Deep learning is particularly important in the field of time series data analysis, and has been applied to tasks such as marine data prediction.However, there is a 'concept drift' problem in marine observation data, which leads to performance degradation and catastrophic forgetting of traditional deep learning models used in online scenarios.For this reason, this paper proposes OL-TCN, an adaptive temporal convolutional neural network online prediction deep learning model, which is more suitable for marine online learning and inference scenarios.We make the following innovations: Inside the model, model performance is enhanced by incorporating a multi-head attention mechanism and introducing an automatic machine learning enhancement in the residual cell.Outside the model, a model repository approach is used to effectively cope with the complexity and evolutionary challenges of data streams.The experimental results verify that the OL-TCN model is effective and feasible in marine time series processing.

In recent years, with the continuous development of artificial intelligence, deep learning algorithms are becoming more and more complex, and the scale of model training is also growing. The artificial intelligence platform also involves large-scale model training in our computing network operating system project. However, with the increasing size of data sets and models, the traditional single-card training makes the training speed very slow, and the training accuracy needs to converge, which has yet to meet people's computational needs. This has led to the development of GPipe, PipeDream, and other famous pipelines. In this paper, an efficient pipeline parallel training optimization method is proposed. In our approach, multiple computing nodes process small batches of data in parallel in a pipeline manner. We have mainly done the following two aspects of work: First, we designed a weight buffer strategy to limit the number of weight versions generated and ensure the model's accuracy. And we also developed a tensor compression mechanism to improve the transmission rate. Secondly, we propose a prefix sum partition algorithm to ensure that the pipeline can achieve balanced partitioning and save the memory of computing resources. Compared with several popular pipeline parallel frameworks, the proposed method can achieve about twice the training acceleration and save about 30% - 40% of the memory usage.

In the context of the rapid development of big data and artificial intelligence, the field of HPC is facing significant challenges in energy consumption. To address this challenge, this paper proposes a power consumption prediction framework for HPC jobs based on self-supervised contrastive learning, named Job2Vec. Firstly, HPC jobs are clustered into different clusters based on job logs and power consumption curve shapes. Next, high-quality features of power consumption data at different granularities within each cluster are captured through self-supervised contrastive learning. Lastly, an incremental update strategy is introduced to handle the dynamic changes in HPC data. The experimental results demonstrate that, on the Jinan Supercomputing dataset comprising 1170 HPC jobs, the Job2Vec framework outperforms the baseline models overall in terms of prediction performance. Additionally, we also investigated the model’s performance in single-step and multi-step forecasting, the impact of different indicators on prediction accuracy, and conducted ablation experiments to confirm the indispensability of each module. This is the first time that self-supervised contrastive learning has been introduced into the HPC field, aiming to provide support for energy scheduling in data centers.

The authors propose a generalised multi‐layer architecture for the phase shifter network (PSN) in hybrid analogue/digital beamforming (BF), which provides flexible trade‐offs between the performance and implementation complexity. By changing the layer number, the authors can adjust the complexity of the PSN and the achievable performance. A two‐stage BF algorithm is adopted, in which the analogue and digital BF matrices are designed separately. Analogue BF is adaptive to the second‐order channel statistics. A simple iterative algorithm is proposed to maximise an upper bound of the average rate of the effective channel after analogue BF. Digital BF is adaptive to the effective channel for multiuser interference cancellation. Both ideal phase shifters and discrete phase shifters with limited phase and amplitude resolutions are considered in the algorithm design. Theoretical analysis and simulation are used to demonstrate the good performance and flexibility of the proposed technique. In a 64‐antenna system, it can use 16 radio‐frequency chains to perform very closely to full digital BF with a fully‐connected PSN, or achieve 70–80% performance with a simple PSN. It is also robust against the limited phase shifter resolution. It enables us to select the proper performance and implementation complexity according to the requirement and capability of real systems.

In downlink frequency division duplex (FDD) systems, the benefits of multi-user (MU) MIMO highly rely on the accuracy of channel quantization based on a predefined codebook. Linear antenna array is widely deployed in realistic systems, including various configurations such as uniformly-spaced linear array (ULA) and cross-polarized linear array (CLA) with closely-spaced or widely-spaced antennas. In this paper, a unified codebook structure is proposed for all these antenna arrays, in a form of two-stage feedback. The average cell spectrum efficiency and cell-edge user spectrum efficiency of the proposed codebook are evaluated through system level simulation and compared to the system performance with long term evolution (LTE) Rel-8 codebook and discrete Fourier transform (DFT) codebook. Simulation results show that the proposed codebook significantly improves channel feedback efficiency as well as system performance.

iVDR (Information Versatile Disk Removable) is new generation large-capacity hard disk for the digital high-definition era. In this paper, it is described that iVDR has excellent performance for large-capacity storage, mobile storage, and DRM (Digital Right Management). Then a solution of iVDR' storage and encryption for Internet video is designed and proposed. And 3 formats of Internet video file, FLV (Flash Video), MP4, and TS (Transport Stream), are analyzed to be encrypted in iVDR by the designed solution. At last, an experiment is done to prove the feasibility of the solution.

Multiple-input multiple-output (MIMO) has been adopted by long term evolution (LTE) and its updated version LTE-Advanced (LTE-A) to increase spectrum efficiency. However, the benefits of multi-user (MU) MIMO highly rely on the accurate channel knowledge at transmitter. So proper codebook design is a key problem for frequency division multiplexing (FDD) downlink systems. In this paper, an efficient codebook design based on Hadamard transform is proposed for uniform circular arrays. Then a two-stage feedback scheme is further proposed to reduce the feedback overhead. The average cell spectrum efficiency and cell-edge user spectrum efficiency of the proposed scheme are evaluated through system level simulation and compared with LTE Rel-10 feedback. The results show that the proposed codebook significantly improves feedback efficiency and system performance.

This work investigates implementing spatial-correlation-information aided feedback in practical multiple-input multiple-output (MIMO) systems. It is presented that the existing spatial-correlation-information aided feedback approach could be simplified in computational complexity with little performance loss. In comparison with Grassmannian codebooks, the feedback codebooks in the 3GPP LTE Release-8 specification is shown to be a near-optimal base codebook. An efficient quantization approach is proposed for quantizing transmit spatial covariances matrices for closely-spaced co-polarized uniform linear arrays (ULA). The good performance of the entire practical feedback solution is demonstrated via system-level simulation.

Multiple-input multiple output (MIMO) has been adopted by long term evolution (LTE) and its updated version LTE-Advanced (LTE-A) to increase system spectrum efficiency. However, the benefits of multi-user (MU) MIMO highly rely on an accurate channel knowledge at the transmitter. So proper codebook design is a key problem for LTE-A FDD downlink system. In this paper, the channel characteristic specific to cross-polarized linear antenna array is analyzed, then its codebook design based on amplitude and phase adjustment is proposed for LTE-A downlink system. The proposed codebooks are compared to the differential scheme [7] on the average cell spectral efficiency and cell-edge user spectral efficiency evaluated through system level simulations. Simulation results show that the proposed codebooks significantly improve channel feedback efficiency and system performance.

Multiple-input multiple-output (MIMO) has been adopted by long term evolution (LTE) and its updated version LTE-Advanced (LTE-A) to increase the spectrum efficiency. However, the benefits of multi-user (MU) MIMO highly rely on the accurate channel knowledge at the transmitter. So proper codebook design is a key problem for LTE-A FDD downlink system. In this paper, a codebook design with phase adjustment in the precoder targeting wideband channel properties is proposed. The average cell spectrum efficiency and cell-edge user spectrum efficiency of the proposed scheme are evaluated through system level simulation and compared to the discrete Fourier transform (DFT) based feedback scheme. The results show that the proposed codebooks significantly improve feedback efficiency and system performance for co-polarized linear antenna array.

We propose a capacity balancing beamforming technique for a multiple inputs and multiple outputs (MIMO) based cognitive radio (CR) network. The proposed algorithm is based on mean square error (MSE) duality and it is aimed at maximizing the worst case user capacity of multiple secondary users (SUs) by jointly designing their transceiver beamformers while ensuring the interference leakage to multiple primary users (PUs) are below a specific set of thresholds. We use an iterative approach to determine the transceiver beamformers so that the capacities achieved by all SUs are balanced and maximized. The performance of the algorithm is demonstrated through Monte Carlo simulation results.

This letter proposes a blind carrier frequency offset (CFO) estimator for orthogonal frequency-division multiplexing (OFDM) systems based on the frequency analysis of the received signal, and derives a closed-form CFO estimate. Since only one OFDM symbol is utilized instead of multi-symbol averaging, the proposed method is effective even when the CFO is time varying. Finally, analysis and simulation results indicate the outstanding performance of the proposed estimator.

Carrier frequency offset (CFO) estimation is a critical problem in orthogonal frequency-division multiplexing (OFDM) systems. This letter proposes a blind CFO estimator based on smoothing the signal power spectrum. A closed-form CFO estimate is also presented, which greatly reduces the computational complexity of the proposed method. Analysis and simulation results show that the proposed estimator is very effective for OFDM systems.

Carrier frequency offset (CFO) is an important problem in multi-input multi-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems. This letter proposes a blind CFO estimator for MIMO-OFDM systems with constant modulus constellations in the frequency-selective fading channels. A closed-form CFO estimate is also provided, which greatly reduces the computational complexity of the proposed method. Analysis and simulation results demonstrate that the proposed estimator is very effective for MIMO-OFDM systems.

Carrier frequency offset (CFO) estimation is critical in practical orthogonal frequency-division multiplexing (OFDM) systems. In this paper, we propose two novel blind CFO estimators named the variance estimator (VE) and the approximate variance estimator (AVE) respectively with high bandwidth efficiency. Both of them are based on minimization of the variance of the received symbols on each subcarrier, but the AVE only utilizes a half of OFDM symbols needed in the VE at the cost of a little degradation in performance. Simulation results show that the proposed estimators are attractively efficient for the future wireless communications.

A new entropy-based thresholding is presented. As a result, a gray level minimizing the difference between the entropy of the object and that of the background will be the desired threshold. Performance of the proposed method has been tested on some ultrasonic images. The results obtained from the method proposed by Kapur et al. (1985) are also given for comparison.