Adaptive Partitioning Strategy Research Articles

DNN Adaptive Partitioning Strategy for Heterogeneous Online Inspection Systems of Substations

With the explosive development of power edge equipment and the continuous improvement in power inspection performance, the requirements of substations and terminal equipment, such as drones with limited resources, cannot meet the strict delay and energy consumption requirements. This paper proposes an adaptive partitioning strategy for heterogeneous substation inspection systems. First, a layer delay prediction model and layer energy consumption prediction model are established on each heterogeneous node, and nonlinear characteristics related to delay and energy consumption are trained. On this basis, a deep neural network (DNN) hybrid partitioning strategy is proposed. The DNN task is divided into synchronous cooperative reasoning between terminal devices and multi-heterogeneous edge nodes. The experimental results show that the average absolute percentage error (MAPE) of the delay model was reduced by 31.49% on average. On drones and mobile edge nodes, the energy consumption model MAPE reduced the average by 21.92%, and the DNN end-to-end latency was reduced by 31.48%. The total cost of the system was reduced and the efficiency of UAV inspection was improved.

Bi-directional ensemble differential evolution for global optimization

Hybridizing multiple mutation strategies has shown much effectiveness in helping differential evolution (DE) algorithms achieve good optimization performance. Though abundant adaptive ensemble strategies have been developed to adaptively employ multiple mutation strategies to evolve the population, most of them ignore to make full use of the properties and characteristics of the multiple mutation strategies. To fill this gap, this paper devises a bi-directional ensemble scheme for DE to adaptively assemble totally 8 mutation strategies with different properties and characteristics. As a result, a novel DE, which we call bi-directional ensemble DE (BDEDE), is developed. Specifically, this paper sorts the 8 mutation strategies roughly from two opposite perspectives, namely the convergence and the diversity. Then, we assign each mutation strategy with two different non-linear probabilities, which are calculated on the basis of its two rankings obtained from the two perspectives. Subsequently, to make full use of these mutation schemes, we first partition the whole population into two separate parts, namely elite individuals and non-elite individuals. Then, for each elite individual, we randomly select a mutation strategy from the 8 candidates based on the probabilities calculated by the convergence rankings, while for each non-elite individual, we stochastically choose a mutation scheme from the same 8 candidates but based on the probabilities computed by the diversity rankings. In this manner, the elite individuals prefer to exploit the located optimal areas, while the non-elite individuals tend to explore the solution space. Therefore, it is likely that BDEDE expectedly maintains a good balance between search diversity and search convergence. To further help BDEDE achieve such a purpose, this paper devises an adaptive partition strategy to dynamically separate the whole population into the two categories. With the above two techniques, BDEDE anticipatedly obtains good optimization performance. To verify its effectiveness and efficiency, we conduct experiments on the CEC2014 and the CEC2017 benchmark sets by comparing BDEDE with totally 14 well-known and state-of-the-art DE variants. Experimental results have shown that BDEDE performs competitively with or even significantly better than the 14 compared DE variants. The source code of BDEDE can be downloaded from https://gitee.com/mmmyq/BDEDE.

An adaptive partitioned reduced order model of peridynamics for efficient static fracture simulation

In this paper, an adaptive partitioning strategy and the proper orthogonal decomposition–Galerkin method are combined to formulate an adaptive partitioned reduced order model (APROM) for fast solution of peridynamic (PD) models involving fractures. A boundary layer separation approach is presented to tackle the inaccurate reproduction of nonlocal boundary conditions in the reduced order model (ROM). Cracks invariably make ROM results more sensitive to the snapshots, i.e. a set of high-fidelity solutions simulated by the full order model (FOM), and this challenge is overcome by an adaptive partitioning strategy. The internal computational domain is divided into two parts: the crack region is modeled by full PD model, and the remaining region is dimensionally reduced. The partitioned configuration is automatically updated with the damage evolution. Several numerical tests are executed to verify the performance of the APROM, which shows that the APROM is able to successfully simulate various fracture phenomena. A significant improvement in computational efficiency is found: the number of degrees of freedom in the full PD model is greatly reduced, leading to an approximately 10 times improvement in CPU time without loss of accuracy. This paper provides strong theoretical support for accelerating PD solutions, thereby promoting the practical application of PD theory in large-scale engineering projects.

Heterogeneous parallel_for Template for CPU–GPU Chips

Heterogeneous processors, comprising CPU cores and a GPU, are the de facto standard in desktop and mobile platforms. In many cases it is worthwhile to exploit both the CPU and GPU simultaneously. However, the workload distribution poses a challenge when running irregular applications. In this paper, we present LogFit, a novel adaptive partitioning strategy for parallel loops, specially designed for applications with irregular data accesses running on heterogeneous CPU–GPU architectures. Our algorithm dynamically finds the optimal chunk size that must be assigned to the GPU. Also, the number of iterations assigned to the CPU cores are adaptively computed to avoid load unbalance. In addition, we also strive to increase the programmer’s productivity by providing a high level template that eases the coding of heterogeneous parallel loops. We evaluate LogFit’s performance and energy consumption by using a set of irregular benchmarks running on a heterogeneous CPU–GPU processor, an Intel Haswell. Our experimental results show that we outperform Oracle-like static and other dynamic state-of-the-art approaches both in terms of performance, up to 57%, and energy saving, up to 31%.

Normalized Scalar Product Approach for Nearshore Bathymetric Estimation From X-Band Radar Images: An Assessment Based on Simulated and Measured Data

This paper investigates the capability of X -band radar systems to estimate nearshore bathymetry fields by considering both simulated and measured radar data. For the first time, a sensitivity analysis is performed to evaluate how sea-state conditions affect bathymetric estimates. For this purpose, sea wave fields generated by means of a numerical model, based on a nonlinear shallow-water equation solver, are used. Starting from the synthetic radar data, which represent the input of the bathymetric estimation algorithm, the bathymetric reconstruction is performed through the normalized scalar product (NSP) estimation strategy, exploiting a spatial partitioning of the radar data. In this way, it is possible to improve the accuracy of the estimates in nearshore areas, where the space-varying behavior of the sea depth and the presence of coastlines or coastal structures typically leads to a spatial inhomogeneity of the wave motion. In this regard, it is shown how the choice of the partitioning settings affects the bathymetric estimates obtained from high-resolution X -band radar images by using the NSP strategy. In addition, an adaptive partitioning strategy that takes into account the wave evolution in nearshore shallow waters is devised. Based on both simulated and measured radar data, the accuracy of the bathymetric estimates achievable through the proposed adaptive partitioning process and that obtained by exploiting the approach using uniform spatial partitioning are compared. The results obtained confirm the robustness of the NSP technique with respect to sea conditions and, moreover, demonstrate that the proposed adaptive partitioning strategy provides more accurate bathymetric estimates than those obtained with the space-invariant partitioning procedure.

Toward more efficient density‐based adaptive QM/MM methods

Adaptive QM/MM (quantum mechanical/molecular mechanical) methods repartition molecules on‐the‐fly into the QM and MM regions based on a partitioning criterion. A density‐based partitioning scheme offers a unique way of carrying out QM and MM assignments on the presence/absence of (non) covalent interactions between non‐QM core molecules and the QM core. Here, we present two improvements to promote our density‐based adaptive partitioning strategy for adaptive QM/MM calculations. First, bond formation/breaking processes between non‐QM core molecules and the QM core are monitored using a density‐based descriptor, to define an adaptive QM core region on the fly. Second, the density‐based adaptive QM/MM partitioning scheme is made efficient enough for large systems by utilizating an adaptive search region and a Voronoi style‐partitioning scheme. The adaptive search region is used to preselect the subset of molecules that may interact strongly enough with the QM core region, and then a density‐based interaction calculation is only carried out on this subset. A Voronoi style‐partitioning scheme achieves an efficient assignment of the density‐based interactions.

ARACNe-AP: gene network reverse engineering through adaptive partitioning inference of mutual information.

Summary: The accurate reconstruction of gene regulatory networks from large scale molecular profile datasets represents one of the grand challenges of Systems Biology. The Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNe) represents one of the most effective tools to accomplish this goal. However, the initial Fixed Bandwidth (FB) implementation is both inefficient and unable to deal with sample sets providing largely uneven coverage of the probability density space. Here, we present a completely new implementation of the algorithm, based on an Adaptive Partitioning strategy (AP) for estimating the Mutual Information. The new AP implementation (ARACNe-AP) achieves a dramatic improvement in computational performance (200× on average) over the previous methodology, while preserving the Mutual Information estimator and the Network inference accuracy of the original algorithm. Given that the previous version of ARACNe is extremely demanding, the new version of the algorithm will allow even researchers with modest computational resources to build complex regulatory networks from hundreds of gene expression profiles.Availability and Implementation: A JAVA cross-platform command line executable of ARACNe, together with all source code and a detailed usage guide are freely available on Sourceforge (http://sourceforge.net/projects/aracne-ap). JAVA version 8 or higher is required.Contact: califano@c2b2.columbia.eduSupplementary information: Supplementary data are available at Bioinformatics online.

An archive-based multi-objective evolutionary algorithm with adaptive search space partitioning to deal with expensive optimization problems: Application to process eco-design

In eco-design, the integration of environmental aspects into the earliest stage of design is considered with the aim of reducing adverse environmental impacts throughout a product's life cycle. An eco-design problem is therefore multi-objective, where several objectives (environmental, economic, and technological) are to be simultaneously optimized.The optimization of industrial processes usually requires solving expensive multi-objective optimization problems (MOPs). Aiming to solve efficiently MOPs, with a limited computational budget, this paper proposes a new framework called AMOEA-MAP. The framework relies on the structure of the NSGAII algorithm and possesses two novel operators: a memory-based adaptive partitioning strategy, which provides an adaptive reticulation of the search space for a quick identification of optimal zones with less computational effort; and a bi-population evolutionary algorithm, tailored for expensive optimization problems.To ascertain its generality, the framework is first tested on several tough benchmarks. Its performance is subsequently validated on a real-world eco-design problem.

Adaptive Partitioning for Irregular Applications on Heterogeneous CPU-GPU Chips

Commodity processors are comprised of several CPU cores and one integrated GPU. To fully exploit this type of architectures, one needs to automatically determine how to partition the workload between both devices. This is specially challenging for irregular workloads, where each iteration's work is data dependent and shows control and memory divergence. In this paper, we present a novel adaptive partitioning strategy specially designed for irregular applications running on heterogeneous CPU-GPU chips. The main novelty of this work is that the size of the workload assigned to the GPU and CPU adapts dynamically to maximize the GPU and CPU utilization while balancing the workload among the devices. Our experimental results on an Intel Haswell architecture using a set of irregular benchmarks show that our approach outperforms exhaustive static and adaptive state-of-the-art approaches in terms of performance and energy consumption.

A parallel adaptive nonlinear elimination preconditioned inexact Newton method for transonic full potential equation

We propose and study a right-preconditioned inexact Newton method for the numerical solution of large sparse nonlinear system of equations. The target applications are nonlinear problems whose derivatives have some local discontinuities such that the traditional inexact Newton method suffers from slow or no convergence even with globalization techniques. The proposed adaptive nonlinear elimination preconditioned inexact Newton method consists of three major ingredients: a subspace correction, a global update, and an adaptive partitioning strategy. The key idea is to remove the local high nonlinearity before performing the global Newton update. The partition used to define the subspace nonlinear problem is chosen adaptively based on the information derived from the intermediate Newton solution. Some numerical experiments are presented to demonstrate the robustness and efficiency of the algorithm compared to the classical inexact Newton method. Some parallel performance results obtained on a cluster of PCs are reported.

Hybrid stochastic simulations of intracellular reaction–diffusion systems

With the observation that stochasticity is important in biological systems, chemical kinetics have begun to receive wider interest. While the use of Monte Carlo discrete event simulations most accurately capture the variability of molecular species, they become computationally costly for complex reaction–diffusion systems with large populations of molecules. On the other hand, continuous time models are computationally efficient but they fail to capture any variability in the molecular species. In this study a hybrid stochastic approach is introduced for simulating reaction–diffusion systems. We developed an adaptive partitioning strategy in which processes with high frequency are simulated with deterministic rate-based equations, and those with low frequency using the exact stochastic algorithm of Gillespie. Therefore the stochastic behavior of cellular pathways is preserved while being able to apply it to large populations of molecules. We describe our method and demonstrate its accuracy and efficiency compared with the Gillespie algorithm for two different systems. First, a model of intracellular viral kinetics with two steady states and second, a compartmental model of the postsynaptic spine head for studying the dynamics of Ca+2 and NMDA receptors.