You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The first step would involve all radius-related arguments to 'relative' radii, a number between 0 and 1 that refers to the range covered within the bounds, rather. The rest could be done in two ways:
A) x0, radius and prior_evals are scaled in each CUATRO instance by bounds. Surrogate fitting and minimization would happen in scaled space. Any inputs would have to be scaled into the original space before being added to results or before being passed to the black-box
B) Keeping CUATRO in original input space, and changing the condition for samples being inside the trust region inside the quadratic minimization constraints `cp.norm(X - trust_center) <= radius' and other functions like samples_in_trust() from
ind = np.where(np.linalg.norm(X - np.array(center), axis = 1, keepdims = True) < radius)[0]
to something like ind = np.where([np.sum([(X_in[j,i]-x_c[i])**2*4/(bounds[i,1]-bounds[i,0])**2/(radius)**2 for i in range(len(x_c))]) <= 1 for j in range(len(X_in))])[0]
The text was updated successfully, but these errors were encountered:
The first step would involve all radius-related arguments to 'relative' radii, a number between 0 and 1 that refers to the range covered within the bounds, rather. The rest could be done in two ways:
A) x0, radius and prior_evals are scaled in each CUATRO instance by bounds. Surrogate fitting and minimization would happen in scaled space. Any inputs would have to be scaled into the original space before being added to results or before being passed to the black-box
B) Keeping CUATRO in original input space, and changing the condition for samples being inside the trust region inside the quadratic minimization constraints `cp.norm(X - trust_center) <= radius' and other functions like samples_in_trust() from
ind = np.where(np.linalg.norm(X - np.array(center), axis = 1, keepdims = True) < radius)[0]
to something like
ind = np.where([np.sum([(X_in[j,i]-x_c[i])**2*4/(bounds[i,1]-bounds[i,0])**2/(radius)**2 for i in range(len(x_c))]) <= 1 for j in range(len(X_in))])[0]
The text was updated successfully, but these errors were encountered: