Small language models (SLMs) are gaining attention for their lower computational and memory needs while maintaining strong performance.However, efficiently deploying SLMs on resource-constrained devices remains a significant challenge.Post-training quantization (PTQ) is a widely used compression technique that reduces memory usage and inference computation, yet existing methods face challenges in inefficient bit-width allocation and insufficient fine-grained quantization adjustments, leading to suboptimal performance, particularly at lower bit-widths.To address these challenges, we propose multi-level weight quantization (MLWQ), which facilitates the efficient deployment of SLMs.Our method enables more effective bit-width allocation by jointly considering inter-layer loss and intralayer salience.Furthermore, we propose a finegrained partitioning of intra-layer salience to support the tweaking of quantization parameters within each group.Experimental results indicate that MLWQ achieves competitive performance compared to state-of-the-art methods, providing an effective approach for the efficient deployment of SLMs while maintaining model accuracy.

Facing the trend of 5G/6G communications toward high spectral efficiency and low power consumption, RF receivers urgently need to overcome the triple challenges of high sensitivity, terahertz (THz) coverage, and system integration. Conventional approaches are limited by electronic thermal noise, band correlation and antenna perturbation effects. In this context, sensors based on the Rydberg atom become a promising alternative, with his large electric dipole moment bringing extremely high sensitivity, inherent frequency selectivity, and the potential to build low-power integrated photonic platforms. This paper describes a Rydberg atomic RF sensor using a quantum coherent mechanism to receive and demodulate quadrature AM signals. The core innovations include: (1) a scheme for 4QAM signal splitting and demodulation within the atomic resonance region via precise signal control; and (2) dynamic, real-time control of the demodulation channels (I/Q paths) by exploiting the phase difference between the local oscillator (LO) and signal (SIG) fields. The proposed architecture can support demodulation of quadrature AM signals, demonstrating its great potential as a versatile next-generation communications platform.

Abstract Graph substitution is a key optimization technique used in deep learning frameworks. Traditional search-based methods are one way to address the problem of graph substitution. However, with the ongoing expansion of deep neural networks (DNNs), the exploration of their vast equivalent graph search space becomes increasingly time-consuming. In this paper, we propose two heuristic methods to accelerate the search process in graph substitution, offering a relatively novel direction compared to existing methods. The first method employs a Memory-Augmented heuristic to optimize computation graphs. To further enhance the efficiency of computation graph optimization, the second method uses the simulated annealing method. This method adds computation graphs with degraded performance into the candidate set with a certain probability. The experimental results show that without significant compromise of inference performance, these two methods can find graph substitutions delivering similar DNN computing performance compared to existing searching methods, while the overall searching time can be reduced from hours to seconds. The source code is available at https://github.com/hudevictor/MAS-SAS .

This paper proposes a model-driven deep learning based channel feedback and multi-user precoding scheme for cell-free massive MIMO systems, where the downlink pilot signals, CSI compressor (from received pilots to quantized bits) at user equipments (UEs), CSI reconstruction at BSs, and multi-user precoding are designed. Specifically, based on the proposed Transformer-based auto-encoder, the non-orthogonal downlink pilots from different BSs, the CSI compressor at UEs, and the CSI reconstruction (from bits to CSI matrix) at different BSs are end-to-end trained in a distributed manner. Moreover, by utilizing the angular-domain reciprocity of downlink/uplink channels, the CSI reconstruction at the BSs can be further improved with the aid of uplink CSI, which can be easily obtained at the UEs' initial access stage. Additionally, we propose a model-driven deep unfolding based multi-user precoding by unfolding the conventional zero-forcing algorithm and integrating learnable parameters, which substantially reduces the computational complexity and improves the robustness to imperfect CSI.

The brain–computer interface (BCI) technology improves the communication efficiency between people and Internet of Things (IoT) devices. BCI based on the steady-state visual evoked potential (SSVEP-BCI) is the preferred scheme for controlling devices because of its convenient operation, low training requirement, and high information transmission rate (ITR). Most signal acquisition devices for BCIs are used for medical diagnosis and scientific research and utilize multiple channels and wet electrodes to obtain high-quality signals. However, the practicability, wearability, and cost of the signal acquisition devices for real-life applications need to be considered, resulting in new requirements for the acquisition mode, the number of electrodes, power consumption, and signal processing methods. This article presents a wearable low-power collaborative sensing system based on a time mask window canonical correlation analysis method (TMW-CCA). An 8-array spring dry electrode signal acquisition device based on a flexible circuit board is designed to address the shortcomings of traditional wet electrode acquisition devices, such as high-power consumption, discomfort, and being unsuitable for long-time use. The proposed TMW-CCA method, which uses a dry electrode sensor to evaluate the time domain’s signal quality dynamically, exhibits 12.5% higher steady-state visual evoked potential recognition accuracy and 40% lower average power consumption (only 740 mW) than the benchmark.

Gas serves as one of the most indispensable energy sources for industrial production and household life. In order to improve the gas utilization efficiency, smart gas meters for the Internet of Things (IoT) has been designed to achieve two-way communication and remote control functions. However, the existing smart gas meter does not consider further reduction of power consumption. Thus, we designed an ultralow-power sensing framework for IoT and applied it to the smart gas meter. Based on a thorough analysis of the metering system, we propose an ultralow power system framework and a low-power peripheral management solution to reach ultralow power consumption. Also, we propose a cooperative sensing scheme to achieve stability and accuracy of gas volume detection. Finally, we implement a real smart gas manage system to evaluate our low-power smart gas meter solution. Verified by a comparative test, the proposed gas meter successfully reduces the power consumption by at least 37% compared to the baseline. The experimental results verify the innovative design and confirm that the proposed gas meter features ultralow power consumption, high precision, and high reliability.

In an aerial hybrid massive multiple-input multiple-output (MIMO) and orthogonal frequency division multiplexing (OFDM) system, how to design a spectral-efficient broadband multi-user hybrid beamforming with a limited pilot and feedback overhead is challenging. To this end, by modeling the key transmission modules as an end-to-end (E2E) neural network, this paper proposes a data-driven deep learning (DL)-based unified hybrid beamforming framework for both the time division duplex (TDD) and frequency division duplex (FDD) systems with implicit channel state information (CSI). For TDD systems, the proposed DL-based approach jointly models the uplink pilot combining and downlink hybrid beamforming modules as an E2E neural network. While for FDD systems, we jointly model the downlink pilot transmission, uplink CSI feedback, and downlink hybrid beamforming modules as an E2E neural network. Different from conventional approaches separately processing different modules, the proposed solution simultaneously optimizes all modules with the sum rate as the optimization object. Therefore, by perceiving the inherent property of air-to-ground massive MIMO-OFDM channel samples, the DL-based E2E neural network can establish the mapping function from the channel to the beamformer, so that the explicit channel reconstruction can be avoided with reduced pilot and feedback overhead. Besides, practical low-resolution phase shifters (PSs) introduce the quantization constraint, leading to the intractable gradient backpropagation when training the neural network. To mitigate the performance loss caused by the phase quantization error, we adopt the transfer learning strategy to further fine-tune the E2E neural network based on a pre-trained network that assumes the ideal infinite-resolution PSs. Numerical results show that our DL-based schemes have considerable advantages over state-of-the-art schemes.

The paper describes the experience of 7 students in a postgraduate course in a higher education setting over the course of 13 weeks. The course utilizes a blended learning approach, employing synchronous and asynchronous forms of communication as well as some face-to-face meetings. A mixed method approach is adopted to investigate the social dimension of various communication media. The coding of online transcripts and the evaluation of student reflections serve as primary data sources for the elaboration of social presence and its utilization within synchronous and asynchronous communication media. Findings confirm the assumption that the medium plays a role in the overall amount of presence shown. However, students use that potential differently. This may be due in part to students' attitudes toward the media. We would like to encourage future research to target factors influencing the different utilization of a medium's potential.

This paper proposes an improved modeling algorithm based on modified fuzzy Petri net for fuzzy reasoning. The asynchronized property of token was neglected due to simple structures of fuzzy Petri net models for fuzzy reasoning in literatures. The property will cause uncertain results of reasoning for complex problems. The modeling algorithm aims mainly at solving this issue based on fuzzy Petri net whose arc power function, transition function and place capacity are modified. Compared with existing fuzzy Petri nets, the modified fuzzy Petri net can be used to reason according to inherent firing mechanism without additional control mechanism. The algorithm uses arc power and module replacement to deal with reasoning uncertainty and structure conflict. The validity of the algorithm for forward/backward reasoning is proved for acyclic fuzzy rule-based systems. An example is given to illustrate the validity of the proposed algorithm for reasoning.

Rapid technological development in the last decade makes it easier than ever to use technologies as collaborative learning tools. Computer video conferencing as a computer-supported collaborative learning (CSCL) technology provides increasing opportunities for learners to share experiences across time and space. The following describes how multipoint desktop video conferencing (MDVC) is used in preservice teacher education programs in Singapore to enhance the professional development of preservice teachers by allowing them to share ideas, experiences and teaching resources in real time with an audience wider than the schools where they teach.

Video conferencing technology has been in use for teaching, training and communication purposes over the past several years. Recent advances in digital streaming technology, increasing bandwidth and software engineering allow the use of video conferencing in the most efficient and cost effective ways. In the teacher training environment, the use of video conferencing has the potential to improve the discourse between the trainee and supervisor and thus increase the overall effectiveness of the training process. The paper describes a research project which utilizes multipoint desktop video conferencing at the National Institute of Education in Singapore to investigate the feasibility and pedagogic values of the technology in teacher training programs. Some preliminary findings are reported.