Rip currents cause over 100 drowning deaths and more than 30,000 rescues annually in the United States, posing a severe threat to beach safety worldwide. However, most existing detection methods are reactive, identifying rip currents only after they form, leaving limited time for intervention. We propose RipAlert, a future-frame-aware framework that forecasts near-future coastal dynamics and proactively identifies rip current risks. We design a region-sensitive optical flow prediction method with a novel entropy-based object detector to capture early-stage reverse-flow anomalies. Unlike static-image approaches, RipAlert leverages temporal motion patterns to detect rip currents up to 5 seconds before they visibly form. To support real-world deployment, we design a lightweight mobile application and release a curated dataset with over 2,000 annotated images. Experiments on the RipVIS benchmark show that our approach achieves state-of-the-art performance. The system has been deployed at high-risk beaches in China, issuing successful early warnings over real-world events. Our work advances AI-driven coastal safety and contributes to SDG 3 (Good Health and Well-Being) and SDG 13 (Climate Action).

Multivariate time series (MTS) forecasting presents significant challenges due to the diverse noise distributions and complex periodic patterns across different channels. Existing Transformer-based models often apply uniform noise reduction techniques and simplistic patch segmentation, resulting in suboptimal performance in capturing fine-grained periodic dependencies. In this paper, we propose PPDformer, which independently denoises each channel’s data and identifies key periodic components using Short Time Fourier Transform (STFT). Additionally, we present a novel period-based patch segmentation strategy with period clustering, which transforms 1D time series data into 2D patches based on the identified periodicity. Furthermore, we design a dual attention mechanism for local and global information aggregation. Extensive experiments on public datasets demonstrate that PPDformer achieves state-of-the-art forecasting accuracy, particularly in scenarios with complex periodicity and noise. Code is available at https://github.com/damonwan1/PPDformer.

Deep-learning-based lossless compression is of immense importance in real-world applications, such as cold data persistence, sensor data collection, and astronomical data transmission. However, existing compressors typically model data using single-byte symbols as tokens, which makes it hard to capture the inherent correlations and cannot effectively utilize the parallel capabilities of GPU and multi-core CPU. This paper proposes SEP, a novel lossless compression framework for most time-series backbone neural networks. We first introduce a semantic enhancement module to capture the complex intra-patch relationships of binary byte streams. To improve the compression speed, we design multi-stream pipelines that dynamically assign parallel tasks to GPU streams and multi-cores. We further propose a novel GPU memory optimization strategy, which reuses GPU memory by a shared pool across streams. We conduct experiments on seven real-world datasets and the results demonstrate that our SEP framework outperforms state-of-the-art compressors with an average speed improvement of 30.0% and an average compression ratio gain of 5.1%, which is further elevated to 7.6% with the use of pre-training models. The GPU memory footprint is reduced by as high as 63.1% and by an average of 36.2%. The source code is available at: https://github.com/damonwan1/SEP.

Over 4.15 million low-income households across nearly 60,000 villages in China benefit from photovoltaic (PV) poverty alleviation power stations. However, weak infrastructure and limited capabilities make these systems vulnerable to fluctuations. One of the United Nations' Sustainable Development Goals (SDG 7) seeks to ensure access to affordable and reliable energy for all, especially in underdeveloped regions. This paper proposes MCloudNet, a multi-modal framework designed to improve ultra-short-term PV prediction in data-scarce, cloud-dynamic environments. MCloudNet explicitly models multi-layer cloud structures from satellite imagery and fuses them with time-series meteorological data to enhance prediction accuracy and interpretability. A province-level dispatch system with MCloudNet has been deployed in Hebei, supporting scheduling across rural PV stations. Experiments conducted in counties such as Shexian and Luxi highlight the framework's effectiveness for use in underdeveloped micro-grids. Operational results show that the system has reduced over 60 million kWh of solar curtailment and generated 24 million CNY in economic value, benefiting approximately 50,000 rural households. By minimizing power fluctuations and improving rural energy scheduling, MCloudNet supports essential services such as lighting, medical facilities, and communications. The source code is available at: https://github.com/AI4SClab/MCloudNet.

Adaptive gradient methods, primarily based on Adam, are prevalent in training neural networks, adjusting step sizes via exponentially decaying averages of gradients and squared gradients. Adam assigns small weights to distant gradients, termed long-tail gradients in this paper. However, these gradients persistently influence update behavior, potentially degrading generalization performance. To address this issue, we incorporate a restart mechanism into moment estimations, proposing AdaR (ADAptive gradient methods via Restarting moment estimations). Specifically, AdaR divides a training epoch into fixed-iteration intervals, alternating between two sets of moment estimations for parameter updates and discarding prior moment estimations at the beginning of each interval. Within each interval, one set updates parameters and will be discarded in the subsequent interval, while the other is reset at the midpoint to estimate moments for updates in the subsequent interval. The restart mechanism cyclically discards distant gradients, initiates fresh moment estimations for parameter updates, and stabilizes training. By prioritizing recent gradients, the method increases estimation accuracy and enhances step size adjustment. Empirically, AdaR outperforms state-of-the-art optimization algorithms on image classification and language modeling tasks, demonstrating superior generalization and faster convergence.

In Audio-Visual Event Localization (AVEL) task, various cross-modal attentions (CMA) were proposed to capture the bilateral correlations of audio and visual segments. However, existing CMA approaches are inefficient since they require two sets of independent attention parameters. Besides, existing works often ignore the semantic alignment between audio and audible objects, leading to the suboptimal localization results. In this letter, a novel network with a cross co-attention (CCA) and a dynamic audio-object semantic alignment (DAOSA) strategy is proposed to tackle these issues. Unlike existing CMA methods, CCA calculates the co-attention between audio and visual segments to capture the bilateral correlations via a group of parameters. To align the semantics of audio and audio-related objects, DAOSA proposes a dynamic threshold scheme to adaptively select the highly relevant audio-object pairs as positivity while regarding other pairs as negativity. Then, DAOSA optimizes the semantic alignment of positive pairs by contrastive learning. Experiments across different datasets demonstrate the effectiveness of proposed method, which also outperforms several state-of-the-art models.

In previous fine-grained image recognition (FGIR) methods, the single global or local semantic fusion view may not be comprehensive to reveal the semantic associations between image and text. Besides, the encoding fusion strategy cannot fuse the semantics finely because the low-order text semantic dependence and the irrelevant semantic concepts are fused. To address these issues, a novel Dual-Semantic Decoding Fusion Networks (2S-DFN) is proposed for FGIR. Specifically, a multilayer text semantic encoder is first constructed to extract the higher-order semantics dependence among text. To obtain sufficient semantic association, two decoding semantic fusion streams are symmetrically designed from the global and local perspectives. Moreover, by decoding way to implant text features to semantic fusion layer as well as cascading it deeply, two streams fuse the semantics of text and image finely. Extensive experiments demonstrate that the effectiveness of the proposed method and 2S-DFN attains the state-of-the-art results on two benchmark datasets.

The Method Of Characteristic (MOC) to solve the Neutron Transport Equation (NTE) is the core of full-core simulation for reactors. High resolution is enabled by discretizing the NTE through massive tracks to traverse the 3D reactor geometry. However, the 3D full-core simulation is prohibitively expensive because of the high memory consumption and the severe load imbalance. To deal with these challenges, we develop ANT-MOC1. Specifically, we build a performance model for memory footprint, computation and communication, based on which a track management strategy is proposed to overcome the resolution bottlenecks caused by limited GPU memory. Furthermore, we implement a novel multi-level load mapping strategy to ensure load balancing among nodes, GPUs, and CUs. ANT-MOC enables a 3D full-core reactor simulation with 100 billion tracks on 16,000 GPUs, with 70.69% and 89.38% parallel efficiency for strong scalability and weak scalability, respectively.

Summary Anomaly detection tries to find out the data that disobeys the rule of majority data or expected patterns. The traditional hierarchical clustering algorithms have been adopted to detect anomaly, but have the disadvantages of low effectiveness and unstability. So we propose an improved agglomerative hierarchical clustering method for anomaly detection. It dynamically adjusts the optimum clustering number according to the self‐defined criterion to save the trouble of manually picking clustering number, and determines the optimum clustering distance mode according to cophenetic correlation coefficient to reduce the procedure of manually testing the suitable distance mode in each iteration. The performances of proposed method are verified on tensile test, HTRU2 and credit card dataset. Compared with the traditional methods, our method possesses the most comprehensive performance (the highest F‐measure with less iterations), which shows effectiveness of anomaly detection. And compared with the traditional methods (single, complete, average, and centroid mode), our method achieves the best performance on tensile test and HTRU2 dataset, showing stronger generalization. Compared with other methods (Decision + Gradient Boosted Tree, Decision Trees + Decision Stump, etc) on credit card dataset, our method obtains similar accuracy, and ranks in the top level in the aspect of sensitivity.