Combined line search failed for a reasonable sized problem
xinwang0312 opened this issue · 1 comments
xinwang0312 commented
The ECOS fails to converge on the following problem, which indicates Combined line search failed :
Could someone take a look and see what is going on? Thanks in advance.
Environment: I'm using CVXPY 1.1.7, ECOS '2.0.7.post1', numpy 1.19.1, python 3.7.5
Following code can be used to reproduce the problem. A and b can be found in the attachement
import numpy as np
import cvxpy as cp
with open('A.npy', 'rb') as f:
A = np.load(f)
with open('b.npy', 'rb') as f:
b = np.load(f)
c = np.array([1786.03738998, 2497.91283978, 2450.9590043, 1065.90899866,
324.45631275, 3588.97920863])
d = 15.620850726403056
m, n = A.shape
x = cp.Variable(n, nonneg=True)
obj_func = c @ x
constraints = [cp.log_sum_exp(A @ x + b) <= d]
prob = cp.Problem(cp.Minimize(obj_func), constraints)
prob.solve(solver=cp.ECOS, verbose=True)
Error message
ECOS 2.0.7 - (C) embotech GmbH, Zurich Switzerland, 2012-15. Web: www.embotech.com/ECOS
It pcost dcost gap pres dres k/t mu step sigma IR | BT
0 +0.000e+00 +1.550e+04 +1e+04 9e-01 1e+00 1e+00 1e+00 --- --- 0 0 - | - -
1 -1.122e+03 +1.436e+04 +9e+03 9e-01 1e+00 1e+00 7e-01 0.4498 4e-01 1 1 1 | 0 0
2 -3.114e+03 +1.234e+04 +9e+03 9e-01 1e+00 1e+00 7e-01 0.0066 1e+00 1 1 3 | 0 0
3 -5.784e+03 +9.575e+03 +9e+03 1e+00 1e+00 1e+00 7e-01 0.1019 7e-01 1 1 1 | 0 0
4 -6.815e+03 +8.465e+03 +9e+03 1e+00 1e+00 1e+00 7e-01 0.0691 7e-01 1 1 1 | 0 0
5 -7.309e+03 +7.821e+03 +9e+03 1e+00 1e+00 1e+00 7e-01 0.0681 7e-01 1 1 1 | 0 0
6 -3.765e+03 +7.239e+03 +7e+03 7e-01 6e-01 1e+00 5e-01 0.9304 7e-01 1 1 1 | 0 0
7 -1.476e+03 +2.683e+03 +2e+03 3e-01 2e-01 5e-01 2e-01 0.7833 7e-02 1 1 1 | 0 1
8 -6.521e+02 +1.144e+03 +8e+02 1e-01 9e-02 2e-01 7e-02 0.6266 3e-02 1 1 1 | 1 2
9 -1.577e+02 +2.797e+02 +2e+02 2e-02 2e-02 6e-02 2e-02 0.7833 2e-02 1 1 1 | 1 1
10 -6.850e+01 +1.135e+02 +8e+01 1e-02 9e-03 3e-02 7e-03 0.6266 6e-02 1 1 1 | 2 2
11 -1.547e+01 +2.679e+01 +2e+01 2e-03 2e-03 7e-03 1e-03 0.7833 1e-02 1 1 1 | 1 1
12 -6.676e+00 +1.131e+01 +7e+00 1e-03 9e-04 3e-03 6e-04 0.6266 5e-02 1 1 1 | 2 2
13 -4.070e+00 +3.127e+00 +2e+00 4e-04 4e-04 2e-03 2e-04 0.9791 3e-01 1 1 1 | 5 0
14 -1.650e+00 +1.821e+00 +9e-01 2e-04 2e-04 1e-03 7e-05 0.6266 7e-02 1 1 2 | 2 2
15 -3.500e-01 +1.368e+00 +3e-01 1e-04 1e-04 6e-04 2e-05 0.9548 3e-01 1 2 2 | 5 0
16 +3.048e-01 +1.243e+00 +1e-01 7e-05 6e-05 4e-04 8e-06 0.6266 1e-01 1 2 2 | 2 2
17 +9.596e-01 +1.620e+00 +5e-02 5e-05 4e-05 3e-04 4e-06 0.6266 2e-01 2 2 2 | 3 2
18 +2.901e+00 +3.475e+00 +2e-02 4e-05 4e-05 4e-04 1e-06 0.9791 4e-01 2 3 2 | 5 0
19 +9.882e+01 +9.965e+01 +5e-04 1e-04 6e-05 2e-02 4e-08 0.9716 1e-04 2 3 3 | 0 0
20 +1.003e+03 +1.012e+03 +1e-04 1e-03 5e-04 1e+00 1e-08 0.6726 4e-02 3 3 3 | 0 0
21 +3.296e+02 +3.331e+02 +6e-05 2e-04 1e-04 2e+00 4e-09 0.9791 4e-01 4 3 3 | 0 0
22 +9.174e+02 +9.263e+02 +2e-05 2e-04 1e-04 7e+00 1e-09 0.7242 3e-02 3 3 3 | 0 0
23 +1.155e+03 +1.163e+03 +2e-05 2e-04 1e-04 7e+00 1e-09 0.0525 7e-01 4 3 3 | 1 12
24 +1.233e+03 +1.241e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0309 9e-01 4 3 3 | 9 14
25 +1.253e+03 +1.261e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0101 1e+00 4 3 3 | 16 19
26 +1.256e+03 +1.264e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0021 1e+00 4 3 3 | 24 26
27 +1.257e+03 +1.265e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0004 1e+00 4 3 3 | 31 33
28 +1.257e+03 +1.265e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0001 1e+00 4 3 3 | 40 42
29 +1.257e+03 +1.265e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0000 1e+00 4 3 3 | 48 50
30 +1.257e+03 +1.265e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0000 1e+00 4 3 3 | 56 57
31 +1.257e+03 +1.265e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0000 1e+00 3 3 3 | 74 75
32 +1.257e+03 +1.265e+03 +2e-05 2e-04 1e-04 6e+00 1e-09 0.0000 1e+00 4 3 3 | 85 87
Combined backtracking failed 0 0 86 4 sigma 1
Combined line search failed, recovering best iterate (17) and stopping.
NUMERICAL PROBLEMS (reached feastol=4.8e-05, reltol=2.9e-02, abstol=4.7e-02).
Runtime: 0.300036 seconds.
rileyjmurray commented
Take a look at my comment on #177.