“…The undetermined parameter set to be optimized was Φ = { w ee , w ie , w ei , w ii , τ i , τ e , r E,b , c, s, k } ( Equation 1 ). The route optimization process started with a set of initial parameters Φ ini = { w ee ,0 , w ie ,0 , w ei ,0 , w ii ,0 , τ i ,0 , τ e ,0 , r E,b ,0 , c 0 , s 0 , k 0 } ( Table S4 ), which were determined as follows: (i) The initial connectivity strength parameters ({ w ee ,0 , w ie ,0 , w ei ,0 , w ii ,0 }) were all set to be 1; (ii) { τ i ,0 , τ e ,0 , r E,b ,0 } were from the anatomical and experimental literature 103,9,52,54,53,104 (iii) { c 0 , s 0 , k 0 } were from fitting the reference firing rate with our previous single-ensemble model of the Vim neural group receiving DBS 22 . Starting from the initial parameter set Φ ini , we did a preliminary model fit based on Stabilized Feature (i.e., stressing data from both low and high frequencies of DBS), and the weight of each DBS frequency was given by the vector g pre ( Equation 10, Table S3 ).…”