The rapid growth of deep learning models and the increasing demand for large-scale datasets have posed unprece dented challenges for data loading and memory management. Existing frameworks (e.g., PyTorch, TensorFlow) often encounter performance bottlenecks when handling large datasets resulting in inefficiencies and excessive memory usage. To address these issues, we propose Lafa, a dynamic metadata loading mechanism optimized for efficient large-scale dataset processing. Lafa introduces the .Lafa format and an adaptive loading strategy with three modes to balance memory usage and loading performance, along with a local shuffle approach that reduces memory overhead and computational complexity while preserving data randomness. Experimental results on GPU (RTX 3090) and Ascend (910A) platforms demonstrate that Lafa significantly improves memory efficiency compared to existing frameworks. Specifically, for every 10 million samples loaded, Lafa reduces additional memory consumption by a factor of 1.33× to 31.34× across various dataset types, relative to the most memory-efficient baseline among PyTorch, TensorFlow, and MindSpore.

Estimating engagement in group interactions is crucial for building socially intelligent systems, such as in human-agent and human-robot interaction. However, precisely modeling the continuous frame-level fluctuations of engagement remains challenging, particularly when considering the complex multi-party signal interactions within groups. Our method employs an encoder that integrates BiLSTM with Transformer to effectively capture both local and global temporal dependencies of multimodal features. Crucially, we explicitly fuse signals from both the target participant and their conversational partners in group to model the holistic group interaction dynamics. Furthermore, we introduce an 8x overlapped optimized sliding window strategy, constructing a ''sliding pipeline'', which significantly enhances the temporal smoothness, continuity, and stability of predictions. In the final regression stage, we replace the traditional multilayer perceptron(MLP) decoder with Kolmogorov-Arnold Network (KAN), leveraging their superior function approximation capability to achieve more accurate engagement predictions. Evaluated on the test sets of NoXi-base, NoXi-addition, MPIIGroupInteraction and NoXi-J datasets from the Multimediate'25 Engagement Challenge, our approach demonstrates significant performance improvements, achieving highly competitive Concordance Correlation Coefficients (CCC) of 0.678 for global, approximately 56.1% higher than the baseline, which shows a significant improvement.

Graphs are able to model interconnected entities in many online services, supporting a wide range of applications on the Web. This raises an important question: How can we train a graph foundational model on multiple source domains and adapt to an unseen target domain? A major obstacle is that graphs from different domains often exhibit divergent characteristics. Some studies leverage large language models to align multiple domains based on textual descriptions associated with the graphs, limiting their applicability to text-attributed graphs. For text-free graphs, a few recent works attempt to align different feature distributions across domains, while generally neglecting structural differences. In this work, we propose a novel Structure Alignment framework for text-free Multi-domain Graph Pre-Training and cross-domain adaptation (SAMGPT). It is designed to learn multi-domain knowledge from graphs originating in multiple source domains, which can then be adapted to address applications in an unseen target domain. Specifically, we introduce a set of structure tokens to harmonize structure-based aggregation across source domains during the pre-training phase. Next, for cross-domain adaptation, we design dual prompts, namely, holistic prompts and specific prompts, which adapt unified multi-domain structural knowledge and fine-grained, domain-specific information, respectively, to a target domain. Finally, we conduct comprehensive experiments on seven public datasets to evaluate and analyze the effectiveness of SAMGPT.

It is important to cooperatively share perception data among vehicles for the safety of autonomous driving systems. However, since vehicles may experience long data transmission delays, especially in case that the number of vehicles participating in cooperative perception is large, the dissemination of large-sized perception data remains a challenging problem. Considering the high transmission cost of perception data in the overlapping perception area of nearby vehicles, we propose a novel fusion-tree-based perception data dissemination scheme in millimeter-wave (mmWave) vehicular networks. Vehicles first aggregate their perception data at a preselected leader along a dynamically constructed fusion tree, and a merge-and- forward policy is adopted to reduce the amount of redundant data. After the aggregation stage ends, the leader broadcasts the merged data back to all vehicles through multi-hop broadcast links. The performance evaluation results show that the proposed scheme yields the best transmission performance among the three comparison schemes in terms of dissemination delay.

In vehicular ad hoc networks (VANETs), quick and reliable multi-hop broadcasting is important for the dissemination of emergency warning messages. By scheduling multiple nodes to transmit messages concurrently and cooperatively, cooperative transmission based broadcast schemes may yield much better broadcast performance than conventional broadcast schemes. However, a cooperative transmission requires multiple relays to achieve strict synchronization on both time and frequency, which may induce high cost for a cooperative transmission process. In this paper, we analyze the cost and benefit of a cooperative transmission for data broadcasting in vehicular networks, and introduce a new metric called the single hop broadcast efficiency (SBE) to evaluate the overall broadcast performance. We propose an efficient, non-deterministic cooperation mechanism to reduce the cooperation cost. The mechanism maximizes the expected broadcast performance by selecting cooperators with the largest expected SBE value for a lead relay, and initiates cooperative broadcasting process when the expected SBE value is larger than that of a single-relay based broadcasting. Based on the non-deterministic mechanism, we propose an efficient, cooperative transmission based opportunistic broadcast (ECTOB) scheme which further utilizes rebroadcast to improve the reliability of the broadcast scheme. Simulation results show that the proposed scheme outperforms the conventional ones.

In VANETs, it is important to support fast and reliable multi-hop broadcast for safety-related applications. The performance of multi-hop broadcast schemes is greatly affected by relay selection strategies. However, the relationship between the relay selection strategies and the expected broadcast performance has not been fully characterized yet. Furthermore, conventional broadcast schemes usually attempt to minimize the waiting time difference between adjacent relay candidates to reduce the waiting time overhead, which makes the relay selection process vulnerable to internal interference, occurring due to retransmissions from previous forwarders and transmissions from redundant relays. In this paper, we jointly take both of the relay selection and the internal interference mitigation into account and propose a fast, reliable, opportunistic multi-hop broadcast scheme, in which we utilize a novel metric called the expected broadcast speed in relay selection and propose a delayed retransmission mechanism to mitigate the adverse effect of retransmissions from previous forwarders and an expected redundancy probability based mechanism to mitigate the adverse effect of redundant relays. The performance evaluation results show that the proposed scheme yields the best broadcast performance among the four schemes in terms of the broadcast coverage ratio and the end-to-end delivery latency.

Network coding and schedule-based flooding trees have been widely deployed to improve the performance of a flooding process in low-duty-cycle wireless sensor networks (WSNs). However, the conventional schemes induce too long flooding delay because there is only one static sink node as a single source. In this paper, we use a mobile sink to travel around the monitoring area. When the mobile sink travels to a position, it serves as a new source. Since multiple sources can simultaneously forward data packets, flooding speed can be accelerated. We investigate the new problems when both a mobile sink and network coding are integrated with a schedule-tree-based flooding scheme. We model the process to deliver multiple encoded packets from the source to a node as a Markov process, and we theoretically investigate how to minimize the total expected flooding delay with a mobile sink by considering all the possible traveling orders of the branches. We propose a heuristic algorithm to obtain an appropriate trajectory by accounting for the effect of traveling delay of the mobile sink. Simulation results show that our proposed flooding scheme with a mobile sink can significantly reduce the flooding delay.

Many applications in vehicular ad hoc networks (VANETs) are based on broadcast to disseminate information among vehicles. In conventional broadcast protocols, only one vehicle is scheduled to rebroadcast a message at a certain time to avoid collisions among vehicles. In this paper, we derive the maximum temporal displacement required by the concurrent transmissions in VANETs and propose a Concurrent Transmission based Broadcast (CTB) protocol. The CTB includes two parts, broadcast in a street and broadcast at intersections. We divide the transmission range in the broadcast direction into segments and schedule the concurrent transmissions of the forwarders in the same segment. In our protocol, even if some selected forwarders fail to receive the message, other forwarders having received the message can still rebroadcast it, which reduces the broadcast delay and increases the broadcast reliability. Simulation results show that our protocol is faster and more reliable, compared to conventional broadcast protocols.