Closed
Description
This is a tracking issue for the accuracy of Lanczos solver. During the development of the spectral embedding interface for cuml, I compared the eigen solver between raft and scipy eigsh
with a normalized Laplacian matrix as input. The result from raft solver is off for some cases. At the end, I have pasted a 16x16 example COO matrix extracted out from spectral embedding.
Related: rapidsai/cuml#3864 rapidsai/cuml#263
std::vector<int32_t> rows{
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1,
1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3,
3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5,
5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7,
7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
8, 9, 9, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10, 10, 10, 11, 11,
11, 11, 11, 11, 11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
13, 13, 13, 13, 13, 13, 13, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
14, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15};
std::vector<int32_t> cols{
13, 11, 10, 9, 5, 3, 2, 1, 14, 15, 6, 12, 4, 7, 8, 0, 3, 9, 7,
14, 8, 0, 11, 2, 1, 12, 0, 1, 8, 3, 9, 11, 2, 1, 0, 12, 8, 2,
11, 9, 3, 7, 5, 14, 8, 13, 0, 15, 12, 6, 4, 4, 7, 0, 6, 10, 15,
14, 5, 10, 7, 5, 8, 13, 0, 12, 14, 4, 15, 6, 10, 5, 1, 12, 4, 8,
6, 14, 15, 0, 7, 13, 11, 10, 9, 3, 2, 1, 15, 7, 14, 6, 4, 12, 0,
8, 1, 0, 12, 8, 2, 11, 3, 9, 8, 0, 7, 6, 15, 14, 5, 10, 1, 0,
12, 8, 9, 3, 2, 11, 11, 9, 7, 3, 2, 14, 15, 6, 8, 13, 0, 4, 12,
14, 15, 8, 0, 6, 12, 4, 13, 13, 12, 10, 1, 8, 7, 4, 0, 5, 6, 15,
14, 13, 12, 10, 5, 8, 7, 0, 4, 14, 6, 15};
std::vector<float> vals{-0.067419983446598053, -0.064282432198524475, -0.07537783682346344, -0.064282432198524475, -0.067419983446598053, -0.067419983446598053, -0.064282432198524475, -0.064282432198524475, -0.10050377994775772, -0.10341754555702209, -0.10341754555702209, -0.10341754555702209, -0.10660035908222198, -0.11396057903766632, -0.093048423528671265, 1, -0.095346257090568542, -0.18181817233562469, -0.080582298338413239, -0.071066908538341522, -0.065795168280601501, -0.064282432198524475, -0.18181817233562469, -0.18181817233562469, 1, -0.073127247393131256, -0.064282432198524475, -0.18181817233562469, -0.065795168280601501, -0.19069251418113708, -0.18181817233562469, -0.18181817233562469, 1, -0.095346257090568542, -0.067419983446598053, -0.076696500182151794, -0.069006554782390594, -0.19069251418113708, -0.19069251418113708, -0.19069251418113708, 1, -0.066815309226512909, -0.079056940972805023, -0.1178511306643486, -0.10910894721746445, -0.15811388194561005, -0.10660035163164139, -0.12126781791448593, -0.12126781791448593, -0.12126781791448593, 1, -0.079056940972805023, -0.08451543003320694, -0.067419983446598053, -0.076696500182151794, -0.22360679507255554, -0.15339300036430359, -0.14907120168209076, 1, -0.085749298334121704, -0.064820379018783569, -0.076696500182151794, -0.10585123300552368, -0.15339300036430359, -0.10341754555702209, -0.11764707416296005, -0.1143324002623558, -0.12126782536506653, -0.11764707416296005, 1, -0.094491124153137207, -0.08451543003320694, -0.080582298338413239, -0.064820379018783569, -0.066815316677093506, -0.11664237827062607, -0.064820379018783569, -0.12598817050457001, -0.12964075803756714, -0.11396057903766632, 1, -0.069006554782390594, -0.065795168280601501, -0.077151678502559662, -0.065795168280601501, -0.069006554782390594, -0.065795168280601501, -0.065795168280601501, -0.10585123300552368, -0.11664237827062607, -0.10286890715360641, -0.10585123300552368, -0.10910895466804504, -0.10585123300552368, -0.093048423528671265, 1, -0.18181817233562469, -0.064282432198524475, -0.073127247393131256, -0.065795168280601501, -0.18181817233562469, -0.18181817233562469, -0.19069251418113708, 1, -0.077151678502559662, -0.07537783682346344, -0.094491124153137207, -0.085749298334121704, -0.085749298334121704, -0.083333335816860199, -0.22360679507255554, 1, -0.18181817233562469, -0.064282432198524475, -0.073127247393131256, -0.065795168280601501, -0.18181817233562469, -0.19069251418113708, -0.18181817233562469, 1, -0.073127247393131256, -0.073127247393131256, -0.064820379018783569, -0.076696500182151794, -0.073127247393131256, -0.057166200131177902, -0.058823537081480026, -0.11764707416296005, -0.10585123300552368, -0.15339300036430359, -0.10341754555702209, -0.12126782536506653, 1, -0.07453560084104538, -0.076696500182151794, -0.069006554782390594, -0.067419983446598053, -0.15339300036430359, -0.15339300036430359, -0.15811388194561005, 1, -0.07453560084104538, -0.057166200131177902, -0.083333335816860199, -0.071066908538341522, -0.10286890715360641, -0.12598817050457001, -0.1178511381149292, -0.10050377994775772, -0.14907120168209076, -0.1143324002623558, -0.1143324002623558, 1, -0.076696500182151794, -0.058823537081480026, -0.085749298334121704, -0.15339300036430359, -0.10585123300552368, -0.12964075803756714, -0.10341754555702209, -0.12126782536506653, -0.1143324002623558, -0.11764707416296005, 1};
Metadata
Metadata
Assignees
Labels
Type
Projects
Status
Done
Activity