Traffic forecasting is essential for intelligent transportation systems. Accurate forecasting relies on continuous observations collected by traffic sensors. However, due to high deployment and maintenance costs, not all regions are equipped with such sensors. This paper aims to forecast for regions without traffic sensors, where the lack of historical traffic observations challenges the generalisability of existing models. We propose a model named **GenCast**, the core idea of which is to exploit external knowledge to compensate for the missing observations and to enhance generalisation. We integrate physics-informed neural networks into GenCast, enabling physical principles to regularise the learning process. We introduce an external signal learning module to explore correlations between traffic states and external signals such as weather conditions, further improving model generalisability. Additionally, we design a spatial grouping module to filter localised features that hinder model generalisability. Extensive experiments show that GenCast consistently reduces forecasting errors on multiple real-world datasets.

Unsupervised anomaly detection has attracted considerable attention in complex data environments due to its independence from costly labeled data. Among various approaches, subspace sampling-based ensemble methods, such as IForest, have been widely adopted for their simplicity and computational efficiency. However, these methods typically operate under single granularity, which limits the diversity of subspaces and hinders the ability to capture hierarchical structures and complex patterns in the data. Moreover, they often overlook uncertainty information such as fuzziness among samples, which constrains their capacity to model complex relationships. To address these limitations, this paper proposes a method called Granular-Ball Subspace-based Fuzzy Neighborhood Anomaly Detector (GSFAD). The proposed method integrates granular-ball subspace ensemble learning with a fuzzy computing framework, achieving a balance between computational efficiency and the ability to model multi-granularity fuzzy structures. Specifically, the algorithm begins by performing multi-granularity aggregation with granular-balls to cover the origin data. Then, regions potentially containing anomalies are filtered out based on granular-ball characteristics before sampling. Building on this, multiple granular-ball subspaces are constructed via repeated sampling, and fuzzy relations between granular-balls are computed within each subspace. Finally, the anomaly score of each sample is assessed by fusing the fuzzy neighborhood information across all subspaces. Experimental results on 20 benchmark datasets demonstrate that GSFAD consistently outperforms existing subspace sampling methods that operate under a single granularity. In addition, it achieves superior performance compared to 15 state-of-the-art anomaly detection techniques. The code is publicly available online athttps://github.com/Caspar-lab/GSFAD.

In information systems lacking decision-making information, effectively leveraging fuzzy rough sets for outlier detection in complex data is challenging, especially in capturing inherent uncertainty and multi-granularity characteristics to construct discriminative outlier scores. However, existing fuzzy rough sets-based outlier detection methods often suffer from three key limitations: (1) Local data distributions are often ignored when calculating fuzzy relation matrices, resulting in inaccurate fuzzy similarity representations; (2) Use of all objects in fuzzy upper and lower approximations can weaken noise resistance and increase computational complexity; (3) Single-granularity data processing reduces efficiency and may fail to capture the multi-granularity nature of data, thereby limiting the adaptability of these methods in complex data environments. To address these issues, we propose to fusesNatural neighbor fuzzy approximations withGranular-ball representation forOutlierDetection (NGOD), which integrates the multi-granularity granular-ball representation and fuzzy rough sets to improve the effectiveness and robustness of unsupervised outlier detection. Specifically, we first define a local distribution-aware fuzzy relation, enabling more discriminative similarity calculations between samples. To improve the effectiveness and robustness of fuzzy upper and lower approximations, we propose a multi-granularity natural neighbor fuzzy approximation model, which effectively utilizes the inherent uncertainty and local abnormal information of data in approximations. Moreover, by introducing natural neighbors, NGOD can adaptively capture local abnormal information in the data without setting neighborhoods manually. Finally, the outlier factors of each sample are calculated in NGOD to measure their outlier degrees. Extensive experiments on diverse datasets demonstrate that NGOD outperforms state-of-the-art methods, validating its superior performance and adaptability. The NGOD code and associated datasets are publicly available athttps://github.com/Mxeron/NGOD.

These recent years have witnessed that convolutional neural network (CNN)-based methods for detecting infrared small targets have achieved outstanding performance. However, these methods typically employ standard convolutions, neglecting to consider the spatial characteristics of the pixel distribution of infrared small targets. Therefore, we propose a novel pinwheel-shaped convolution (PConv) as a replacement for standard convolutions in the lower layers of the backbone network. PConv better aligns with the Gaussian-like spatial distribution of infrared small target, improves feature extraction, significantly expands the receptive field, and introduces only a minimal increase in parameters. Additionally, while recent loss functions combine scale and location losses, they do not adequately account for the varying sensitivity of these losses across different target scales, limiting detection performance on dim-small targets. To overcome this, we propose a scale-based dynamic (SD) Loss that dynamically adjusts the influence of scale and location losses based on target size, improving the network's ability to detect targets of varying scales. We construct a new benchmark, SIRST-UAVB, which is the largest and most challenging dataset to date for real-shot single-frame infrared small target detection. Lastly, by integrating PConv and SD Loss into the latest small target detection algorithms, we achieved significant performance improvements on IRSTD-1K and our SIRST-UAVB dataset, validating the effectiveness and generalizability of our approach.

Traffic forecasting is crucial for transportation systems optimisation. Current models minimise the mean forecasting errors, often favouring periodic events prevalent in the training data, while overlooking critical aperiodic ones like traffic incidents. To address this, we propose DualCast, a dual-branch framework that disentangles traffic signals into intrinsic spatial-temporal patterns and external environmental contexts, including aperiodic events. DualCast also employs a cross-time attention mechanism to capture high-order spatial-temporal relationships from both periodic and aperiodic patterns. DualCast is versatile. We integrate it with recent traffic forecasting models, consistently reducing their forecasting errors by up to 9.6% on multiple real datasets.

Existing density-based outlier detection methods process data at the single-granularity level of individual samples, requiring pairwise distance calculations between all samples and exhibiting high sensitivity to noise. The single-granularity-based processing paradigm fails to mine the information at multiple levels of granularity in data, and most of these methods ignore the potential uncertainty information in data, such as fuzziness, resulting in an inability to effectively detect potential outliers in data. As a novel granular computing method, Granular-Ball Computing (GBC) is characterized by its multi-granularity and robustness, which makes it able to make up for the above drawbacks well. In this study, we propose local Granular-Ball Density-based Outlier (GBDO) detection to improve the performance of the density-based methods. In GBDO, we first identify the $k\text {-}$ similarity Granular-Ball (GB) neighborhoods of each GB via the fuzzy relations among them. Subsequently, the local reachability similarity density of the GBs is calculated through the reachability similarity we defined. Finally, the local GB outlier factors of the samples are calculated based on the local reachability similarity density of the GBs. We adopt a multi-granularity processing paradigm using GBs as the basic units, which reduces computational complexity and improves robustness to noisy data by leveraging the multi-granularity nature of GBs. The experimental results demonstrate the effectiveness of GBDO by comparing it with state-of-the-art methods. The source code and datasets are publicly available at https://github.com/Mxeron/GBDO.

Abstract In this study, we explored the effects that the font size and line spacing of simplified Chinese characters had on their readability on smartphones. One hundred and fifteen participants were recruited to complete Chinese text comprehension tasks and provide user preferences on a 5.9-inch smartphone. Nine test conditions were studied, consisting of three font sizes (10-, 12- and 14-point) and three line spacing variations (1.25-F, 1.5-F and 2-F). The results showed that both font size and line spacing significantly affect reading time, but only font size significantly affects reading accuracy; font size, line spacing and the interaction between them have significant effects on the difficulty of reading and the degree of visual fatigue. Medium and large font sizes are more comfortable to read with large line spacing, while small and medium font sizes are more attractive with large line spacing. The results provide useful information for mobile text interface design.