Quantization emerges as one of the most promising compression technologies for deploying efficient large models for various real time application in recent years. Considering that the storage and IO of weights take up the vast majority of the overhead inside a large model, weight only quantization can lead to large gains. However, existing quantization schemes suffer from significant accuracy degradation at very low bits, or require some additional computational overhead when deployed, making it difficult to be applied to large-scale applications in industry. In this paper, we propose decoupleQ, achieving a substantial increase in model accuracy, especially at very low bits. decoupleQ abandons the traditional heuristic quantization paradigm and decouples the model parameters into integer and floating-point parts, thus transforming the quantization problem into a traditional mathematical optimization problem with constraints, which is then solved alternatively by off-the-shelf optimization methods. Quantization via decoupleQ is linear and uniform, making it hardware-friendlier than non-uniform counterpart, and enabling the idea to be migrated to high-bit quantization to enhance its robustness. Our method has achieved well on-line accuracy near fp16/bf16 on the 2-bit quantization of large speech models in ByteDance. The code is available at https://github.com/bytedance/decoupleQ
Multi-band radiomap reconstruction (MB-RMR) is a key component in wireless communications for tasks such as spectrum management and network planning. However, traditional machine-learning-based MB-RMR methods, which rely heavily on simulated data or complete structured ground truth, face significant deployment challenges. These challenges stem from the differences between simulated and actual data, as well as the scarcity of real-world measurements. To address these challenges, our study presents RadioGAT, a novel framework based on Graph Attention Network (GAT) tailored for MB-RMR within a single area, eliminating the need for multi-region datasets. RadioGAT innovatively merges model-based spatial-spectral correlation encoding with data-driven radiomap generalization, thus minimizing the reliance on extensive data sources. The framework begins by transforming sparse multi-band data into a graph structure through an innovative encoding strategy that leverages radio propagation models to capture the spatial-spectral correlation inherent in the data. This graph-based representation not only simplifies data handling but also enables tailored label sampling during training, significantly enhancing the framework's adaptability for deployment. Subsequently, The GAT is employed to generalize the radiomap information across various frequency bands. Extensive experiments using raytracing datasets based on real-world environments have demonstrated RadioGAT's enhanced accuracy in supervised learning settings and its robustness in semi-supervised scenarios. These results underscore RadioGAT's effectiveness and practicality for MB-RMR in environments with limited data availability.
Despite the development of various deep learning methods for Wi-Fi sensing, package loss often results in noncontinuous estimation of the Channel State Information (CSI), which negatively impacts the performance of the learning models. To overcome this challenge, we propose a deep learning model based on Bidirectional Encoder Representations from Transformers (BERT) for CSI recovery, named CSI-BERT. CSI-BERT can be trained in an self-supervised manner on the target dataset without the need for additional data. Furthermore, unlike traditional interpolation methods that focus on one subcarrier at a time, CSI-BERT captures the sequential relationships across different subcarriers. Experimental results demonstrate that CSI-BERT achieves lower error rates and faster speed compared to traditional interpolation methods, even when facing with high loss rates. Moreover, by harnessing the recovered CSI obtained from CSI-BERT, other deep learning models like Residual Network and Recurrent Neural Network can achieve an average increase in accuracy of approximately 15\% in Wi-Fi sensing tasks. The collected dataset WiGesture and code for our model are publicly available at https://github.com/RS2002/CSI-BERT.
The LoRA-finetuning quantization of LLMs has been extensively studied to obtain accurate yet compact LLMs for deployment on resource-constrained hardware. However, existing methods cause the quantized LLM to severely degrade and even fail to benefit from the finetuning of LoRA. This paper proposes a novel IR-QLoRA for pushing quantized LLMs with LoRA to be highly accurate through information retention. The proposed IR-QLoRA mainly relies on two technologies derived from the perspective of unified information: (1) statistics-based Information Calibration Quantization allows the quantized parameters of LLM to retain original information accurately; (2) finetuning-based Information Elastic Connection makes LoRA utilizes elastic representation transformation with diverse information. Comprehensive experiments show that IR-QLoRA can significantly improve accuracy across LLaMA and LLaMA2 families under 2-4 bit-widths, e.g., 4- bit LLaMA-7B achieves 1.4% improvement on MMLU compared with the state-of-the-art methods. The significant performance gain requires only a tiny 0.31% additional time consumption, revealing the satisfactory efficiency of our IRQLoRA. We highlight that IR-QLoRA enjoys excellent versatility, compatible with various frameworks (e.g., NormalFloat and Integer quantization) and brings general accuracy gains. The code is available at https://github.com/htqin/ir-qlora.
Knowledge Distillation (KD) emerges as one of the most promising compression technologies to run advanced deep neural networks on resource-limited devices. In order to train a small network (student) under the guidance of a large network (teacher), the intuitive method is regularizing the feature maps or logits of the student using the teacher's information. However, existing methods either over-restrict the student to learn all information from the teacher, which lead to some bad local minimum, or use various fancy and elaborate modules to process and align features, which are complex and lack generality. In this work, we proposed an abstract and general paradigm for the KD task, referred to as DIMensionality Reduction KD (RdimKD), which solely relies on dimensionality reduction, with a very minor modification to naive L2 loss. RdimKD straightforwardly utilizes a projection matrix to project both the teacher's and student's feature maps onto a low-dimensional subspace, which are then optimized during training. RdimKD achieves the goal in the simplest way that not only does the student get valuable information from the teacher, but it also ensures sufficient flexibility to adapt to the student's low-capacity reality. Our extensive empirical findings indicate the effectiveness of RdimKD across various learning tasks and diverse network architectures.
In the sixth-generation (6G) networks, massive low-power devices are expected to sense environment and deliver tremendous data. To enhance the radio resource efficiency, the integrated sensing and communication (ISAC) technique exploits the sensing and communication functionalities of signals, while the simultaneous wireless information and power transfer (SWIPT) techniques utilizes the same signals as the carriers for both information and power delivery. The further combination of ISAC and SWIPT leads to the advanced technology namely integrated sensing, communication, and power transfer (ISCPT). In this paper, a multi-user multiple-input multiple-output (MIMO) ISCPT system is considered, where a base station equipped with multiple antennas transmits messages to multiple information receivers (IRs), transfers power to multiple energy receivers (ERs), and senses a target simultaneously. The sensing target can be regarded as a point or an extended surface. When the locations of IRs and ERs are separated, the MIMO beamforming designs are optimized to improve the sensing performance while meeting the communication and power transfer requirements. The resultant non-convex optimization problems are solved based on a series of techniques including Schur complement transformation and rank reduction. Moreover, when the IRs and ERs are co-located, the power splitting factors are jointly optimized together with the beamformers to balance the performance of communication and power transfer. To better understand the performance of ISCPT, the target positioning problem is further investigated. Simulations are conducted to verify the effectiveness of our proposed designs, which also reveal a performance tradeoff among sensing, communication, and power transfer.
Benefiting from tens of GHz of bandwidth, terahertz (THz) communications has become a promising technology for future 6G networks. However, the conventional hybrid beamforming architecture based on frequency-independent phase-shifters is not able to cope with the beam split effect (BSE) in THz massive multiple-input multiple-output (MIMO) systems. Despite some work introducing the frequency-dependent phase shifts via the time delay network to mitigate the beam splitting in THz wideband communications, the corresponding issue in reconfigurable intelligent surface (RIS)-aided communications has not been well investigated. In this paper, the BSE in THz massive MIMO is quantified by analyzing the array gain loss. A new beamforming architecture has been proposed to mitigate this effect under RIS-aided communications scenarios. Simulations are performed to evaluate the effectiveness of the proposed system architecture in combating the array gain loss.
Integrated sensing and communication (ISAC) capability is envisioned as one key feature for future cellular networks. Classical half-duplex (HD) radar sensing is conducted in a "first-emit-then-listen" manner. One challenge to realize HD ISAC lies in the discrepancy of the two systems' time scheduling for transmitting and receiving. This difficulty can be overcome by full-duplex (FD) transceivers. Besides, ISAC generally has to comprise its communication rate due to realizing sensing functionality. This loss can be compensated by the emerging reconfigurable intelligent surface (RIS) technology. This paper considers the joint design of beamforming, power allocation and signal processing in a FD uplink communication system aided by RIS, which is a highly nonconvex problem. To resolve this challenge, via leveraging the cutting-the-edge majorization-minimization (MM) and penalty-dual-decomposition (PDD) methods, we develop an iterative solution that optimizes all variables via using convex optimization techniques. Besides, by wisely exploiting alternative direction method of multipliers (ADMM) and optimality analysis, we further develop a low complexity solution that updates all variables analytically and runs highly efficiently. Numerical results are provided to verify the effectiveness and efficiency of our proposed algorithms and demonstrate the significant performance boosting by employing RIS in the FD ISAC system.
Edge artificial intelligence (AI) has been a promising solution towards 6G to empower a series of advanced techniques such as digital twin, holographic projection, semantic communications, and auto-driving, for achieving intelligence of everything. The performance of edge AI tasks, including edge learning and edge AI inference, depends on the quality of three highly coupled processes, i.e., sensing for data acquisition, computation for information extraction, and communication for information transmission. However, these three modules need to compete for network resources for enhancing their own quality-of-services. To this end, integrated sensing-communication-computation (ISCC) is of paramount significance for improving resource utilization as well as achieving the customized goals of edge AI tasks. By investigating the interplay among the three modules, this article presents various kinds of ISCC schemes for federated edge learning tasks and edge AI inference tasks in both application and physical layers.