Michael W. Cole

New version of Actflow Toolbox just released (colelab.github.io/ActflowToolbox/), adding glasso FC functionality. Demo: colelab.github.io/ActflowToolbox… We found that glasso FC is an improvement on the fMRI field-standard functional connectivity measure (correlation): doi.org/10.1101/2023.0…

The Science paper: science.org/doi/10.1126/sc… (now viral online, unfortunately!) A welcome correction that came out today: zenodo.org/records/180021…

@RaviMill This study is now peer reviewed and published at Nature Communications! Congratulations to first author @RaviMill Check out the paper here (open access): nature.com/articles/s4146…

Hence, progressive strengthening of conjunctive representations in cortex serves as a neural computational signature of effective task learning. Thanks to first author @RaviMill! Preprint here: doi.org/10.1101/2023.0…

Lab's latest: "Neural representation dynamics reveal computational principles of cognitive task learning", wherein we tracked dynamic changes in cortical-subcortical cognitive representations over practice, revealing computational principles underlying task learning: 1/n

Excited to see our paper with @TheColeLab finally out in peer-reviewed form @NatureComms ! Please check it out: “Dynamically shifting from compositional to conjunctive brain representations supports cognitive task learning”, doi.org/10.1038/s41467…

The graphical lasso functional connectivity approach that performed best in the paper can be implemented using the Activity Flow Toolbox: colelab.github.io/ActflowToolbox/ also using the paper's code release: github.com/ColeLab/Reliab… [12/n]

Together, results demonstrated vast improvements in fMRI functional connectivity estimation using regularized partial correlation. Thanks to first author Kirsten Peterson, and coauthors Ruben Sanchez-Romero and Ravi Mill! 11/n doi.org/10.1162/IMAG.a…

Lab’s latest out in Imaging Neuroscience, led by K. Peterson: “Regularized partial correlation provides reliable functional connectivity estimates while correcting for widespread confounding”, where we show a major improvement to standard fMRI func. connectivity (correlation) 1/n

@PessoaBrain @carrisa_cocuzza I wish you could have told that to some of our reviewers – they made it clear that strong localization of function is still assumed in some circles 🙃

@TheColeLab @carrisa_cocuzza I could have have told you that! 😉

Lab’s latest out at PLOS Comp Biol, led by @carrisa_cocuzza: “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]

@dj_bolger @carrisa_cocuzza Haha! Well, it was a matter of degree. Sure, much is distributed, but there were always those pesky localized visual category selective regions. I still think that functionality is localized, but due to converging distributed processes (not just within-region processes).

@TheColeLab @carrisa_cocuzza Dude! You spent your grad school years in the heart of PDP territory with a visual word researcher next to you. Surprised about distributed processing? 🤪

The full peer-reviewed (and open access) article is available here: doi.org/10.1371/journa… "Distributed network flows generate localized category selectivity in human visual cortex" Thanks to coauthors @carrisa_cocuzza @RUBENSARO @taku_ito1 @BrianKeaneLab and @RaviMill!

While surprising, this is broadly consistent with recent convolutional neural network models of vision (featuring cross-region/layer vector transformations), though it is less compatible with recurrent neural network models (featuring within-region processing) [10/N]