Time series forecasting involves justifying assertions scientifically regarding potential states or predicting future trends of an event based on historical data recorded at various time intervals. The field of time series forecasting, supported by diverse deep learning models, has made significant advancements, rendering it a prominent research area. The broad spectra of available time series datasets serve as valuable resources for conducting extensive studies in time series analysis with varied objectives. However, the complexity and scale of time series data present challenges in constructing reliable prediction models. In this paper, our objectives are to introduce and review methodologies for modeling time series data, outline the commonly used time series forecasting datasets and different evaluation metrics. We delve into the essential architectures for trending an input dataset and offer a comprehensive assessment of the recently developed deep learning prediction models. In general, different models likely serve different design goals. We boldly examine the performance of these models under the same time series input dataset with an identical hardware computing system. The measured performance may reflect the design flexibility among all the ranked models. And through our experiments, the SCINet model performs the best in accuracy with the ETT energy input dataset. The results we obtain could give a glimpse in understanding the model design and performance relationship. Upon concluding the paper, we shall provide further discussion on future deep learning research directions in the realm of time series forecasting.

The Macao Government provides web-based streaming videos for the public to monitor live traffic and road conditions across the city. This allows individuals to review the latest road traffic conditions online before planning their travels. To let road user makes better and faster decisions, it is desirable to design an automated subsystem in an Intelligent Transportation System (ITS). And the subsystem should analyze available live video streams and recommend multiple travel routes, if possible, to each query instantly. In the paper, we propose a real-time road traffic condition evaluation system. Its design is based on a combination of deep learning models (YOLO and BoTSORT), and a modified Non-Maximum Suppression (mNMS) algorithm. The mNMS strategy removes the needs to manually tune the NMS parameters. By deploying YOLO with our mNMS, the object detection efficiency on live videos improves significantly. Together with the BoTSORT method, we can track the moving vehicles, create the corresponding motion trajectories, and identify traffic lanes with high correctness. The generated trajectory then operates as a filtering mechanism in assessing real-time road traffic conditions. By separating the lanes based on observation angles and using a per-lane status score independently, we further enhance the overall system performance. Through thorough experiments on the live videos, our design correctly estimates traffic status with high accuracy and without needing any manual parametric adjustments.

Quality of service (QoS) is commonly measured in terms of signal to interference plus noise ratio (SINR), where multiuser interference is mitigated in order to improve the performance. As opposed to conventional suppression, interference can be exploited constructively to enhance the desired signal. With the aid of channel state information (CSI) at the transmitter and data information, we study symbol-level downlink beamforming problems based on noise robustness and outage probability, respectively, subject to power constraints. We further show that an equivalence relationship between the noise robustness and outage probability symbol-level downlink beamforming problems can be obtained. Finally, we provide an analytic symbol error rate (SER) upper bound of the worst user by solving the outage probability-based problem. Our simulations demonstrate that the proposed techniques provide substantial performance improvements over conventional downlink beamforming techniques.

Harmful algal blooms (HABs) pose a serious hazard to the resilience of global aquatic ecosystems, coastal economies, and public health, underscoring an urgent need for advanced, high-efficiency monitoring and forecasting technologies. This study systematically reviews remote sensing technologies for detection and prediction of HABs, with a particular focus on investigating recent advancements related to deep learning. Traditionally, field sampling approaches require laboratory analysis that are not responsive to sudden bloom events, and are characterized by high costs, time inefficiency with limited spatial coverage. But with the wide-area and continuous multi-spectral and hyper-spectral sensing capability, remote sensing is an indispensable means for algal bloom surveillance. With deep learning technology transcending the limitations of conventional methods, sensed data from sentinel-2 satellites, for example, can be used to build a pixel-level annotated dataset for conducting systemic performance comparisons among different spectral indexing methods and learning models. The results indicate that deep learning methods exhibit superior performance over other index-based approaches across all assessment criteria. Despite persistent challenges in areas such as data labeling and model explainability, the integration of remote sensing and deep learning represents a pivotal direction for future research in algal bloom monitoring and forecasting, offering robust technological support for the smart management of aquatic environment.

Rotated object detection in aerial images presents unique challenges due to the variability in object orientation and aspect ratios. In object detection, the common practice is to use boxes to enclose the spatial locations of discovered objects. Traditionally, these bounding boxes are rectangular in shape with edges parallel to the horizontal or vertical axes. However, with the ever-increasing content complexity of images, it would be desirable to rotate these bounding boxes properly to better fit the sizes and shapes of all semantic object locations in images with random orientations. Lately, through different deep learning model developments, there are noticeable improvements on the detection of rotated objects in images. Certainly, it always faces challenges with the most stringent expectations and accuracy requirements. Hence, the primary objective of this study is to analyze and compare state-of-the-art methodologies for rotated object detection in aerial imagery, and provide a comprehensive survey of the latest research status of it. From four different working principles, we systematically review recent advancements in deep learning architectures, including R-CNN variants, YOLO variants and Transformer variants adapted for rotated bounding boxes. Then we further carry out quantitative performance comparisons among the latest state-of-the-art rotated object detection algorithms using the benchmark dataset DOTA and computing facility. Among them, the RTMDet ^a outperforms all other models ^b . There are numerous important applications based on aerial images, for example, urban management, precision agriculture, environment monitoring, emergency rescue, and disaster relief, etc. This study contributes to advancing the understanding of rotated object detection in aerial imagery by providing insights into effective methodologies and highlighting areas for future research. The findings guide practitioners and researchers in selecting appropriate techniques for applications requiring robust rotated object detection in challenging aerial environments.