def test_assign_opt4():
P0 = np.array([[
32.2839240000000, 3.21131500000000, 32.2839240000000, 32.2839240000000,
32.2839240000000, 32.2839240000000, 32.2839240000000, 22.2960390000000,
32.2839240000000, np.inf, np.inf, np.inf
],
[
32.2839240000000, 1.92542300000000, 32.2839240000000,
32.2839240000000, 32.2839240000000, 32.2839240000000,
32.2839240000000, 11.4484150000000, 32.2839240000000,
6.16967300000000, 6.16967300000000, 6.16967300000000
],
[
0, 32.2839240000000, 32.2504870000000, 32.2839240000000,
32.2839240000000, 32.2839240000000, 32.2839240000000,
32.2839240000000, 32.2839240000000, 6.16967300000000,
6.16967300000000, 6.16967300000000
]])
(S0, C0) = g_sub.assign_opt(P0)
exp_C0 = 9.380988
exp_S0 = np.array([1, 9, 0])
test.assert_allclose(C0, exp_C0)
test.assert_array_equal(S0, exp_S0)
def test_assign_opt1():
P0 = np.array([
26.1142509467173, 26.1142509467173, 26.1142509467173, 26.1142509467173,
0
])
(S0, C0) = g_sub.assign_opt(P0)
exp_C0 = 0
exp_S0 = np.array([4])
test.assert_allclose(C0, exp_C0)
test.assert_array_equal(S0, exp_S0)
def test_assign_opt3():
P0 = np.array(
[[np.inf, np.inf, np.inf, np.inf, np.inf, np.inf, np.inf, np.inf],
[
np.inf, 32.9419076165403, 32.9419076165403, 32.9419076165403,
32.9419076165403, 32.9419076154638, np.inf, 6.82765666982308
]])
(S0, C0) = g_sub.assign_opt(P0)
exp_C0 = 6.82765666982308
exp_S0 = np.array([-1, 7])
test.assert_allclose(C0, exp_C0)
test.assert_array_equal(S0, exp_S0)
def test_assign_opt2():
P0 = np.array([[
32.9419076165403, 32.9419076165403, 32.9419076165403, 17.4668722633371,
32.9419076165403, 32.9419076165403, 6.82765666982308, 6.82765666982308
],
[
0, 32.9419076165403, 32.9419076165403, 32.9419076165403,
32.9419076165403, 32.9419076154638, 6.82765666982308,
6.82765666982308
]])
(S0, C0) = g_sub.assign_opt(P0)
exp_C0 = 6.82765666982308
exp_S0 = np.array([6, 0])
test.assert_allclose(C0, exp_C0)
test.assert_array_equal(S0, exp_S0)
def __murty_helper(p0, m):
(s0, c0) = subs.assign_opt(p0)
s0 = s0.T
if m == 1:
return (s0.reshape((1, s0.size)), np.array([c0]))
(n_rows, n_cols) = p0.shape
# preallocate arrays
blk_sz = 1000
ans_lst_P = np.zeros((n_rows, n_cols, blk_sz))
ans_lst_S = np.zeros((n_rows, blk_sz), dtype=int)
ans_lst_C = np.nan * np.ones(blk_sz)
ans_lst_P[:, :, 0] = p0
ans_lst_S[:, 0] = s0.T
ans_lst_C[0] = c0
ans_nxt_ind = 1
assigns = np.nan * np.ones((n_rows, m))
costs = np.zeros(m)
for ii in range(0, m):
# if cleared break
if np.isnan(ans_lst_C).all():
assigns = assigns[:, 0:ans_nxt_ind]
costs = costs[0:ans_nxt_ind]
break
# find lowest cost solution
idx_top = np.nanargmin(ans_lst_C[0:ans_nxt_ind])
assigns[:, ii] = ans_lst_S[:, idx_top]
costs[ii] = ans_lst_C[idx_top]
P_now = ans_lst_P[:, :, idx_top]
S_now = ans_lst_S[:, idx_top]
ans_lst_C[idx_top] = np.nan
for (aw, aj) in enumerate(S_now):
if aj >= 0:
P_tmp = P_now.copy()
if aj <= n_cols - n_rows - 1:
P_tmp[aw, aj] = np.inf
else:
P_tmp[aw, (n_cols - n_rows):] = np.inf
(S_tmp, C_tmp) = subs.assign_opt(P_tmp)
S_tmp = S_tmp.T
if (S_tmp >= 0).all():
# allocate more space as needed
if ans_nxt_ind >= len(ans_lst_C):
ans_lst_P = np.concatenate(
(ans_lst_P, np.zeros((n_rows, n_cols, blk_sz))),
axis=2)
ans_lst_S = np.concatenate(
(ans_lst_S, np.zeros((n_rows, blk_sz), dtype=int)),
axis=1)
ans_lst_C = np.hstack(
(ans_lst_C, np.nan * np.ones(blk_sz)))
ans_lst_P[:, :, ans_nxt_ind] = P_tmp
ans_lst_S[:, ans_nxt_ind] = S_tmp
ans_lst_C[ans_nxt_ind] = C_tmp
ans_nxt_ind += 1
v_tmp = P_now[aw, aj]
P_now[aw, :] = np.inf
P_now[:, aj] = np.inf
P_now[aw, aj] = v_tmp
return (assigns.T, costs)