The consideration of channel correlation is crucial for improving the performance of multivariate time series forecasting. However, existing models fail to capture it in homogeneous and highly correlated channels. In this work, we introduce CompRestacking (Compression Restacking), a strikingly intuitive and effective method to address this problem. The approach consists of three main components: (1) PCC-Restacking for correlation-aware channel ordering, (2) Temporal embedding for time encoding, and (3) Aggregation compression for compact token generation. CompRestacking consistently outperforms in experiment results. The results demonstrate that CompRestacking leverages strong channel correlations for improved performance.

In this work, C2R-KD is proposed, applying a Complex-to-Real projection to map complex domain features into the real domain. C2R-KD mitigates complex-real domain mismatch to strengthen the representational capacity of the student model and further improves the knowledge distillation model performance through the hybrid distillation of features and logits simultaneously. Experimental result demonstrates higher accuracy than the conventional KD across all test environments.

Complex-valued data, unlike real-valued data, requires consideration of intricate correlations and patterns. In particular, within learning frameworks based on real-valued computations, conventional input representations for complex numbers may fail to account for the correlations between the real and imaginary components, leading to potential inefficiencies. To address this issue, we propose a novel framework called ComplexRep, aimed at efficiently processing complex-valued data and enhancing its transparency. This framework transforms complex sequence data into a format similar to images, allowing for the consideration of inter-component correlations while improving overall model performance. The ComplexRep framework employs advanced techniques such as Information Addition and the proposed Learned Representation Integration (LRI) to strengthen low model complexity and high Initial Trial Success Probability (ITSP). Additionally, we enhance the reliability of our experiments by utilizing both public datasets and data collected from real-world environments. Extensive evaluation results demonstrate that our framework excels even under low signal-to-noise ratio (SNR) conditions, increasing the overall system efficiency. Notably, compared to previously used input formats, ComplexRep improves ITSP performance and reduces model complexity, thus proving its efficiency. Further experiments across various models confirm the framework’s compatibility with several state-of-the-art models. All experiments include additional tests on real-world 5G data, validating the applicability of the proposed approach. This study presents the potential to effectively manage complex-valued data and maximize performance while offering directions for future complex-valued data processing research.

We applied Risk-averse Reinforcement Learning (RL) to optimize investment portfolios while incorporating risk constraints. Given that portfolios must adhere to risk constraints set by investors and regulators, enforcing hard constraints is essential for practical portfolio optimization. Traditional techniques often lack the flexibility to model the complexities of dynamic financial markets. To address this, we used the Augmented Lagrangian Multiplier (ALM) to impose constraints on the agent, reducing risk during decision-making. Our risk-constrained RL algorithm demonstrated no constraint violations during testing and outperformed other Risk-averse RL methods, indicating its potential for optimizing portfolios for risk-averse investors.

This paper addresses contention window optimization for multi-access scenarios. Our investigation into state-of-the-art models revealed that a limited number of nodes dominate the communication channels. Such monopolization issues are critical in networks as they can lead to significant disruptions. To mitigate this monopolization problem, we propose an imitation learning-based backoff mechanism. The proposed model is a reinforcement learning-based contention window optimization method. It imitates the expert's policy to ensure fair policy convergence for the agent and includes opportunities for weight adjustment to boost performance. The proposed model shows a fairness improvement of approximately 20% to 41% across various scenarios.

We explored the application of Risk-averse Reinforcement Learning (Risk-averse RL) in Constrained Markov Decision Process (CMDP) in optimizing investment portfolios, incorporating constraints assessment. The investment portfolio must be always constrained with risk characteristics by investors and regulators. Therefore, the hard constraint is necessary for the practical Portfolio optimization. Moreover, traditional portfolio optimization techniques lack flexibility to model complex dynamic financial market. To address this issue, Augmented Lagrangian Multiplier (ALM) was employed to enforce constraints on the agent, mitigating the impact of risk in the decision process. Our proposal of the risk-constrained RL algorithm demonstrated no constraint violations during the testing phase, and outperformance compared to other Risk-averse RL algorithms, fulfilling our primary goal. This suggests that incorporating a risk-constrained RL technique holds promise for portfolio optimization, particularly for risk-averse investors.

Machine Learning (ML) is recently considered a key technology for bringing outstanding performance to wireless communications. Conventional research has highlighted the potential of Support Vector Machines (SVMs), which train their model based on optimization theory, to enhance the performance of wireless communications. However, there are practical issues that makes SVM difficult to apply to a wireless communication system. SVM generally entails a heavy training process with high computational complexity, and the model requires a significant amount of time for training. Also, the entire dataset needs to be trained at once, requiring a substantial amount of memory for data storage. To enable SVM in wireless communications, we propose Real-Time Channel Vector Classifier (RTCVC), which employs a light-weight SVM model capable of training and processing incoming data in real-time. A novel input data pre-processing technique is implemented to reduce the computational overhead associated with calculating non-linear functions. The rearranged formulation of the original problem also allows each SVM sub-model to be trained distributively over time based on incremental parameters. For performance evaluation, we implement the RTCVC inter-operating with 5G beam index detection, whose detection probability has been theoretically proven to be significantly enhanced by SVM. The software modules of the RTCVC are based on LibSVM, a well-known open-source library for implementing SVM sub-models. The experimental results confirm that RTCVC significantly reduces training time while maintaining suitable performance for 5G beam index detection.

We present an imagification approach for multivariate time-series data tailored to constrained NN-based forecasting model training environments. Our imagification process consists of two key steps: Re-stacking and time embedding. In the Re-stacking stage, time-series data are arranged based on high correlation, forming the first image channel using a sliding window technique. The time embedding stage adds two additional image channels by incorporating real-time information. We evaluate our method by comparing it with three benchmark imagification techniques using a simple CNN-based model. Additionally, we conduct a comparison with LSTM, a conventional time-series forecasting model. Experimental results demonstrate that our proposed approach achieves three times faster model training termination while maintaining forecasting accuracy.

This work proposes and analyzes IncepSeqNet which is a new model combining the Inception Module with the innovative Multi-Shape Augmentation technique. IncepSeqNet excels in feature extraction from sequence signal data consisting of a number of complex numbers to achieve superior classification accuracy across various SNR(Signal-to-Noise Ratio) environments. Experimental results demonstrate IncepSeqNet’s outperformance of existing models, particularly at low SNR levels. Furthermore, we have confirmed its applicability in practical 5G systems by using real-world signal data.

The utilization of sequence signals in real-world mobile communications plays a crucial role in the design and optimization of communication methods. Through our own performance evaluation, we have confirmed that conventional augmentation techniques, mainly designed for image or photograph data, are unsuitable for sequence signal applications due to inherent differences in data characteristics. To address this practical limitation, multishape augmentation (MuShAug) employs sequence signal-to-image (SSI) to represent sequence signals in image format, enabling the extraction of diverse signal features. To evaluate the practical applicability of our proposed method, we conduct experiments using real-world sequence signals collected from operational Fifth Generation (5G) mobile communication systems. In experimental trials, MuShAug consistently demonstrates robust generalization performance, achieving high levels of classification accuracy. Furthermore, through the incorporation of random phase transformation (RPT), our method achieves further enhanced performance within advanced data augmentation techniques.

Artificial intelligence has recently permeated every field, and current research trends in wireless communications naturally involve leveraging machine learning (ML) for communications processing. Numerous previous research studies have theoretically demonstrated that intelligentizing physical layer holds the promise of enhancing performance for the 6G era. In this article, we demonstrate the practical implementation of intelligentizing random access (RA) through real-time channel learning (CL). Real-time CL adapts an ML model to the variations of wireless channel in real time and requires extremely short and low-complexity training. Our research encompasses algorithmic design through the implementation of an off-the-shelf testbed which incorporates real-time CL in softwarized modem. We initially review the fundamental algorithms of communications processing for RA and then proceed to design a proper ML model for real-time CL. The overall intelligentized RA detector is then implemented using the concept of softwarized modem. The implementation achieves real-time CL via efficient interoperation of communications processing and the ML model. Experiments with the implementation confirm that intelligentization through real-time CL is feasible and results in an SNR gain up to 2.7 dB for RA scenarios.

This paper describes an iterative learning framework consisting of multi-layer prediction processes for underwater link adaptation. To obtain a dataset in real underwater environments, we implemented OFDM (Orthogonal Frequency Division Multiplexing)-based acoustic communications testbeds for the first time. Actual underwater data measured in Yellow Sea, South Korea, were used for training the iterative learning model. Remarkably, the iterative learning model achieves up to 25% performance improvement over the conventional benchmark model.

This study demonstrates a signal‐to‐noise ratio (SNR) analysis and estimation algorithm for efficient phase noise mitigation that can be practically applied to single‐carrier frequency‐domain‐equalisation (SC‐FDE) systems that operate in millimetre‐wave bands. First, the effect of phase noise in SC‐FDE systems is investigated on each of the packet reception processes, namely, channel estimation, SNR estimation, and data‐field reception. According to the analysis, an SNR estimation algorithm is proposed. The performance of minimum‐mean‐square‐error equalisation and conventional phase noise mitigation algorithm can be enhanced using the proposed SNR estimation. The effectiveness of the proposed analysis and SNR estimation algorithm is verified through the link‐level simulation. Compared with the conventional SNR estimation and the iterative phase noise mitigation algorithms, the proposed algorithm provides a lower packet‐error rate without any iterative decoding process.

The explosively growing demand of Internet of Things (IoT) has rendered broadscale advancements in the fields across sensors, radio access, network, and hardware/software platforms for mass market applications. In spite of the recent advancements, limited coverage and battery for persistent connections of IoT devices still remains a critical impediment to practical service applications. In this paper, we introduces a cost-effective IoT solution consisting of device platform, gateway, IoT network, and platform server for smart railway infrastructure. Then, we evaluate and demonstrate the applicability through an in-depth case study related to IoT-based maintenance by implementing a proof of concept and performing experimental works. The IoT solution applied for the smart railway application makes it easy to grasp the condition information distributed over a wide railway area. To deduce the potential and feasibility, we propose the network architecture of IoT solution and evaluate the performance of the candidate radio access technologies for delivering IoT data in the aspects of power consumption and coverage by performing an intensive field test with system level implementations. Based on the observation of use cases in interdisciplinary approaches, we figure out the benefits that the IoT can bring.

To improve system throughput and extend coverage, multi-hop relaying is a promising technology in wireless communication systems. Thus, multi-hop relaying is world-widely gaining huge attention as a key research item. Nevertheless, it may cause longer delay and degrade resource efficiency at the cost of the benefits due to multi-hop relaying. In this paper, we propose a traffic adaptive uplink scheduling algorithm for relay station, which could be applied to the centralized multi-hop system such as IEEE 802.16 MMR efficiently. The proposed algorithm can overcome those defects of multi-hop relaying technology and enhance system performance.

In high speed downlink packet access (HSDPA), packet scheduler is a key element for high-speed and efficient transmissions. In this paper, we propose an enhanced packet scheduling algorithm which computes the priority according to the proportional fair algorithm. That is, for users to need hybrid ARQ retransmission in case of erroneous packets, the priority of the proportional fair algorithm is elegantly modified to reduce the transfer delay without degradation of system throughput.