Universality of neural population coarse-graining exponents

Apply phenomenological renormalization-group coarse-graining (iteratively merging maximally correlated neurons; Meshulam and co-workers, 2019) to recordings of $N \ge 10^{3}$ neurons and measure the variance-scaling exponent, the covariance-eigenvalue exponent and the dynamic exponent. Determine whether they are universal across regions, species and behavioral states, and identify the model class whose fixed point produces them.

The same analysis applied to many large recordings and to model networks with known critical and non-critical states, showing which exponents are universal.