The paper proposes a test‑time optimization technique that converts a frozen sequence reconstructor into an imputer for missing time‑series channels. By treating missing entries as optimization variables, clamping observed data, and backpropagating to minimize reconstruction error only on observed coordinates, the method avoids the need for artificial mask augmentation or full‑channel dropout during training. Experiments on five real‑world datasets show that this refinement improves mean absolute error in most dataset‑backbone combinations and can outperform mask‑aware imputers on some datasets, though it requires more inference time and depends on cross‑channel coupling.
Complete sensor-channel failures create a difficult form of time-series missingness: local interpolation is unavailable, and deploying a separately trained imputer for every failure pattern may be imp...
arXiv:2606. 04857v1 Announce Type: new Abstract: Standard IMVC evaluation retrains separate models for different missing-data configurations.
arXiv:2610.20342v1 Announce Type: new Abstract: Missing data is a pervasive challenge in machine learning, often degrading predictive performance and complicating model deployment. Most existing appr...
arXiv:2606. 06328v1 Announce Type: new Abstract: In healthcare, multimodal time series tasks often operate on incomplete observations in practice, for example when ECG segments are lost because electrodes detach or an entire respiratory channel is unavailable during overnight monitoring.
arXiv:2507. 07754v3 Announce Type: replace-cross Abstract: Machine unlearning is usually evaluated by what the classifier outputs: forget-set accuracy, confidence, membership-inference scores.
arXiv:2608. 12592v1 Announce Type: new Abstract: Continuous physiological time series underpin modern clinical monitoring, yet many of the most informative signals are invasive, expensive, or simply unavailable for a given patient.
The paper investigates whether a model that has undergone class unlearning can still recover forgotten classes without access to original data. It introduces a white‑box audit method that generates synthetic probes in representation space, filters them by confidence, and relabels boundary‑adjacent probes as the forgotten class. The authors define a Relearning Score to quantify recovery while preserving retain performance, and demonstrate that several unlearning techniques on CIFAR‑10, CIFAR‑100, and TinyImageNet can be fully recovered in a source‑free setting, sometimes even outperforming a retrained reference.
The paper introduces the concept of protocol divergence, showing that identical nominal missing rates can lead to vastly different learning regimes in incomplete multi‑view clustering. It critiques existing evaluation practices that ignore observation structure and proposes CRAFT, a train‑once framework that fuses observed views with mask‑aware attention, enabling efficient deployment across multiple missing‑view protocols. Experiments on CUB, MultiFashion, and other benchmarks demonstrate CRAFT’s superior performance and significant computational savings through checkpoint reuse.
The paper introduces Curriculum‑Aware Interpolate‑then‑Refine (CAIR), a two‑stage framework for imputing physiological time‑series data. CAIR first learns a coarse base curve with a bidirectional‑GRU interpolator and then refines it through three Transformer passes, trained under a random‑gap curriculum that mimics realistic missingness. Evaluations on continuous glucose monitoring and arterial pressure datasets show CAIR outperforms all baselines across MCAR, MAR, and NMAR mechanisms, especially for long gaps and value‑dependent dropout, while also preserving clinically relevant burden metrics.
CLOSURE introduces a two‑task framework for multivariate time‑series anomaly detection that separately evaluates recoverability and cross‑channel consistency. The recoverability task uses a diffusion reconstructor conditioned on a discrete memory of normal temporal bases, while the consistency task employs a relation‑closure operator that predicts each channel from its peers using sparse signed gates. Experiments demonstrate that combining these two scores outperforms recent reconstruction, frequency, and topology‑based detectors on public benchmarks, highlighting the importance of assessing consistency alongside recoverability.
Continuous physiological time series underpin modern clinical monitoring, yet many of the most informative signals are invasive, expensive, or simply unavailable for a given patient. Conditional generation offers a remedy: an absent signal can be synthesized from co-recorded signals and routine clinical variables.
The paper introduces a new evaluation protocol for deep multivariate imputation models on wearable device data, addressing the issue of structured missingness where sensor features drop out together. Using a Garmin smartwatch dataset from an epilepsy patient, the authors generate realistic block-missing patterns from training data and show that matching the training protocol to this distribution reduces BRITS’ mean absolute error by 43%. They also extend BRITS with time‑of‑day encoding and compare it to linear interpolation and SAITS, finding that no single model dominates and that model rankings vary with evaluation design.