def candidate_difference_squared(beta, index1, index2):
B_betaJ, B_betaI, y_beta = calc_B_y_beta_selection(A,
y,
svdAAt_U,
svdAAt_S,
beta,
I=support,
J=[index1, index2])
hit_candidates, used_signs = calc_hit_cand_selection(
B_betaI, B_betaJ, y_beta, signum)
return (hit_candidates[0] - hit_candidates[1])**2
def get_substitute_for_last_entry_changed(beta):
B_betaJ, B_betaI, y_beta = calc_B_y_beta_selection(
A,
y,
svdAAt_U,
svdAAt_S,
beta,
I=support,
J=np.array([last_entry_changed]).astype('int'))
return calc_hit_cand_selection(B_betaI,
B_betaJ,
y_beta,
signum,
prescribed_sign=sign_last_entry_changed)
def candidate_difference_squared(beta, index1, index2):
I_in_support = np.where(np.in1d(support, np.array([index1,
index2])))[0]
J = np.setdiff1d([index1, index2], support[I_in_support])
B_betaJ, B_betaI, y_beta = calc_B_y_beta_selection(A,
y,
svdAAt_U,
svdAAt_S,
beta,
I=support,
J=J)
if index1 in support and index2 in support:
candidates = calc_cross_cand_selection(B_betaI,
y_beta,
signum,
indices=I_in_support)
elif index1 in support:
cross_cand = calc_cross_cand_selection(B_betaI,
y_beta,
signum,
indices=I_in_support)
hit_cand, used_signs = calc_hit_cand_selection(
B_betaI, B_betaJ, y_beta, signum)
candidates = [cross_cand, hit_cand]
elif index2 in support:
cross_cand = calc_cross_cand_selection(B_betaI,
y_beta,
signum,
indices=I_in_support)
hit_cand, used_signs = calc_hit_cand_selection(
B_betaI, B_betaJ, y_beta, signum)
candidates = [cross_cand, hit_cand]
else:
candidates, used_signs = calc_hit_cand_selection(
B_betaI, B_betaJ, y_beta, signum)
return (candidates[0] - candidates[1])**2