CaReTS: New Multi-Task Framework Combines Classification and Regression for Time Series Forecasting
Researchers have developed CaReTS, a multi-task learning framework that uses both classification and regression to improve time series forecasting accuracy and interpretability. The framework uses a dual-stream architecture where one branch predicts trends and another estimates deviations from baseline values, with uncertainty-aware weighting to balance the tasks. The approach outperforms existing state-of-the-art methods on real-world datasets while providing more interpretable predictions.
CaReTS introduces a novel approach to multi-step time series forecasting by combining classification and regression tasks within a unified framework. The system employs a dual-stream architecture: a classification branch learns stepwise trends into the future, while a regression branch estimates deviations from the latest observation, thereby disentangling macro-level trends from micro-level variations. To optimize learning across these interconnected tasks, the researchers designed a multi-task loss function with uncertainty-aware weighting that adaptively balances each task's contribution. The framework supports four variants incorporating mainstream temporal modeling encoders—CNNs, LSTMs, and Transformers—allowing flexibility in implementation. Experimental validation on real-world datasets demonstrates that CaReTS achieves superior forecasting accuracy compared to state-of-the-art algorithms while simultaneously improving trend classification performance.
What's missing
The paper does not specify which real-world datasets were used for evaluation, the magnitude of performance improvements over baselines, computational complexity or runtime comparisons, or limitations of the approach such as performance on non-stationary time series or very long-term forecasting horizons.
What different sources said
- arXiv cs.LGCenter
CaReTS: A Multi-Task Framework Unifying Classification and Regression for Time Series Forecasting
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.