Table 1.

Parameters of models used in different figures.

Parameter	Fig. 1	Figs. 2–3	Fig. 4	Fig. 5	Fig. S1	Unit	Description
n	2					-	power-law I/O exponent
k	0.0194					mV−2 · ms	power-law I/O pre-factor
τ corr	5					ms	noise correlation time
τ AMPA	5					ms	AMPA decay time
τ GABA	7					ms	GABAA decay time
τ NMDA	100					ms	NMDA decay time
ρ N	0.39	0 – 0.5	0.39	0.3 – 0.5	0.45	-	NMDA share of excitation
J EE	124	100 – 300	124	100 – 300	165	mV	total E → E connection weight
J IE	116	100 – 300	116	100 – 300	123	mV	total E → I connection weight
J EI	103	50 – 150	103	50 – 150	114	mV	total I → E connection weight
J II	59.3	50 – 150	59.3	50 – 150	57.1	mV	total I → I connection weight
g E	21.9	10 – 30	21.9	10 – 30	21.7	mV/s	E feedforward current per 1% contrast
g I	10.3	5 – 15	10.3	5 – 15	10.6	mV/s	I feedforward current per 1% contrast
λEE	-	-	0.72	0.25 – 0.75†	0	-	locality of E → E connections
λIE	-	-	0.70	0.25 – 0.75†	0	-	locality of E → I connections
σ EE	-	-	0.296	0.15 – 0.60	0.265	mm	range of E → E connections
σ IE	-	-	0.554	0.15 – 0.60	0.294	mm	range of E → I connections
σ EI	-	-	0.09			mm	range of I → E connections
σ II	-	-	0.09			mm	range of I → I connections
N col	1	1	172			-	number of cortical columns
L	-	-	6.4			mm	retinotopic network width
Δx	-	-	0.4			mm	cortical column width
M	-	-	2			mm/degrees	cortical magnification factor
w RF	-	-	0.04°			degrees	grating input’s margin width
σ Gabor	-	-	0.5°			degrees	Gabor stimulus sigma

In figures 2,3 and 5, parameters were sampled independently and uniformly from the ranges given in the table, except for enforcing three inequality constraints (i.e., sampled parameter sets violating any of these inequalities were rejected). See the main text (Methods) for details.

^†:these were the ranges for sampled λ_EE and λ_IE of the columnar model; these parameters were zero in the non-columnar model.