arXiv:2512.03298v1 Announce Type: new 
Abstract: Regime transitions routinely break stationarity in time series, making calibrated uncertainty as important as point accuracy. We study distribution-free uncertainty for regime-switching forecasting by coupling Deep Switching State Space Models with Adaptive Conformal Inference (ACI) and its aggregated variant (AgACI). We also introduce a unified conformal wrapper that sits atop strong sequence baselines including S4, MC-Dropout GRU, sparse Gaussian processes, and a change-point local model to produce online predictive bands with finite-sample marginal guarantees under nonstationarity and model misspecification. Across synthetic and real datasets, conformalized forecasters achieve near-nominal coverage with competitive accuracy and generally improved band efficiency.

قدمت دراسة جديدة استنتاجًا متوافقًا تكيفيًا (ACI) بالاشتراك مع نماذج الفضاء العميق للتبديل لتحسين توقعات تغيير الأنظمة. تتناول هذه الطريقة التحديات التي تطرحها عدم الثبات في بيانات السلاسل الزمنية، مما يسمح بوجود عدم اليقين المعاير إلى جانب دقة النقاط.

Un nuevo estudio ha presentado la Inferencia Conformal Adaptativa (ACI) combinada con Modelos de Espacio de Estado de Conmutación Profunda para mejorar las predicciones de cambio de régimen. Este enfoque aborda los desafíos que plantea la no estacionariedad en los datos de series temporales, permitiendo una incertidumbre calibrada junto con la precisión puntual.

Une nouvelle étude a introduit l'Inference Conformale Adaptative (ACI) combinée avec des Modèles d'Espace d'État à Commutation Profonde pour améliorer les prévisions de changement de régime. Cette approche répond aux défis posés par la non-stationnarité dans les données de séries temporelles, permettant une incertitude calibrée en plus de la précision des points.

A new study has introduced Adaptive Conformal Inference (ACI) combined with Deep Switching State Space Models to enhance regime-switching forecasting. This approach addresses the challenges posed by nonstationarity in time series data, allowing for calibrated uncertainty alongside point accuracy.

Adaptive Regime-Switching Forecasts with Distribution-Free Uncertainty: Deep Switching State-Space Models Meet Conformal Prediction

arXiv:2508.12409v3 Announce Type: replace 
Abstract: Semi-supervised semantic segmentation (S4) has advanced remote sensing (RS) analysis by leveraging unlabeled data through pseudo-labeling and consistency learning. However, existing S4 studies often rely on small-scale datasets and models, limiting their practical applicability. To address this, we propose S5, the first scalable framework for semi-supervised semantic segmentation in RS, which unlocks the potential of vast unlabeled Earth observation data typically underutilized due to costly pixel-level annotations. Built upon existing large-scale RS datasets, S5 introduces a data selection strategy that integrates entropy-based filtering and diversity expansion, resulting in the RS4P-1M dataset. Using this dataset, we systematically scale up S4 into a new pretraining paradigm, S4 pre-training (S4P), to pretrain RS foundation models (RSFMs) of varying sizes on this extensive corpus, significantly boosting their performance on land cover segmentation and object detection tasks. Furthermore, during fine-tuning, we incorporate a Mixture-of-Experts (MoE)-based multi-dataset fine-tuning approach, which enables efficient adaptation to multiple RS benchmarks with fewer parameters. This approach improves the generalization and versatility of RSFMs across diverse RS benchmarks. The resulting RSFMs achieve state-of-the-art performance across all benchmarks, underscoring the viability of scaling semi-supervised learning for RS applications. All datasets, code, and models will be released at https://github.com/MiliLab/S5

تم تقديم إطار عمل جديد يسمى S5 للتمييز الدلالي شبه المراقب القابل للتوسع في الاستشعار عن بعد، مما يعزز تحليل بيانات مراقبة الأرض من خلال الاستفادة من كميات كبيرة من البيانات غير المعلّمة باستخدام تقنيات مبتكرة مثل التسمية الزائفة وتعلم التناسق. يعتمد هذا الإطار على مجموعات بيانات كبيرة الحجم الموجودة ويقدم مجموعة بيانات RS4P-1M، التي تستخدم استراتيجية اختيار البيانات لتحسين أداء النموذج.

Se ha introducido un nuevo marco llamado S5 para la segmentación semántica semi-supervisada escalable en la teledetección, mejorando el análisis de datos de observación de la Tierra al utilizar grandes cantidades de datos no etiquetados mediante técnicas innovadoras como el etiquetado pseudo y el aprendizaje de consistencia. Este marco se basa en conjuntos de datos a gran escala existentes e introduce el conjunto de datos RS4P-1M, que emplea una estrategia de selección de datos para mejorar el rendimiento del modelo.

Un nouveau cadre nommé S5 a été introduit pour la segmentation sémantique semi-supervisée évolutive dans la télédétection, améliorant l'analyse des données d'observation de la Terre en utilisant de vastes quantités de données non étiquetées grâce à des techniques innovantes telles que le pseudo-étiquetage et l'apprentissage de la cohérence. Ce cadre s'appuie sur des ensembles de données à grande échelle existants et introduit l'ensemble de données RS4P-1M, qui utilise une stratégie de sélection de données pour améliorer les performances des modèles.

A new framework named S5 has been introduced for scalable semi-supervised semantic segmentation in remote sensing, enhancing the analysis of Earth observation data by utilizing vast amounts of unlabeled data through innovative techniques like pseudo-labeling and consistency learning. This framework builds upon existing large-scale datasets and introduces the RS4P-1M dataset, which employs a data selection strategy for improved model performance.

Adaptive Regime-Switching Forecasts with Distribution-Free Uncertainty: Deep Switching State-Space Models Meet Conformal Prediction

Was this article worth reading? Share it

LucidQuery AI

Portfolio Backtest

Https