Scipy 最小化和步长约束

Question

I use the scipy.optimize.minimize function and I have a 1D array as x0 with different parameters inside it. Since I work with different physical parameters, I don't want, for example, to make an iteration of 0.1 for temperature of the order of magnitude of 2500K. So I want to put constraints on the function iterations for each inputs. I've tried this already:

x0 = [0.1, 10.0, 0.90,1200] #tau,star_diam,amC,Td
bounds = [ [0.005,1],[7,13],[0.05,1], [800,1500] ]

cons = []
for factor in range(len(bounds)):
    lower, upper = bounds[factor]
    l = {'type': 'ineq',
         'fun': lambda x, lb=lower, i=factor: x[i] - lb}
    u = {'type': 'ineq',
         'fun': lambda x, ub=upper, i=factor: ub - x[i]}
    cons.append(l)
    cons.append(u)


res = minimize(DUSTY_RUN,x0,args=(data_charac,lambda_max,lambda_min,DATA_DIR,INPUT_DIR), constraints=cons, method='COBYLA', options={'rhobeg': [0.01,0.01,0.01,100], 'maxiter': 5000, 'disp': False, 'catol': 1e-6})

In the options of the function, the rhoberg option is considered as float , I've tried to change it as a list/array but this didn't change anything as we may guess.

Answer 1

rhoberg is the initial step, so it doesn't directly control following steps. You can add scaling of parameters in your DUSTY_RUN function - since you wrote [0.01,0.01,0.01,100] you could multiply first three by 1e4 .