Abstract ：

The performance of an evolutionary algorithm (EA) is deeply affected by its control parameter's setting. It has become a trend in recent studies to treat the control parameter as a random variable. In these studies, the associated distribution of the control parameter is updated at each generation and new parameter setting is sampled from the distribution. The distribution's parameter (called hyper-parameter) is thus critical to the algorithmic performance. In this paper, we propose a variational learning framework to tune the hyper-parameters of EA, in which the expectation-maximization (EM) algorithm and a reinforcement learning algorithm are combined. To verify the effectiveness of the proposed method which is named Reinforcement EM (REM), we apply it to tune the hyper-parameters of the distributions of two important parameters, i.e. the scaling parameter (F) and crossover rate (CR), of differential evolution (DE) and an adaptive DE algorithm. In addition, we propose to use the meta-learning technique to learn good initial distributions for the hyper-parameters of F and CR so that the REM can effectively adapt to a new optimization problem. Experimental results obtained on the CEC 2018 test suite show that with the tuned hyper-parameters, DE and the adaptive DE can achieve significantly better performance than their counterparts with empirical parameter settings and with parameters tuned by some widely-used tuning methods, including ParamILS, F-Race and Bayesian optimization algorithm.

Keyword ：

evolutionary algorithm expectation-maximization parameter tuning reinforcement learning Variational inference

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Recording moving magnetoencephalograms (MEGs), in which a person's head can move freely as the brain's magnetic field is recorded, has been a key subject in recent years. Here, we describe a method based on an optically pumped atomic co-magnetometer (OPACM) for recording moving MEGs. In the OPACM, hyper-polarized nuclear spins produce a magnetic field that blocks the background fluctuation low-frequency magnetic field noise while the rapidly changing MEG signal is recorded. In this study, the magnetic field compensation was studied theoretically, and we found that the compensation is closely related to several parameters such as the electron spin magnetic field, nuclear spin magnetic field, and holding magnetic field. Furthermore, the magnetic field compensation was optimized based on a theoretical model . We also experimentally studied the magnetic field compensation and measured the responses of the OPACM to different magnetic field frequencies. We show that the OPACM clearly suppresses low-frequency (under 1 Hz) magnetic fields. However, the OPACM responses to magnetic field frequencies around the band of the MEG. A magnetic field sensitivity of 3 fT/Hz1/2 was achieved. Finally, we performed a simulation of the OPACM during utilization for moving MEG recording. For comparison, the traditional compensation system for moving MEG recording is based on a coil that is around 2 m in dimension, while our compensation system is only 2 mm in dimension . © 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement.

Keyword ：

Brain mapping Electrospinning Magnetic field measurement Magnetic fields Magnetic moments Magnetoencephalography Magnetometers Spin dynamics

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Hyper compressors are key facilities for producing the low-density polyethylene with discharge pressure up to 350 MPa. Such high pressure brings great challenges to the design of the hyper compressor in many aspects. In this paper, a 3D transient computational fluid dynamics (CFD) model with inlet and outlet pipelines is built to investigate the thermodynamic performance of a hyper compressor. To realize the interaction between the thermodynamic processes and the pressure pulsation through valve dynamics, the pressures across the valve surfaces were monitored to the dynamic equation of the poppet valve. Then, structured grids were generated for the flow domain inside the valve, and the entire numerical model was solved by a commercial code: ANSYS Fluent. Consequently, the p-V diagram, the valve motion and pressure pulsation could be acquired simultaneously. The results of the numerical model showed that the exponents of the expansion and compression processes were 5.12 and 13.22, which were much larger than the common compressor. The maximal pressure pulsations were 13.25% and 22.07%, which occurred in the suction and discharge chambers, respectively. Severe flutter happened during the opening process of the suction valve due to the high incompressibility of the ethylene.

Keyword ：

CFD hyper compressor pressure pulsation thermodynamic process valve dynamics

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Adaptive parameter control is critical in the design and application of evolutionary algorithm (EA), so does in differential evolution. In the past decade, many adaptive evolutionary algorithms have been proposed, in which online information collected until current generation during the evolutionary search procedure is used to determine the algorithmic parameters for the next generation. Recent studies often assume that the algorithmic parameters follow some distributions, while the distributions' parameters (called hyper-parameters) are updated by the collected information. Performances of these adaptive EAs depend highly on the hyper-parameters. Notice that the experiences obtained from optimizing some related problems could provide useful guidelines on how to adaptively control the distributions' parameters. However, few existing studies sufficiently used such experiences. To fill the gap, we propose a general framework for adaptive parameter control by modeling its evolution procedure as a Markov decision process. In the framework, a neural network is employed to act as the controller. The natural evolution strategies is applied to train the neural network. The proposed framework is applied on two well-known differential evolutions (DEs), namely JADE and LSHADE. By incorporating the learned controller, two DEs, named JADE/AC and LSHADE/AC, are formed. Experimental results on the CEC 2018 benchmark suite show that in general JADE/AC and LSHADE/AC perform significantly better than their counterparts. Moreover, in comparison with some well-known EAs including three suggested best DEs in a review paper (including LSHADE, cDE and CoBiDE), the championship algorithm in the CEC 2018 competitions, a recently-developed learnable DE and recently proposed DEs, our study shows that LSHADE/AC performs the best amongst them without sacrificing much computation time.

Keyword ：

Evolutionary algorithm History learning to optimize Markov decision process natural evolution strategies Next generation networking Q-learning Search problems Sociology Statistics Tuning

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Accurate prediction of remaining useful life (RUL) is necessary to ensure stable and safe operations for rocket engines. The paper proposed a multi-head attention network coupled with adaptive meta-transfer learning for RUL prediction. By combining the convolution-based branch with an attention-based branch, the multi-head attention network is proposed for accurate RUL prediction of cryogenic bearings in rocket engines under the steady stage. In addition, an adaptive model-agnostic meta-transfer learning strategy is developed to further improve the performance under small sample circumstances with adaptive hyper-parameters. To demonstrate the superiority, the proposed method is compared with typical benchmark algorithms using real monitoring data from a high-precision cryogenic rocket engine experiment platform. Results indicate that the proposed method achieves better performance compared with existing models under multiple evaluation indexes. © 2022 Elsevier Ltd

Keyword ：

Cryogenics Deep neural networks Forecasting Rockets

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Achieving adequate precision in deep learning based communications often requires large network architectures, which results into unacceptable time delay and power consumption. This paper introduces the dynamic neural network (DyNN) into the design of wireless communications systems. DyNN allocates different samples with computation resources on demand by preforming dynamic inferences, thereby reducing the redundant computational cost and enhancing the network efficiency. We design a dynamic depth architecture that allows samples to adaptively skip layers with various dynamic strategies, from which we further develop a confidence criterion based dynamic improved DetNet (CD-IDetNet) and a policy network based dynamic improved DetNet (PD-IDetNet) for multiple-input multiple-output (MIMO) detection. Specifically, in CD-IDetNet, a confidence criterion is adopted to control samples exiting early, while in PD-IDetNet, policy networks are trained by reinforcement learning to selectively skip layers for varying samples. Simulation results demonstrate that CD-IDetNet and PD-IDetNet detectors can respectively reduce 17.4% and 31.1% computational costs while preserving the full accuracy of IDetNet. Desirable tradeoffs between accuracy and computational complexity can be further achieved by fine-tuning the hyper-parameters of CD-IDetNet and PD-IDetNet. Moreover, over-the-air (OTA) tests are conducted to validate the effectiveness of the proposed detectors in practical systems. © 1983-2012 IEEE.

Keyword ：

Complex networks Cost reduction Deep learning Memory architecture MIMO systems Network architecture Neural networks Reinforcement learning

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Background and aims: Hypercholesterolemia is characterized by the elevation of plasma total cholesterol level, especially low-density lipoprotein (LDL) cholesterol. This disease is usually caused by a mutation in genes such as LDL receptor, apolipoprotein B, or proprotein convertase subtilisin/kexin type 9. However, a considerable number of patients with hypercholesterolemia do not have any mutation in these candidate genes. In this study, we examined the difference in the metabolic level between patients with hyper-cholesterolemia and healthy subjects, and screened the potential biomarkers for this disease. Methods: Analysis of plasma metabolomics in hypercholesterolemia patients and healthy controls was performed by gas chromatography-mass spectrometry and metabolic correlation networks were constructed using Gephi-0.9.2. Results: First, metabolic profile analysis confirmed the distinct metabolic footprints between the patients and the healthy ones. The potential biomarkers screened by orthogonal partial least-squares discrimination analysis included L-lactic acid, cholesterol, phosphoric acid, D-glucose, urea, and D-allose (Variable importance in the projection > 1). Second, arginine and methionine metabolism were significantly perturbed in hypercholesterolemia patients. Finally, we identified that L-lactic acid, L-lysine, L-glutamine, and L-cysteine had high scores of centrality parameters in the metabolic correlation network. Conclusion: Plasma L-lactic acid could be used as a sensitive biomarker for hypercholesterolemia. In addition, arginine biosynthesis and cysteine and methionine metabolism were profoundly altered in patients with hypercholesterolemia. (C) 2021 Elsevier Inc. All rights reserved.

Keyword ：

Biomarker Correlation network analysis Hypercholesterolemia Metabolomics analysis

Cite：

Copy from the list or Export to your reference management。

Abstract ：

An efficient approach for handling hyperspectral image (HSI) denoising issue is to impose weights on different HSI pixels to suppress negative influence brought by noisy elements. Such weighting scheme, however, largely depends on the prior understanding or subjective distribution assumption on HSI noises, making them easily biased to complicated real noises, and hardly generalizable to diverse practical scenarios. Against this issue, this paper proposes a new scheme aiming to capture general weighting principle in a data-driven manner. Specifically, such weighting principle is delivered by an explicit function, called hyper-weight-net (HWnet), mapping from an input noisy image to its properly imposed weights. A Bayesian framework as well as a variational inference algorithm for inferring HWnet parameters is elaborately designed, expecting to extract the latent weighting rule for general diverse and complicated noisy HSIs. Comprehensive experiments substantiate that the learned HWnet can be not only finely generalized to different noise types from those used in training, but also effectively transferred to other weighted models. Besides, as a sounder guidance, HWnet can help to more faithfully and robustly achieve deep hyperspectral prior(DHP). The extracted weights by HWnet are verified to be able to effectively capture complex noise knowledge underlying input HSI, revealing its working insight in experiments. © 2021 IEEE

Keyword ：

Inference engines Spectroscopy

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Cross-contrast image translation is an important task for completing missing contrasts in clinical diagnosis. However, most existing methods learn separate translator for each pair of contrasts, which is inefficient due to many possible contrast pairs in real scenarios. In this work, we propose a unified Hyper-GAN model for effectively and efficiently translating between different contrast pairs. Hyper-GAN consists of a pair of hyper-encoder and hyper-decoder to first map from the source contrast to a common feature space, and then further map to the target contrast image. To facilitate the translation between different contrast pairs, contrast-modulators are designed to tune the hyper-encoder and hyper-decoder adaptive to different contrasts. We also design a common space loss to enforce that multi-contrast images of a subject share a common feature space, implicitly modeling the shared underlying anatomical structures. Experiments on two datasets of IXI and BraTS 2019 show that our Hyper-GAN achieves state-of-the-art results in both accuracy and efficiency, e.g., improving more than 1.47 and 1.09 dB in PSNR on two datasets with less than half the amount of parameters.

Keyword ：

Multi-contrast MR Unified hyper-GAN Unpaired image translation

Cite：

Copy from the list or Export to your reference management。

Abstract ：

Impulsive blind deconvolution (IBD) is a fundamental ill-posed inverse problem in fault diagnosis community. Current IBD methods mainly utilize the intrinsic prior knowledge of impulsive sources to design various regularization terms (e.g., kurtosis, sparsity) to alleviate its ill-posedness. However, the great potentiality of statistical distribution structures embedded in observation data hasn't been exploited to establish more effective model and algorithm for IBD problem. Leveraging recent plug-and-play (PnP) strategy, a convolutional sparse optimization framework (dubbed COPS) is proposed to address it. Firstly, based on the fact that the absolute envelope of Gaussian noises follows a Rayleigh distribution and sparse impulsive components can be viewed as its outliers, a noise-aware statistical threshold is introduced to design a data-driven sparse penalty. Secondly, from a Bayesian perspective, a mapping relation between residual distribution and model hyper-parameter is unveiled, and then an adaptive parameter penalty is established to dynamically select model hyper-parameters. Lastly, two penalties are plugged into ADMM solver by PnP strategy to guarantee algorithmic convergence. Comprehensive numerical simulations confirm the COPS's advantages in terms of robustness, convergence, scalability and effectiveness. Diagnostic results of planetary gearbox faults further corroborate the COPS retains better deconvolutional accuracy than the state-of-the-art IBD techniques. © 2021 Elsevier Ltd

Keyword ：

Convolution Fault detection Gaussian noise (electronic) Gears Inverse problems

Cite：

Copy from the list or Export to your reference management。