Predictive Surprise as Self-Grounding Concept Bottleneck for Interpretable Time Series
Organizations: Nanyang Technological University, Singapore · Institute of Advanced Intelligence and Computing (IAIC), A*STAR, Singapore · Department of Computer Science, Khalifa University, UAE · Singapore University of Technology and Design, Singapore
Abstract
Time-series classifiers deployed in risk-sensitive domains must be (1) accurate, (2) interpretable, and (3) able to abstain when evidence is weak. Existing methods do not deliver all 3 at once, and the obstacle is structural, rather than logistical. Meaningful temporal patterns cannot be defined without knowing segment boundaries, yet meaningful boundaries cannot be placed without knowing which patterns to expect. Existing approaches break this circularity by using fixed windows or expert-annotated vocabularies, thereby sacrificing at least one of the 3 properties. We introduce ConceptTime, which closes the loop through a single self-supervised signal. A frozen probabilistic forecaster predicts a distribution over the next observation at every timestep of the sequence. We call the mismatch between its prediction and the realized signal the predictive surprise. This signal both locates segment boundaries and characterizes each resulting segment through the forecaster. Segments are thus grounded by how they behaved relative to expectation, not by what they look like. These summaries cluster into a vocabulary of dynamical regimes that we call concepts. A lightweight head classifies from the concept sequence, and the distance to the nearest concept yields reliability by design. Across 17 public benchmarks, ConceptTime is competitive with state-of-the-art black-box accuracy, beats every interpretable baseline, and retains near-98% accuracy on Epilepsy with 1% of labels. It further remains competitive with softmax, prototype-distance, and SHAP/TimeX/TimeX++ on OOD detection and deletion faithfulness, without a single annotation, language model call, or domain expert. Code is available at: https://github.com/Sachithx/ConceptTime.git.
Figures & tables
| UCI-HAR | Epilepsy | Sleep-EDF | UEA (10 avg) | ||||||||||
| Group | Method | 1% | 5% | Full | 1% | 5% | Full | 1% | 5% | Full | 1% | 5% | Full |
| Black-box | ROCKET | 82.21 | 89.65 | 93.91 | 93.68 | 92.25 | 97.15 | 70.21 | 73.67 | 76.38 | 55.69 | 59.87 | 72.30 |
| MiniRocket | 86.55 | 95.63 | 97.31 | 94.71 | 96.86 | 98.26 | 79.53 | 82.15 | 85.91 | 51.90 | 54.89 | 73.80 | |
| InceptionTime | 86.38 | 91.03 | 95.32 | 95.14 | 97.28 | 98.21 | 77.64 | 81.11 | 84.62 | 48.37 | 50.80 | 67.42 | |
| TS-TCC | 70.23 | 75.82 | 93.72 | 91.20 | 93.20 | 97.12 | 75.26 | 78.98 | 85.67 | 62.37 | 63.87 | 72.71 | |
| TimesNet | 85.33 | 90.05 | 91.82 | 93.74 | 96.33 | 97.57 | 52.36 | 55.62 | 75.88 | 60.01 | 62.28 | 68.76 | |
| Setting | A B | A C | A D | B A | B C | B D | C A | C B | C D | D A | D B | D C | Mean | #1st |
| Panel A. Zero-shot and few-shot transfer | ||||||||||||||
| Direct, 0% | 36.25 | 44.46 | 29.58 | 50.62 | 81.82 | 98.57 | 52.05 | 87.84 | 94.20 | 57.19 | 93.60 | 85.15 | 67.61 | 0 |
| Source- , 5% | 90.14 | 75.84 | 91.06 | 90.40 | 84.02 | 98.53 | 96.66 | 97.43 | 97.62 | 88.60 | 92.00 | 90.74 | 91.09 | 2 |
| Target- , 5% | 95.42 | 91.20 | 97.29 | 93.73 | 90.54 | 98.72 | 96.08 | 98.97 | 98.28 | 87.94 | 95.06 | 89.17 | 94.37 | 7 |
| TSLANet, 5% | 85.89 | 82.04 | 86.36 | 86.73 | 87.87 | 99.56 | 89.11 | 96.30 | 96.00 | 90.65 | 99.85 | 88.78 | 90.76 | 3 |
| Panel B. Density model and vocabulary adaptation with 5% target labels | ||||||||||||||
| Signature components | Vocabulary size ( ) | Boundary signal | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Dataset | Dist. | Ent. | Morph. | Mom. | Full | Surp.f. | 8 | 16 | 32 | 64 | 128 | Ent. | KL | Surp. |
| UCI-HAR | 72.8 | 61.2 | 42.7 | 87.7 | 82.8 | 85.7 | 80.2 | 84.4 | 85.7 | 86.3 | 86.8 | 83.4 | 81.6 | 82.5 |
| Epilepsy | 97.2 | 96.1 | 94.6 | 96.4 | 96.7 | 98.0 | 98.1 | 98.2 | 98.0 | 98.4 | 98.5 | 97.4 | 97.6 | 98.1 |
| Sleep-EDF | 72.3 | 65.4 | 59.0 | 74.4 | 78.3 | 74.8 | 73.7 | 75.0 | 74.0 | 75.4 | 76.2 | 73.8 | 75.0 | 74.5 |
Appendix figures & tables19 assets
Supplementary material from the paper’s appendix.
Appendix
| Property | Post-hoc | CBMs | Adaptive Patch | Prototype / Shapelet | IB / Repr. Learning | ConceptTime (Ours) |
|---|---|---|---|---|---|---|
| Decision routes through concepts | ○ | ● | ○ | ◐ | ○ | ● |
| No external semantic supervision required | ● | ◐ | ● | ● | ● | ● |
| Concepts discovered from data, not predefined | ◐ | ◐ | – | ◐ | ● | ● |
| Adaptive segmentation | ○ | ○ | ● | ○ | ○ | ● |
| Local, segment-level explanations | ◐ | ◐ | ○ | ● | ○ | ● |
| Built-in abstention without auxiliary head | ○ | ◐ | ○ | ○ | ◐ | ● |
| Dataset | Domain / Modality | #Train | #Test | #Ch. | Task / Classes | |
| UCI HAR [ 3 ] | Wearable inertial sensing | 7352 | 2947 | 9 | 128 | 6 activities |
| UCI Epilepsy [ 2 ] | EEG | 9200 | 2300 | 1 | 178 | Seizure / non-seizure |
| Sleep-EDF [ 19 , 12 ] | Sleep EEG | 25612 | 8910 | 1 | 3000 | 5 sleep stages |
| UEA subset [ 6 , 63 ] | Mixed multivariate time series | Varies | Varies | Varies | Varies | 10 classification datasets |
| Fault Diagnosis [ 31 , 49 ] | Bearing fault diagnosis | 8184 | 2728 | 1 | 5120 | 3 classes; 4 transfer domains |
| Dataset | #Ch. | Split | Domain | |
|---|---|---|---|---|
| EthanolConcentration | 3 | 1751 | 261 / 0 / 263 | Alcohol industry |
| FaceDetection | 144 | 62 | 5890 / 0 / 3524 | Face / sensor signals |
| Handwriting | 3 | 152 | 150 / 0 / 850 | Handwriting |
| Heartbeat | 61 | 405 | 204 / 0 / 205 | Heartbeat |
| JapaneseVowels | 12 | 29 | 270 / 0 / 370 | Voice |
| PEMS-SF | 963 | 144 | 267 / 0 / 173 | Transportation |
| Dataset | Labels | TSLANet | ConceptTime -Dynamic |
|---|---|---|---|
| UCI-HAR | Full | 95.41 0.63 | 88.91 3.40 |
| 1% | 86.31 1.42 | 74.60 1.76 | |
| 5% | 90.99 1.08 | 79.60 1.31 | |
| Epilepsy | Full | 97.26 0.51 | 98.39 0.35 |
| 1% | 95.41 1.19 | 97.91 0.96 | |
| 5% | 96.28 0.87 | 98.01 0.71 |
| Unlabeled fraction | 1% labels | 5% labels | Val. NLL |
|---|---|---|---|
| 1% | 95.64 0.16 | 96.74 0.50 | -1.422 |
| 5% | 95.78 1.81 | 97.12 0.78 | -1.631 |
| 10% | 96.46 0.93 | 97.42 0.37 | -1.839 |
| 25% | 97.07 0.60 | 97.22 0.29 | -1.970 |
| 50% | 97.33 0.41 | 97.19 0.25 | -2.223 |
| 100% | 97.36 0.35 | 97.30 0.26 | -2.233 |
| Dataset | Size | Layers | Parameters | Accuracy (%) | |
|---|---|---|---|---|---|
| Epilepsy | Tiny | 2 | 64 | 110,149 | 98.65 0.47 |
| Default | 4 | 128 | 810,629 | 98.39 0.35 | |
| Larger | 6 | 128 | 1,204,357 | 97.22 0.19 | |
| UCI-HAR | Tiny | 2 | 64 | 108,597 | 89.20 3.32 |
| Default | 4 | 128 | 807,413 | 88.91 3.40 | |
| Larger | 6 | 128 | 1,201,141 | 86.94 2.06 |
| UCI-HAR | Epilepsy | |||
| Method | 1% | 5% | 1% | 5% |
| CT-FullU | 65.73 0.45 | 68.86 0.31 | 96.64 0.41 | 96.96 0.24 |
| CT-LabelOnly | 69.16 2.44 | 67.33 2.61 | 96.22 0.46 | 97.36 0.09 |
| CT-DensityFull/GMM-LabelOnly | 64.25 1.27 | 67.14 1.47 | 96.16 0.27 | 97.10 0.37 |
| TS-TCC | 46.85 5.57 | 85.25 1.25 | 81.43 2.60 | 93.25 0.40 |
| TS-TCC-SSL | 72.51 4.46 | 80.55 1.72 | 94.28 0.93 | 95.30 0.17 |
| Protocol | 1% | 5% |
|---|---|---|
| CT-FullU | 74.63 | 79.88 |
| CT-LabelOnly | 75.25 | 76.09 |
| CT-DensityFull/GMM-LabelOnly | 77.13 | 79.91 |
| Source | Pairs | Direct | Source- | Gain | Target- | Gain |
|---|---|---|---|---|---|---|
| FD-A/B/C | 9 | 63.93 26.58 | 91.30 7.52 | +27.37 | 95.58 3.16 | +31.65 |
| FD-D | 3 | 78.65 19.06 | 90.45 1.72 | +11.80 | 90.72 3.81 | +12.08 |
| All | 12 | 67.61 24.98 | 91.09 6.47 | +23.48 | 94.37 3.84 | +26.76 |
| Score | AUROC | AUPR-OOD | FPR95 | Det. acc. |
|---|---|---|---|---|
| Fidelity | 0.666 0.187 | 0.674 0.186 | 0.719 0.343 | 0.605 0.145 |
| Mean density NLL | 0.533 0.291 | 0.583 0.224 | 0.770 0.351 | 0.562 0.116 |
| Max density NLL | 0.513 0.163 | 0.522 0.122 | 0.887 0.170 | 0.509 0.030 |
| 0.528 0.146 | 0.561 0.100 | 0.935 0.109 | 0.546 0.047 | |
| Predictive entropy | 0.517 0.152 | 0.560 0.108 | 0.922 0.122 | 0.546 0.047 |
| Mean predictive variance | 0.522 0.255 | 0.543 0.160 | 0.787 0.312 | 0.505 0.019 |
| Method | Confidence score | AURC | E-AURC |
|---|---|---|---|
| ConceptTime | Fidelity | 0.01060 0.00164 | 0.01032 0.00165 |
| Mean prototype proximity | 0.01062 0.00198 | 0.01034 0.00198 | |
| Max softmax | 0.00144 0.00012 | 0.00117 0.00012 | |
| TOTEM | Mean codebook proximity | 0.01514 0.00284 | 0.01319 0.00289 |
| Max codebook proximity | 0.01597 0.00381 | 0.01401 0.00386 | |
| Max softmax | 0.01547 0.00252 | 0.01351 0.00247 |