Jacqueline He, Mengzhou Xia, Christiane Fellbaum, Danqi Chen. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022.

Disaster monitoring is a primary task for wireless sensor networks. Systems for the rapid reporting of earthquake information are a crucial aspect of disaster monitoring. Furthermore, during emergency rescue after a large earthquake, wireless sensor networks can provide pictures and sound information to save lives. Therefore, when accompanied by multimedia data flow, the alert and seismic data sent by the seismic monitoring nodes must be sufficiently fast. We present herein the architecture of a collaborative disaster-monitoring system that can obtain seismic data in a highly energy-efficient manner. In this paper, a hybrid superior node token ring MAC scheme is proposed for disaster monitoring in wireless sensor networks. This scheme consists of set-up and steady-state stages. A clustering approach was proposed for heterogeneous networks during the set-up stage. The proposed MAC operates in the duty cycle mode at the steady-state stage and is based on the virtual token ring of ordinary nodes, the polling all the superior nodes in one period, and alert transmissions with a low-power listening and shortened preamble approach during the sleep state. The proposed scheme can simultaneously satisfy the requirements of three types of data in disaster-monitoring applications. Based on embedded Markov chains, a model of the proposed MAC was developed and the mean queue length, mean cycle time, and mean upper bound of the frame delay were obtained. Using simulations under various conditions, the clustering approach performed better than the pLEACH approach, and the theoretical results of the proposed MAC were verified. We found that alerts and superior data have outstanding delay and throughput performances even under heavy traffic intensity, and the proposed MAC can provide a data rate of several hundred kb/s for superior and ordinary data. Considering all three types of data, the frame delay performances of the proposed MAC are better than those of the WirelessHART and DRX schemes, and the alert data of the proposed MAC have a maximum frame delay of 15 ms. These satisfy the application requirements of disaster monitoring.

Communication and information technologies play crucial roles in the smart grid system. Wireless communications offer many unique features to utilities. Real-time capability and high efficiency under heavy load are vital capabilities for smart grid wireless communications. However, the existing medium access control (MAC) protocols for low-rate and short-range wireless transfer in the literature mainly aim to achieve the objectives of low energy consumption and self-configuration and rarely address these requirements under heavy traffic intensity. This paper presents a real-time, efficient, and lightweight MAC (RE-MAC) protocol to support smart grid applications, based on priority node polling and a hybrid scheme. The upper bounds of packet delay are determined using the embedded Markov chain method, and the simulation results demonstrate that the protocol can achieve predictable, real-time, and efficient performance.

Wireless sensor networks (WSNs) can provide crucial and real-time information in many crisis response and management scenarios. For example, in volcano monitoring and the rapid reporting of aftershock information, WSNs detect the ground vibration and transmit the information sufficiently fast to help people survive the disaster. However, the processing and transmission of ground vibration data present a heavy burden for the small nodes of WSNs, and a single sensor node cannot provide accurate results for ground vibration detection. In this paper, we present a promising energy-efficient collaborative ground vibration measurement scheme based on efficient compressed sensing method and belief propagation method using the difference in the signal strength to distribute wireless communication load among sensor nodes, improve the detection probability, and extend the lifetime of this system. The performance was studied via theoretical analysis and simulation experiments, and the results indicated that the system could reconstruct the seismic signal even with a low measurement count; other benefits included decreased energy consumption of the sensor nodes and a considerably increased detection probability because of the proposed belief propagation method.

The vast diversity of styles, domains, and quality levels present in language model pre-training corpora is essential in developing general model capabilities, but efficiently learning and deploying the correct behaviors exemplified in each of these heterogeneous data sources is challenging. To address this, we propose a new method, termed MEtadata COnditioning then COoldown (MECO), to incorporate additional learning cues during pre-training. MeCo first provides metadata (e.g., URLs like en.wikipedia.org) alongside the text during training and later uses a cooldown phase with only the standard text, thereby enabling the model to function normally even without metadata. MeCo significantly accelerates pre-training across different model scales (600M to 8B parameters) and training sources (C4, RefinedWeb, and DCLM). For instance, a 1.6B language model trained with MeCo matches the downstream task performance of standard pre-training while using 33% less data. Additionally, MeCo enables us to steer language models by conditioning the inference prompt on either real or fabricated metadata that encodes the desired properties of the output: for example, prepending wikipedia.org to reduce harmful generations or factquizmaster.com (fabricated) to improve common knowledge task performance. We also demonstrate that MeCo is compatible with different types of metadata, such as model-generated topics. MeCo is remarkably simple, adds no computational overhead, and demonstrates promise in producing more capable and steerable language models.1