What question did this study set out to answer?

This research focuses on understanding how strong and localized recurrence affects the dimensionality of neural activity across different brain areas.

February 8, 2026Open Access

Strong and localized recurrence controls dimensionality of neural activity across brain areas

Key Points

This research focuses on understanding how strong and localized recurrence affects the dimensionality of neural activity across different brain areas.
Code collection reproducing results from electrophysiological data analyses.
Pipeline setup for dimensionality analysis involving hmm fitting and preprocessed datasets.
Analyses include theoretical results and synaptic physiology data evaluations.
Demonstrated dimensionality control through localized recurrence in neural networks.
Reproduced analyses with notable figures shown in the respective output of the pipeline.
Findings support theoretical implications about synaptic physiology and neural activity patterns.

Abstract

This code collection contains the code to reproduce the results in the paper Dahmen, Recanatesi et al. "Strong and localized recurrence controls dimensionality of neural activity across brain areas". The collection contains multiple code repositories also available on github at the links below. Some of the required data are available in this collection. The missing ones can be found, along with the code and the final figures, at: https: //codeocean. technionneuroailab. com/collections/e85debba-078a-4e64-85be-8e72a175c02d? filter=all The results of the paper map to this collection according to three streams: all the results that pertain analysis of electrophisiology data (hmm, dimensionality analysis, Fig 1-2 of the paper etc. ) are reproduced by running the pipeline. Generation of the entire results takes about 4-5 days when running the pipeline from scratch. To facilitate the reproduction of the figures we have included an already preprocessed dataset (the results of the hmm fit) and attached to the pipeline. In order for the pipeline to avoid using such data (rerunning all results from scratch) the following flags in the config. py of the capsule hmmfitting need to be set to True (RUNSTEP2PREPROCESS = True, RUNSTEP3HMMXVAL = False, RUNSTEP4HMMANALYSIS = False). Please note that the output figures of this entire process can already be observed in the results of the pipeline. Github repositories of the capsules in this pipeline, with further details, can be found at: https: //github. com/TechnionNeuroAILab/DahmenRecanatesiStrongCouplingₕmmfittingstagewithrestart. git https: //github. com/TechnionNeuroAILab/DahmenRecanatesiStrongCouplingfigures. git a second stream of analysis is the one pertatining theoretical results. Most of such results are generated running the capsule TheoreticalandCalciumfigures. Please note that the output figures of this entire process can already be observed in the results of the capsule. A public github repository for this capsule, with a more detailed description, can be found at https: //github. com/TechnionNeuroAILab/DahmenRecanatesiStrongCouplingTheoreticalandCalciumfigures. git finally a third stream of results, pertaining the analysis of synaptic physiology data, can be reproduced by running the aisynphys capsule. A public github repository for this capsule, with a more detailed description, can be found at: https: //github. com/TechnionNeuroAILab/DahmenRecanatesiStrongCouplingₐisynphys. git

Strong and localized recurrence controls dimensionality of neural activity across brain areas

Key Points

Abstract

Cite This Study