problem (despite other issues) 7/n
This pattern of results was mirrored in empirical resting-state fMRI data across 4 validation measures. Regularization was key to estimating individual subject-level networks with reduced confounding. 6/n
September 14, 2025 at 9:34 PM
This pattern of results was mirrored in empirical resting-state fMRI data across 4 validation measures. Regularization was key to estimating individual subject-level networks with reduced confounding. 6/n
In simulations, pairwise (standard) correlation led to many false connections, but so did partial correlation. Regularized partial correlation (glasso) better recovered the true network organization 5/n
September 14, 2025 at 9:34 PM
In simulations, pairwise (standard) correlation led to many false connections, but so did partial correlation. Regularized partial correlation (glasso) better recovered the true network organization 5/n
We hypothesized that low reliability of partial correlation is due to overfitting to noise, with regularization (model simplification) improving reliability. 4/n
September 14, 2025 at 9:34 PM
We hypothesized that low reliability of partial correlation is due to overfitting to noise, with regularization (model simplification) improving reliability. 4/n
Pairwise correlations are known to be susceptible to false positives in theory. For example, region A causing activity in unconnected regions B and C (B<-A->C) can lead to a false B-C connection. Partial correlation can correct for this error, but not reliably 3/n
September 14, 2025 at 9:34 PM
Pairwise correlations are known to be susceptible to false positives in theory. For example, region A causing activity in unconnected regions B and C (B<-A->C) can lead to a false B-C connection. Partial correlation can correct for this error, but not reliably 3/n
Lab’s latest is out in Imaging Neuroscience, led by Kirsten Peterson: “Regularized partial correlation provides reliable functional connectivity estimates while correcting for widespread confounding”, where we demonstrate a major improvement to standard fMRI functional connectivity (correlation) 1/n
September 14, 2025 at 9:34 PM
Lab’s latest is out in Imaging Neuroscience, led by Kirsten Peterson: “Regularized partial correlation provides reliable functional connectivity estimates while correcting for widespread confounding”, where we demonstrate a major improvement to standard fMRI functional connectivity (correlation) 1/n
“Cognitive flexibility as the shifting of brain network flows by flexible neural representations”, a solo paper by yours truly, making the case that brain activity flow shifts are essential to mental flexibility (and quite interesting too!)
Open access: www.sciencedirect.com/science/arti...
Open access: www.sciencedirect.com/science/arti...
December 3, 2024 at 6:12 PM
“Cognitive flexibility as the shifting of brain network flows by flexible neural representations”, a solo paper by yours truly, making the case that brain activity flow shifts are essential to mental flexibility (and quite interesting too!)
Open access: www.sciencedirect.com/science/arti...
Open access: www.sciencedirect.com/science/arti...
But is the visual hierarchy important for category selectivity? Yes, adding V1-indirect flows improved flow model accuracy. This suggests V1-direct flows generate category selectivity, but complex hierarchical flows add additional category selective neural activity [8/N]
November 21, 2024 at 2:18 PM
But is the visual hierarchy important for category selectivity? Yes, adding V1-indirect flows improved flow model accuracy. This suggests V1-direct flows generate category selectivity, but complex hierarchical flows add additional category selective neural activity [8/N]
We found that the model using direct connectivity from V1 to each category-selective visual region was effective at generating category selectivity, demonstrating activity flows directly from V1 are sufficient for generating visual category selectivity [7/N]
November 21, 2024 at 2:18 PM
We found that the model using direct connectivity from V1 to each category-selective visual region was effective at generating category selectivity, demonstrating activity flows directly from V1 are sufficient for generating visual category selectivity [7/N]
This whole-brain model did not reflect the visual system’s hierarchical organization (starting in V1), however. We then tested a series of V1-initiated models, testing the role of distributed V1-initiated activity flows in generating localized visual responses [6/N]
November 21, 2024 at 2:18 PM
This whole-brain model did not reflect the visual system’s hierarchical organization (starting in V1), however. We then tested a series of V1-initiated models, testing the role of distributed V1-initiated activity flows in generating localized visual responses [6/N]
We first tested a whole-brain model, with all empirical stimulus-evoked activity outside the ROI generating (via activity flows) localized category selective responses. Selective responses were generated for all visual categories (faces, places, body parts, and tools) [5/N]
November 21, 2024 at 2:18 PM
We first tested a whole-brain model, with all empirical stimulus-evoked activity outside the ROI generating (via activity flows) localized category selective responses. Selective responses were generated for all visual categories (faces, places, body parts, and tools) [5/N]
We used a rigorous empirical approach with fMRI to estimate activity flows generating stimulus-evoked activity: activity flow mapping (colelab.org/pubs/2024_Ac...). Further, causal principles were used to identify more valid functional connections for modeling activity flows [4/N]
November 21, 2024 at 2:18 PM
We used a rigorous empirical approach with fMRI to estimate activity flows generating stimulus-evoked activity: activity flow mapping (colelab.org/pubs/2024_Ac...). Further, causal principles were used to identify more valid functional connections for modeling activity flows [4/N]
A thread...
Lab’s latest at PLOS Comp Biol, led by
@carrisacocuzza.bsky.social: “Distributed network flows generate localized category selectivity in human visual cortex”. This one changed how I think the brain works! Even "localized" functions are likely generated by distributed processes [1/N]
Lab’s latest at PLOS Comp Biol, led by
@carrisacocuzza.bsky.social: “Distributed network flows generate localized category selectivity in human visual cortex”. This one changed how I think the brain works! Even "localized" functions are likely generated by distributed processes [1/N]
November 21, 2024 at 2:18 PM
A thread...
Lab’s latest at PLOS Comp Biol, led by
@carrisacocuzza.bsky.social: “Distributed network flows generate localized category selectivity in human visual cortex”. This one changed how I think the brain works! Even "localized" functions are likely generated by distributed processes [1/N]
Lab’s latest at PLOS Comp Biol, led by
@carrisacocuzza.bsky.social: “Distributed network flows generate localized category selectivity in human visual cortex”. This one changed how I think the brain works! Even "localized" functions are likely generated by distributed processes [1/N]
Finally, to demonstrate how activity flow models can help explain how cognitive effects emerge from brain network interactions, we define a PC-based model for a dorsolateral prefrontal cortex region during a 2- vs 0-back working memory task—and compared with correlation (8/11)
February 19, 2024 at 10:16 PM
Finally, to demonstrate how activity flow models can help explain how cognitive effects emerge from brain network interactions, we define a PC-based model for a dorsolateral prefrontal cortex region during a 2- vs 0-back working memory task—and compared with correlation (8/11)
Superior performance of mechanistic activity flow models with PC-networks was also confirmed for generating the pattern of activations across conditions (24) for individual brain regions (360). A notable improvement was observed for language and somatomotor networks (7/11)
February 19, 2024 at 10:16 PM
Superior performance of mechanistic activity flow models with PC-networks was also confirmed for generating the pattern of activations across conditions (24) for individual brain regions (360). A notable improvement was observed for language and somatomotor networks (7/11)
To show the notable improvements in prediction from causally-valid FC architectures, we compared the field-standard correlation vs PC—the tested method with the strongest causal principles—with whole-brain predictions, and visualized a motor and a memory task (6/11)
February 19, 2024 at 10:16 PM
To show the notable improvements in prediction from causally-valid FC architectures, we compared the field-standard correlation vs PC—the tested method with the strongest causal principles—with whole-brain predictions, and visualized a motor and a memory task (6/11)
We confirmed in empirical fMRI data the benefits of causal FC methods, showing that activity flow models parameterized with causal FC networks can accurately generate task-evoked activity, using a small set of network sources with a more valid mechanistic interpretation (5/11)
February 19, 2024 at 10:16 PM
We confirmed in empirical fMRI data the benefits of causal FC methods, showing that activity flow models parameterized with causal FC networks can accurately generate task-evoked activity, using a small set of network sources with a more valid mechanistic interpretation (5/11)
We tested this continuum in simulations, showing that causal FC methods—combinedFC & PC (Peter-Clark)—recovered networks with high precision, resulting in accurate activity flow models with a small set of predictors and, for PC-based models, direction of influence (4/11)
February 19, 2024 at 10:16 PM
We tested this continuum in simulations, showing that causal FC methods—combinedFC & PC (Peter-Clark)—recovered networks with high precision, resulting in accurate activity flow models with a small set of predictors and, for PC-based models, direction of influence (4/11)
Leveraging their causal validity, we can use FC methods to build directed activity flow models in which the task-evoked activity of a target region is predicted from the activity of its direct causal source regions, and thus have a stronger mechanistic interpretation (3/11)
February 19, 2024 at 10:16 PM
Leveraging their causal validity, we can use FC methods to build directed activity flow models in which the task-evoked activity of a target region is predicted from the activity of its direct causal source regions, and thus have a stronger mechanistic interpretation (3/11)
We began by proposing a continuum to order FC methods according to their causal principles. Methods with more causal validity better control for confounders, chains and colliders and can infer direction of influence, bringing us closer to the true functional networks (2/11)
February 19, 2024 at 10:16 PM
We began by proposing a continuum to order FC methods according to their causal principles. Methods with more causal validity better control for confounders, chains and colliders and can infer direction of influence, bringing us closer to the true functional networks (2/11)
Lab’s latest out at NeuroImage (submitted prior to editor changes; led by Ruben Sanchez-Romero): “Causally informed activity flow models provide mechanistic insight into network-generated cognitive activations” (1/11)
February 19, 2024 at 10:16 PM
Lab’s latest out at NeuroImage (submitted prior to editor changes; led by Ruben Sanchez-Romero): “Causally informed activity flow models provide mechanistic insight into network-generated cognitive activations” (1/11)
SFN poster 4) Neural representation dynamics reveal computational principles of cognitive task learning” abstractsonline.com/pp8/#!/10892..., VV39, Wed 11/15 8am-12pm
If you’re going to SFN please come check out these posters! N/N
If you’re going to SFN please come check out these posters! N/N
November 9, 2023 at 7:00 PM
SFN poster 4) Neural representation dynamics reveal computational principles of cognitive task learning” abstractsonline.com/pp8/#!/10892..., VV39, Wed 11/15 8am-12pm
If you’re going to SFN please come check out these posters! N/N
If you’re going to SFN please come check out these posters! N/N
SFN poster 3) “Regularized partial correlation provides reliable functional connectivity estimates while correcting for widespread confounding” abstractsonline.com/pp8/#!/10892..., XX42, Mon 11/13 1-5pm
3/N
3/N
November 9, 2023 at 6:59 PM
SFN poster 3) “Regularized partial correlation provides reliable functional connectivity estimates while correcting for widespread confounding” abstractsonline.com/pp8/#!/10892..., XX42, Mon 11/13 1-5pm
3/N
3/N