TEMPO: Regime-Aware Operator Learning with Locally Adapted POD Bases
Abstract
Operator learning approximates the solution map of a parametric partial differential equation from data, enabling fast prediction at new parameter values without repeated simulation. The dominant approach fits a single global low-rank basis, typically a proper orthogonal decomposition (POD), to all solution snapshots. This fails when the snapshots span qualitatively different solution shapes, from smooth diffusive profiles to near-discontinuous shocks. We propose TEMPO (Trajectory Expectation-Maximization Mixture of POD Operators), a two-phase framework for regime-aware operator learning. In the offline phase, TEMPO recovers latent regimes via expectation-maximization, fitting a dedicated linear or nonlinear basis to each discovered regime. In the online phase, a gating network and per-regime branch networks are trained jointly, routing new inputs to the regimes at inference. We test TEMPO on Burgers' flow, Sod shock tubes, Coanda flow, Darcy flow, Navier-Stokes flow and the heterogeneous Helmholtz equation. With 64 total POD modes, TEMPO gives 18.61% test error on Burgers', compared with 19.61% for global POD. It also gives lower errors on Coanda and Helmholtz.
Figures & tables
Appendix figures & tables23 assets
Supplementary material from the paper’s appendix.
Appendix
| Method | Train | Modes | Mean | Time (ms) | |||
| Single-viscosity training (original split) | |||||||
| DeepONet (0.553M) | 0.001 | – | 53.563 | 47.407 | 38.173 | 46.381 | |
| DeepONet (0.553M) | 0.1 | – | 53.451 | 47.110 | 34.918 | 45.160 | |
| DeepONet (0.553M) | 1.0 | – | 54.729 | 48.300 | 28.274 | 43.768 | |
| POD-DeepONet (0.468M) | 0.001 | 27.266 | 29.697 | 63.896 | 40.286 | ||
| POD-DeepONet (0.468M) | 0.01 | 27.438 | 27.377 | 61.436 | 38.750 | ||
| Method | Modes | Error (%) | Time (ms) |
|---|---|---|---|
| TEMPO(POD) (11,866) | |||
| TEMPO(NeuralPOD) (1.207M) | |||
| DeepONet (0.529M) | – | ||
| POD-DeepONet ( ) (0.037M) | |||
| POD-DeepONet (11,952) | |||
| NeuralPOD-DeepONet (1.168M) |
| Method | Modes | Error (%) | Time (ms) |
|---|---|---|---|
| TEMPO(POD) (0.104M) | |||
| TEMPO(NeuralPOD) (0.646M) | |||
| DeepONet (0.528M) | – | ||
| POD-DeepONet ( ) (0.035M) | |||
| POD-DeepONet (0.107M) | |||
| NeuralPOD-DeepONet (0.628M) |
| Method | Modes | Test (%) |
|---|---|---|
| DeepONet (2.197M) | – | 5.697 |
| FNO (1.193M) | – | 0.628 |
| CNO (2.007M) | – | 0.397 |
| POD-DeepONet (1.346M) | 64 | 2.254 |
| NeuralPOD-DeepONet (3.234M) | 64 | 2.280 |
| TEMPO(POD) ( 1.343M) | 2.528 |
| Method | Train Re | Test Re | Infer. (ms) | |||
| Per-Re specialists (single-Re training) | ||||||
| DeepONet [ 4 ] (2.691M) | 100 | 1.000 | 1.002 | 1.003 | 1.003 | 0.381 |
| 1000 | 1.000 | 1.000 | 1.000 | 1.000 | 0.381 | |
| 3600 | 1.001 | 1.000 | 1.000 | 1.000 | 0.381 | |
| 10000 | 1.001 | 1.000 | 1.000 | 1.000 | 0.381 | |
| Method | Modes | Error (%) | Time (ms) |
|---|---|---|---|
| TEMPO(POD) (6.704M) | |||
| TEMPO(NeuralPOD) (8.123M) | |||
| DeepONet (3.140M) | – | ||
| POD-DeepONet ( ) (1.341M) | |||
| POD-DeepONet (6.579M) | |||
| NeuralPOD-DeepONet (7.935M) |
| Task ( ) | Basis | Modes | Best local LS | Prediction | Outside share (%) |
|---|---|---|---|---|---|
| Burgers’ (1200) | POD | 14.7901 | 19.5692 | 54.44 | |
| NeuralPOD | 23.3221 | 22.8662 | 80.65 | ||
| Sod (408) | POD | 3.1633 | 3.2113 | 88.74 | |
| NeuralPOD | 4.1899 | 4.1909 | 89.06 | ||
| Coanda (62) | POD | 0.1396 | 0.8078 | 1.95 | |
| NeuralPOD | 0.6915 | 0.8026 | 53.09 |
| Task | Before | After |
|---|---|---|
| Burgers’ | 26.601 | 21.534 |
| Sod | 5.967 | 4.243 |
| Coanda | 0.698 | 0.680 |
| Navier–Stokes | 9.131 | 8.956 |
| Helmholtz | 17.984 | 5.832 |
| Model | Modes | Standard | Parallel heads | CUDA graph |
|---|---|---|---|---|
| TEMPO(POD) (1.397M) | 0.405 | 0.235 | 0.184 | |
| Global POD (1.346M) | 64 | 0.194 | – | 0.142 |
| TEMPO(NeuralPOD) (2.917M) | 0.534 | – | 0.210 | |
| Global NeuralPOD (6.449M) | 32 | 0.131 | 0.139 | 0.104 |
| Benchmark | Basis | EM iters | Skip | Incr. | Full | |
|---|---|---|---|---|---|---|
| 1D Burgers’ | POD | 3 | 19 | |||
| 1D Burgers’ | NeuralPOD | 3 | 13 |
| Regime 1 | Regime 2 | Regime 3 | |
|---|---|---|---|
| 0.001 | 0.32 | 0.28 | 0.40 |
| 0.1 | 0.32 | 0.26 | 0.42 |
| 1.0 | 0.33 | 0.26 | 0.41 |
| 0.32 | 0.27 | 0.41 |
| Task | TEMPO(NeuralPOD) | TEMPO(POD) |
|---|---|---|
| Burgers’ | ||
| Navier–Stokes |
| Task/model | Known parameter | Estimated parameter |
|---|---|---|
| Burgers’, TEMPO(POD), | 0.32978 | 0.33283 |
| Navier–Stokes, TEMPO(POD) | 0.09648 | 0.09724 |
| Navier–Stokes, TEMPO(NeuralPOD) | 0.08956 | 0.09039 |
PDE Operator Learning
Partial differential equation (PDE) operator learning builds models that map inputs such as coefficients and initial conditions to solution fields.
See all 250 papers