Junru Luo

Long-horizon spatio-temporal forecasting under strong seasonal distribution drift, mixed temporal periodicity, and time-varying inter-node dependency remains a challenging problem, with ozone concentration prediction serv ing as a representative and demanding instance. We propose TriAtt-STNet, a channel–temporal–spatial tri-attention network with seasonality-aware em bedding that targets this general class of problems and is instantiated here for long-horizon ozone forecasting. The framework integrates hybrid-statistics reversible normalization to stabilize distribution drift by fusing instance-level and climatological priors, seasonality-aware embedding to preserve temporal context and spatial identity, and stacked spatio-temporal attention blocks to jointly capture dominant periodic structures and dynamic inter-node cou pling. Experiments on gridded ozone data over East China show that TriAtt STNet achieves the best overall performance among the compared baselines, while cross-horizon, ablation, and visualization analyses further demonstrate its robustness and interpretability. Beyond ozone, the design is broadly ap plicable to spatio-temporal forecasting tasks that exhibit seasonal drift and multi-scale periodicity, such as air-quality, traffic, hydrological, and energy demand prediction, providing both an effective data-driven basis for envi ronmental decision support and a general modeling perspective for the time series and spatio-temporal learning community.

In contextual multi-armed bandits, the relationship between contextual information and rewards is typically unknown, complicating the trade-off between exploration and exploitation. A common approach to address this challenge is the Upper Confidence Bound (UCB) method, which constructs confidence intervals to guide exploration. However, the UCB method becomes computationally expensive in environments with numerous arms and dynamic contexts. This paper presents an adaptive noise exploration framework to reduce computational complexity and introduces two novel algorithms: EAD (Exploring Adaptive Noise in Decision-Making Processes) and EAP (Exploring Adaptive Noise in Parameter Spaces). EAD injects adaptive noise into the reward signals based on arm selection frequency, while EAP adds adaptive noise to the hidden layer of the neural network for more stable exploration. Experimental results on recommendation and classification tasks show that both algorithms significantly surpass traditional linear and neural methods in computational efficiency and overall performance.

This article provides a new framework to evaluate the status of Renminbi internationalization. It proposes that the trading patterns of a currency in the global foreign exchange market embody the currency's position in the international monetary system. Based on foreign exchange trading data provided by CLS Group, the article constructs a ranking of major international currencies including the Renminbi. It finds that the Renminbi shares more similarities in foreign exchange trading patterns with the established global currencies like the US dollar and the Euro than with those regional currencies. The article also explores the policy implications that the new evaluation approach provides.