Update Scheme Research Articles

Point-to-Point Learning Tracking Control via Fading Communication Using Reference Update Strategy.

The networked structure has attracted significant attention due to high demand for industrial systems and rapid developments of network communication. Among various network randomness, fading is a common phenomenon, which can lead to signal attenuation, distortion, loss, and interference. This study concentrates on the point-to-point tracking problem via fading communications by proposing a reference update strategy. Using this strategy, the tracking performance is continuously improved even with faded information as the number of iterations increases. A learning control scheme is established and proved convergent in both mean-square and almost-sure senses under mild conditions. The convergence rate is accelerated by introducing the virtual reference compared with the traditional update approach. Illustrative simulations verify the theoretical results.

Reconfiguration of distributed power distribution networks based on improved gravitational search algorithms

In order to solve the problem of weak global search ability and premature convergence of GSA, a hybrid algorithm combining GSA and PSO is proposed. This algorithm can effectively use the advantages of PSO in global search. In addition, the update strategy of the gravitational constant is optimized, which improves the early search ability of the algorithm. Taking the reduction of network loss as the objective function, the improved algorithm is used to reconstruct the distribution network with distributed generation. The simulation results of the IEEE33 system show that this algorithm can efficiently decrease the network loss and enhance the node voltage.

A Reinforcement Learning Approach for Flexible Job Shop Scheduling Problem With Crane Transportation and Setup Times.

Flexible job shop scheduling problem (FJSP) has attracted research interests as it can significantly improve the energy, cost, and time efficiency of production. As one type of reinforcement learning, deep Q-network (DQN) has been applied to solve numerous realistic optimization problems. In this study, a DQN model is proposed to solve a multiobjective FJSP with crane transportation and setup times (FJSP-CS). Two objectives, i.e., makespan and total energy consumption, are optimized simultaneously based on weighting approach. To better reflect the problem realities, eight different crane transportation stages and three typical machine states including processing, setup, and standby are investigated. Considering the complexity of FJSP-CS, an identification rule is designed to organize the crane transportation in solution decoding. As for the DQN model, 12 state features and seven actions are designed to describe the features in the scheduling process. A novel structure is applied in the DQN topology, saving the calculation resources and improving the performance. In DQN training, double deep Q-network technique and soft target weight update strategy are used. In addition, three reported improvement strategies are adopted to enhance the solution qualities by adjusting scheduling assignments. Extensive computational tests and comparisons demonstrate the effectiveness and advantages of the proposed method in solving FJSP-CS, where the DQN can choose appropriate dispatching rules at various scheduling situations.

Two-stream neural network with different gradient update strategies

Deep neural networks will be affected by various noises in different scenes. Traditional deep neural networks often use gradient descent algorithms to update parameter weights. When the gradient falls to a certain range, it is easy to fall into the local optimal solution. Although the impulse method and other methods can escape from local optimization in some scenarios, they still have some limitations, which will greatly reduce the application effect of the actual scenes. To solve the above problems, a two-stream neural network with different gradient update strategies was proposed. Combined with the gradient ascent algorithm, this method alleviated the disadvantage of deep neural networks falling into local optimality and increased the robustness of neural networks to a certain extent. The experimental results on the CIFAR10 dataset verify that the proposed method can improve the accuracy of various gradient descent optimizers by about 1%, such as SGD, Adagrad, RM-Sprop and Adam. The experimental results on the COCO dataset show that the accuracy of the proposed method is also improved compared with the baseline models PAA and EfficientDet. The method proposed can be widely used in various neural network structures and has good practical significance and application prospects.

Enhanced nitrous acid (HONO) formation via NO2 uptake and its potential contribution to heavy haze formation during wintertime

A full understanding of the sources of atmospheric nitrous acid (HONO) in the polluted urban atmosphere remains a challenge. In this study, ambient HONO and relevant species were measured during January 2019 at an urban site in Beijing, China, and a budget analysis of HONO was conducted using a box model combined with field observations. Large nighttime “missing sources” of HONO were identified on heavily polluted days based on traditional sources, which had a significant correlation with the relative humidity, ammonia (NH3), and aerosol surface area, and the promotional effect of NH3 for nitrogen dioxide (NO2) uptake on the wet aerosol surface was discussed. Then, an updated parameterization scheme for quantifying the enhanced heterogeneous reactions of NO2 on aerosol surfaces is proposed, and the missing nighttime sources of HONO could be substantially compensated after the new scheme was incorporated. Further evaluation on the contributions of HONO to hydroxyl radicals was conducted, and the authors found that the photolysis of HONO played a dominant role in the primary OH production on the polluted days (78%–90%). The study reveals great potential of an NH3-enhanced uptake coefficient of NO2 on the aerosol surface in the nocturnal HONO budget, and highlights the significance of HONO in the strong atmospheric oxidation capability during episodes with a heavily polluted atmosphere.摘要本文研究了重污染过程中气态亚硝酸 (HONO) 的关键来源并基于化学箱模型对其源汇过程进行了闭合分析. 结果显示, 现有的HONO 源汇机制并不能解释观测期间出现的高HONO事件, 尤其是重污染过程夜间存在大量的HONO "未知源". 本文提出了一套新的HONO 源汇参数化方案, 可以较好的模拟观测期间HONO日间和夜间来源, 并进一步评估了HONO 对羟基自由基 (OH) 的贡献, 发现 HONO 光解主导了重污染天OH生成 (78%–90%). 研究结果揭示了华北城市富氨大气环境促进 NO2在气溶胶表面的摄取和非均相反应进而主导了夜间HONO生成, 凸显了HONO化学对于重污染过程维持大气氧化能力的重要作用.

Satellite-Based Background Aerosol Optical Depth Determination via Global Statistical Analysis of Multiple Lognormal Distribution

Determining background aerosol optical depth threshold value (BAOD) is critical to aerosol type identification and air pollution control. This study presents a statistical method to select the best BAOD threshold value using the VIIRS DB AOD products at 1 × 1 degree resolution from 2012 to 2019 as a major testbed. A series of multiple lognormal distributions with 1 to 5 peaks are firstly applied to fit the AOD histogram at each grid point, and the distribution with the highest correlation coefficient (R) gives preliminary estimations of BAOD, which is defined as either the intersection point of the first two normal distribution curves when having multiple peaks, or the midpoint between the peak AOD and the first AOD with non-zero probability when the mono peak is the best fit. Then, the lowest 1st to 100th percentile AOD distributions are compared with the preliminary BAOD distribution on a global scale. The final BAOD is obtained from the best cutoff percentile AOD distributions with the lowest bias compared with preliminary BAOD. Results show that the lowest 30th percentile AOD is the best estimation of BAOD for different AOD datasets and different seasons. Analysis of aerosol chemical information from MERRA-2 further supports this selection. Based on the BAOD, we updated the VIIRS aerosol type classification scheme, and the results show that the updated scheme is able to achieve reliable detection of aerosol type change in low aerosol loading conditions.

Multi-Strategy Boosted Fick's Law Algorithm for Engineering Optimization Problems and Parameter Estimation.

To address the shortcomings of the recently proposed Fick's Law Algorithm, which is prone to local convergence and poor convergence efficiency, we propose a multi-strategy improved Fick's Law Algorithm (FLAS). The method combines multiple effective strategies, including differential mutation strategy, Gaussian local mutation strategy, interweaving-based comprehensive learning strategy, and seagull update strategy. First, the differential variation strategy is added in the search phase to increase the randomness and expand the search degree of space. Second, by introducing the Gaussian local variation, the search diversity is increased, and the exploration capability and convergence efficiency are further improved. Further, a comprehensive learning strategy that simultaneously updates multiple individual parameters is introduced to improve search diversity and shorten the running time. Finally, the stability of the update is improved by adding a global search mechanism to balance the distribution of molecules on both sides during seagull updates. To test the competitiveness of the algorithms, the exploration and exploitation capability of the proposed FLAS is validated on 23 benchmark functions, and CEC2020 tests. FLAS is compared with other algorithms in seven engineering optimizations such as a reducer, three-bar truss, gear transmission system, piston rod optimization, gas transmission compressor, pressure vessel, and stepped cone pulley. The experimental results verify that FLAS can effectively optimize conventional engineering optimization problems. Finally, the engineering applicability of the FLAS algorithm is further highlighted by analyzing the results of parameter estimation for the solar PV model.

In defense and revival of Bayesian filtering for thermal infrared object tracking

Deep learning-based methods monopolize the latest research in the field of thermal infrared (TIR) object tracking. However, relying solely on deep learning models to obtain better tracking results requires carefully selecting feature information that is beneficial to representing the target object and designing a reasonable template update strategy, which undoubtedly increases the difficulty of model design. Thus, recent TIR tracking methods face many challenges in complex scenarios. This paper introduces a novel Deep Bayesian Filtering (DBF) method to enhance TIR tracking in these challenging situations. DBF is distinctive in its dual-model structure: the system and observation models. The system model leverages motion data to estimate the potential positions of the target object based on two-dimensional Brownian motion, thus generating a prior probability. Following this, the observation model comes into play upon capturing the TIR image. It serves as a classifier and employs infrared information to ascertain the likelihood of these estimated positions, creating a likelihood probability. According to the guidance of the two models, the position of the target object can be determined, and the template can be dynamically updated. Experimental analysis across several benchmark datasets reveals that DBF achieves competitive performance, surpassing most existing TIR tracking methods in complex scenarios.

Reinforcement Learning-Based Energy Management for Fuel Cell Electrical Vehicles Considering Fuel Cell Degradation

The service life and fuel consumption of fuel cell system (FCS) are the main factors limiting the commercialization of fuel cell electric vehicles (FCEV). Effective energy management strategies (EMS) can reduce fuel consumption during the cycle and prolong the service life of FCS. This paper proposes an energy management strategy based on the deep reinforcement learning (DRL) algorithm, deep Q-learning (DQL). Considering the unstable performance of conventional DQL during the training process, a new algorithm called Double Deep Q Learning (DDQL) is introduced. The DDQL uses a target evaluation network to evaluate output actions and a delayed update strategy to improve the convergence and stability of DRL. This article trains the strategy using UDDS cycle, tests it using combined cycles UDDS-WLTC-NEDC, and compares it with traditional ECM-based EMS. The results demonstrate that under the combined cycle, the strategy effectively reduced FCS voltage degradation by 50%, maintained fuel economy, and ensured consistency between the initial and final state of charge (SOC) of LIB.

Unified path planning for composite UAVs via Fermat point-based grouping particle swarm optimization

This paper is concerned with the problem of path planning for a class of composite unmanned aerial vehicles (UAVs). The composite UAV in the paper refers to an aerial vehicle consisting of a carrier UAV and several parasite UAVs to be launched, which combines the advantages of long flight range and high flexibility. A Fermat point-based grouping particle swarm optimization (FP-GPSO) algorithm is proposed to simultaneously determine the air-launch position and optimize the resulting multi-segmented paths. In the proposed approach, the particles are initialized into three groups considering the spatial distribution of the air-launch position, and one of the groups makes full use of the geometric property of the Fermat point. An update strategy of particles is then devised, in which not only the global best particle but the group-best particles are utilized to guide the other particles. By this means, as compared to conventional methods, the developed algorithm is more capable of effectively avoiding the local optimum, as such finding the optimal paths. Simulations are provided to demonstrate the superiority of FP-GPSO in path planning for composite UAVs.

Introduction of Artificial Intelligence Technologies in Russian Economy: A Practitioner’s View

The article presents the results of a generalization of the practical activities of companies developing and/or implementing artificial intelligence technologies (hereinafter referred to as AI), as well as companies that use these technologies: what scenarios for the use of AI technologies exist and in which industries, what kind of problems organizations implementing AI face, how members of the expert community of the artificial intelligence sphere consider solving these issues and what the state bodies offer. As to the state’s policy regarding the development of artificial intelligence, the article contains information about the factors built in the updated national AI development strategy, reflects the relations between AI technologies and state sovereignty, demonstrates the impact of artificial intelligence on the competitiveness of a company and human creativity. The article also presents the main instructions of the government of the Russian Federation on the development of artificial intelligence, some statistical data on the use of AI in economic and social sectors. It identifies measures to support developers and “implementers” of AI technologies offered by development institutions within the framework of the federal project “Artificial Intelligence” as a part of the national program “Digital Economy of the Russian Federation”. Much attention is paid to the issue of human resourcing in the AI sphere – what kind of specialists companies need, what their level of training should be and what they should be able to do, what requirements are imposed on AI teachers, what leading companies ask applicants for interviews and what the trajectory of “growing” talents in the field of artificial intelligence is. The final part of the article provides recommendations to students on how to prepare for the widespread use of artificial intelligence technologies.

FedDUS: Lung tumor segmentation on CT images through federated semi-supervised with dynamic update strategy

Background and ObjectiveLung tumor annotation is a key upstream task for further diagnosis and prognosis. Although deep learning techniques have promoted automation of lung tumor segmentation, there remain challenges impeding its application in clinical practice, such as a lack of prior annotation for model training and data-sharing among centers. MethodsIn this paper, we use data from six centers to design a novel federated semi-supervised learning (FSSL) framework with dynamic model aggregation and improve segmentation performance for lung tumors. To be specific, we propose a dynamically updated algorithm to deal with model parameter aggregation in FSSL, which takes advantage of both the quality and quantity of client data. Moreover, to increase the accessibility of data in the federated learning (FL) network, we explore the FAIR data principle while the previous federated methods never involve. ResultThe experimental results show that the segmentation performance of our model in six centers is 0.9348, 0.8436, 0.8328, 0.7776, 0.8870 and 0.8460 respectively, which is superior to traditional deep learning methods and recent federated semi-supervised learning methods. ConclusionThe experimental results demonstrate that our method is superior to the existing FSSL methods. In addition, our proposed dynamic update strategy effectively utilizes the quality and quantity information of client data and shows efficiency in lung tumor segmentation. The source code is released on (https://github.com/GDPHMediaLab/FedDUS).

Exploring Sparse Visual Prompt for Domain Adaptive Dense Prediction

The visual prompts have provided an efficient manner in addressing visual cross-domain problems. Previous works introduce domain prompts to tackle the classification Test-Time Adaptation (TTA) problem by placing image-level prompts on the input and fine-tuning prompts for each target domain. However, since the image-level prompts mask out continuous spatial details in the prompt-allocated region, it will suffer from inaccurate contextual information and limited domain knowledge extraction, particularly when dealing with dense prediction TTA problems. To overcome these challenges, we propose a novel Sparse Visual Domain Prompts (SVDP) approach, which applies minimal trainable parameters (e.g., 0.1%) to pixels across the entire image and reserves more spatial information of the input. To better apply SVDP in extracting domain-specific knowledge, we introduce the Domain Prompt Placement (DPP) method to adaptively allocates trainable parameters of SVDP on the pixels with large distribution shifts. Furthermore, recognizing that each target domain sample exhibits a unique domain shift, we design Domain Prompt Updating (DPU) strategy to optimize prompt parameters differently for each sample, facilitating efficient adaptation to the target domain. Extensive experiments were conducted on widely-used TTA and continual TTA benchmarks, and our proposed method achieves state-of-the-art performance in both semantic segmentation and depth estimation tasks.

PPO-Clip Attains Global Optimality: Towards Deeper Understandings of Clipping

Proximal Policy Optimization algorithm employing a clipped surrogate objective (PPO-Clip) is a prominent exemplar of the policy optimization methods. However, despite its remarkable empirical success, PPO-Clip lacks theoretical substantiation to date. In this paper, we contribute to the field by establishing the first global convergence results of a PPO-Clip variant in both tabular and neural function approximation settings. Our findings highlight the O(1/√T ) min-iterate convergence rate specifically in the context of neural function approximation. We tackle the inherent challenges in analyzing PPO-Clip through three central concepts: (i) We introduce a generalized version of the PPO-Clip objective, illuminated by its connection with the hinge loss. (ii) Employing entropic mirror descent, we establish asymptotic convergence for tabular PPO-Clip with direct policy parameterization. (iii) Inspired by the tabular analysis, we streamline convergence analysis by introducing a two-step policy improvement approach. This decouples policy search from complex neural policy parameterization using a regression-based update scheme. Furthermore, we gain deeper insights into the efficacy of PPO-Clip by interpreting these generalized objectives. Our theoretical findings also mark the first characterization of the influence of the clipping mechanism on PPO-Clip convergence. Importantly, the clipping range affects only the pre-constant of the convergence rate.

An active learning Kriging model with approximating parallel strategy for structural reliability analysis

With the ever-increasing complexity of engineering problems, active learning functions fused with Kriging models are receiving significant attention and are extensively applied in various reliability analysis methods. It is argued unfavorably that the existing fusion-type methods usually obtain only a single optimal sample point during model updating process. Meanwhile, the information of the used sample points is not sufficiently utilized. To these gaps, this study proposes an adaptive Kriging reliability analysis method based on an approximate parallel computing strategy, namely AP-AK for short. To start with, a new Kriging model update strategy is proposed, where the first selected sample points help to introduce the intermediate model and further search for optimal points closer to the limit state surface. An improved learning function is then proposed to limit the locations of points based on the information of used sample points. The new function is characterized by setting the acceptance and rejection domains of the sample points, which prevents computational waste incurred by local sample point aggregation. Superior performance of the AP-AK method is demonstrated against other Kriging-based methods through five numerical cases and one solid attitude orbit control engine case. The comparison results show that the proposed method exhibits high efficiency and robustness for solving complex reliability analysis problems.

Efficient Asynchronous Federated Learning with Prospective Momentum Aggregation and Fine-Grained Correction

Asynchronous federated learning (AFL) is a distributed machine learning technique that allows multiple devices to collaboratively train deep learning models without sharing local data. However, AFL suffers from low efficiency due to poor client model training quality and slow server model convergence speed, which are a result of the heterogeneous nature of both data and devices. To address these issues, we propose Efficient Asynchronous Federated Learning with Prospective Momentum Aggregation and Fine-Grained Correction (FedAC). Our framework consists of three key components. The first component is client weight evaluation based on temporal gradient, which evaluates the client weight based on the similarity between the client and server update directions. The second component is adaptive server update with prospective weighted momentum, which uses an asynchronous buffered update strategy and a prospective weighted momentum with adaptive learning rate to update the global model in server. The last component is client update with fine-grained gradient correction, which introduces a fine-grained gradient correction term to mitigate the client drift and correct the client stochastic gradient. We conduct experiments on real and synthetic datasets, and compare with existing federated learning methods. Experimental results demonstrate effective improvements in model training efficiency and AFL performance by our framework.

GAD-PVI: A General Accelerated Dynamic-Weight Particle-Based Variational Inference Framework

Particle-based Variational Inference (ParVI) methods approximate the target distribution by iteratively evolving finite weighted particle systems. Recent advances of ParVI methods reveal the benefits of accelerated position update strategies and dynamic weight adjustment approaches. In this paper, we propose the first ParVI framework that possesses both accelerated position update and dynamical weight adjustment simultaneously, named the General Accelerated Dynamic-Weight Particle-based Variational Inference (GAD-PVI) framework. Generally, GAD-PVI simulates the semi-Hamiltonian gradient flow on a novel Information-Fisher-Rao space, which yields an additional decrease on the local functional dissipation. GAD-PVI is compatible with different dissimilarity functionals and associated smoothing approaches under three information metrics. Experiments on both synthetic and real-world data demonstrate the faster convergence and reduced approximation error of GAD-PVI methods over the state-of-the-art.

Generalizable Fourier Augmentation for Unsupervised Video Object Segmentation

The performance of existing unsupervised video object segmentation methods typically suffers from severe performance degradation on test videos when tested in out-of-distribution scenarios. The primary reason is that the test data in real- world may not follow the independent and identically distribution (i.i.d.) assumption, leading to domain shift. In this paper, we propose a generalizable fourier augmentation method during training to improve the generalization ability of the model. To achieve this, we perform Fast Fourier Transform (FFT) over the intermediate spatial domain features in each layer to yield corresponding frequency representations, including amplitude components (encoding scene-aware styles such as texture, color, contrast of the scene) and phase components (encoding rich semantics). We produce a variety of style features via Gaussian sampling to augment the training data, thereby improving the generalization capability of the model. To further improve the cross-domain generalization performance of the model, we design a phase feature update strategy via exponential moving average using phase features from past frames in an online update manner, which could help the model to learn cross-domain-invariant features. Extensive experiments show that our proposed method achieves the state-of-the-art performance on popular benchmarks.

ODTrack: Online Dense Temporal Token Learning for Visual Tracking

Online contextual reasoning and association across consecutive video frames are critical to perceive instances in visual tracking. However, most current top-performing trackers persistently lean on sparse temporal relationships between reference and search frames via an offline mode. Consequently, they can only interact independently within each image-pair and establish limited temporal correlations. To alleviate the above problem, we propose a simple, flexible and effective video-level tracking pipeline, named ODTrack, which densely associates the contextual relationships of video frames in an online token propagation manner. ODTrack receives video frames of arbitrary length to capture the spatio-temporal trajectory relationships of an instance, and compresses the discrimination features (localization information) of a target into a token sequence to achieve frame-to-frame association. This new solution brings the following benefits: 1) the purified token sequences can serve as prompts for the inference in the next video frame, whereby past information is leveraged to guide future inference; 2) the complex online update strategies are effectively avoided by the iterative propagation of token sequences, and thus we can achieve more efficient model representation and computation. ODTrack achieves a new SOTA performance on seven benchmarks, while running at real-time speed. Code and models are available at https://github.com/GXNU-ZhongLab/ODTrack.

Abstract 5515: Low body mass index (BMI) induces neuronal NPY and promotes brain metastasis of lung cancer

Abstract Brain metastases (BrMs) are a common occurrence in lung cancer with a dismal outcome. To develop a more effective therapeutic strategy, it is crucial to understand the underlying pathological mechanism of brain metastasis. Recent clinical data has unveiled a significant association between low body mass index (BMI) and an increased risk of brain metastasis in lung cancer patients, leading to reduced survival rates compared to other primary tumors. However, the molecular mechanism underlying this paradoxical relationship, often referred to as the "obesity paradox," in the context of lung cancer mortality presents a complex and intriguing puzzle and warrants in-depth investigation. Here, we examined biological mechanism of how low BMI promotes lung cancer brain metastasis. In this study, we conducted a retrospective pan-analysis of 5,516 patient cohorts with brain metastasis originating from lung, breast, melanoma, and renal cancers with their BMI status. We found that lung cancer patients with low BMI have significantly higher brain metastatic incidence compared to the high BMI patients, in contrast to breast, melanoma, and renal cancers. Furthermore, we found that low BMI plays a pivotal role in mediating neuron activation within the brain and that the secretion of neuronal neuropeptide Y (NPY), a protein involved in appetite and energy homeostasis via GHSR-receptor manner, promotes metabolic switching through NPY/NPY5R axis in the tumor cells in the brain thereby enabling metastasis. Elevated levels of neuronal NPY in the brains of cancer-free subjects with low BMI suggest its potential utility as a promising prognostic biomarker for identifying the increased risk of metastatic disease in lung cancer patients with low BMI. We also demonstrated that reversing low BMI or blocking the NPY/Y5R interactions effectively abrogated brain metastasis in our animal model. Our findings suggest a novel pro-metastatic role of low BMI-induced NPY in the progression of brain metastasis. We emphasize the importance of elucidating the clinical implications of this relationship to develop an updated intervention strategy for the clinical care of patients with lung cancer and low BMI. Citation Format: Abhishek Tyagi, Shih-Ying Wu, Ravindra Deshpande, Kerui Wu, Liang Liu, Kounosuke Watabe. Low body mass index (BMI) induces neuronal NPY and promotes brain metastasis of lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 5515.