Graph Search Research Articles

Automatic computation of bending sequences for wire bending machines

ABSTRACT Determining a feasible bending sequence, i.e., ensuring absence of wire self-collisions and wire-machine collisions, or even an optimal bending sequence, i.e., minimising time or energy required to perform it, is a difficult and time-consuming task for complex workpieces, even for expert operators. Introducting algorithms for the computation of wire-bending sequences is thus crucial to increase productivity and production flexibility, and to decrease production costs. This work proposes an algorithm to automatically determine an appropriate bending sequence for a given workpiece, bending tool, and machine 3D CAD model, which leverages on a representation of the wire as a robotic manipulator and of a bending sequence as a tree, and on the adoption of A* as graph search algorithm. A cost and a heuristic function, suitable for the wire-bending problem, and an approach to parallelise the execution of A* are introduced, as well. In this way, a computationally simple and efficient wire-bending sequence computation algorithm is devised, able to determine a solution in an amount of time less or equal to the time used by an expert operator, without the need of high computational power. The effectiveness of this algorithm is assessed on two different test cases, relevant to industrial workpieces.

OCT Retinal and Choroidal Layer Instance Segmentation Using Mask R-CNN.

Optical coherence tomography (OCT) of the posterior segment of the eye provides high-resolution cross-sectional images that allow visualization of individual layers of the posterior eye tissue (the retina and choroid), facilitating the diagnosis and monitoring of ocular diseases and abnormalities. The manual analysis of retinal OCT images is a time-consuming task; therefore, the development of automatic image analysis methods is important for both research and clinical applications. In recent years, deep learning methods have emerged as an alternative method to perform this segmentation task. A large number of the proposed segmentation methods in the literature focus on the use of encoder–decoder architectures, such as U-Net, while other architectural modalities have not received as much attention. In this study, the application of an instance segmentation method based on region proposal architecture, called the Mask R-CNN, is explored in depth in the context of retinal OCT image segmentation. The importance of adequate hyper-parameter selection is examined, and the performance is compared with commonly used techniques. The Mask R-CNN provides a suitable method for the segmentation of OCT images with low segmentation boundary errors and high Dice coefficients, with segmentation performance comparable with the commonly used U-Net method. The Mask R-CNN has the advantage of a simpler extraction of the boundary positions, especially avoiding the need for a time-consuming graph search method to extract boundaries, which reduces the inference time by 2.5 times compared to U-Net, while segmenting seven retinal layers.

Fault Tolerant Depth First Search in Undirected Graphs: Simple Yet Efficient

Fault Tolerant Depth First Search in Undirected Graphs: Simple Yet Efficient

Clique Search in Graphs of Special Class and Job Shop Scheduling

In this paper, we single out the following particular case of the clique search problem. The vertices of the given graph are legally colored with k colors and we are looking for a clique with k nodes in the graph. In other words, we want to decide if a given k-partite graph contains a clique with k nodes. The maximum clique problem asks for finding a maximum clique in a given finite simple graph. The problem of deciding if the given graph contains a clique with k vertices is called the k-clique problem. The first problem is NP-hard and the second one is NP-complete. The special clique search problem, we propose, is still an NP-complete problem. We will show that the k-clique problem in the special case of k-partite graphs is more tractable than in the general case. In order to illustrate the possible practical utility of this restricted type clique search problem we will show that the job shop scheduling problem can be reduced to such a clique search problem in a suitable constructed graph. We carry out numerical experiments to assess the efficiency of the approach. It is a common practice that before one embarks on a large scale clique search typically one attempts to simplify and tidy up the given graph. This procedure is commonly referred as preconditioning or kernelization of the given graph. Of course, the preconditioning or kernelization is meant with respect to the given type of clique search problem. The other main topic of the paper is to describe a number of kernelization methods tailored particularly to the proposed special k-clique problem. Some of these techniques works in connection with the generic k-clique problem. In these situations, we will see that they are more efficient in the case of k-partite graphs. Some other preconditioning methods applicable only to k-partite graphs. We illustrate how expedient these preconditioning methods can be by solving non-trivial scheduling problems to optimality employing only kernelization techniques dispensing with exhaustive clique search algorithms altogether.

Cross-Border E-Commerce Intelligent Information Recommendation System Based on Deep Learning.

In order to improve the effect of cross-border e-commerce intelligent information recommendation, this paper applies deep learning to the intelligent information processing and intelligent recommendation of e-commerce and proposes an improved version of the topic model to solve the problem of feature extraction of the text of the recommendation system. In order to deal with translation problems, this paper proposes an end-to-end sequence-to-sequence learning method. In addition, this study uses the long tail theory to excavate the mass commodities in the niche and recommends these products to users as suggestions. Finally, this paper proposes a niche product recommendation algorithm based on the graph search strategy based on the graph model. The experiment shows that the cross-border e-commerce intelligent information recommendation system based on deep learning proposed in this paper has a good recommendation effect and meets the recommendation needs of cross-border e-commerce.

Coordinated Multi-Robotic Vehicles Navigation and Control in Shop Floor Automation.

In this paper, we propose a global navigation function applied to model predictive control (MPC) for autonomous mobile robots, with application to warehouse automation. The approach considers static and dynamic obstacles and generates smooth, collision-free trajectories. The navigation function is based on a potential field derived from an E* graph search algorithm on a discrete occupancy grid and by bicubic interpolation. It has convergent behavior from anywhere to the target and is computed in advance to increase computational efficiency. The novel optimization strategy used in MPC combines a discrete set of velocity candidates with randomly perturbed candidates from particle swarm optimization. Adaptive horizon length is used to improve performance. The efficiency of the proposed approaches is validated using simulations and experimental results.

Accelerating SSSP for Power-Law Graphs.

The single-source shortest path (SSSP) problem is one of the most important and well-studied graph problems widely used in many application domains, such as road navigation, neural image reconstruction, and social network analysis. Although we have known various SSSP algorithms for decades, implementing one for large-scale power-law graphs efficiently is still highly challenging today, because ① a work-efficient SSSP algorithm requires priority-order traversal of graph data, ② the priority queue needs to be scalable both in throughput and capacity, and ③ priority-order traversal requires extensive random memory accesses on graph data. In this paper, we present SPLAG to accelerate SSSP for power-law graphs on FPGAs. SPLAG uses a coarse-grained priority queue (CGPQ) to enable high-throughput priority-order graph traversal with a large frontier. To mitigate the high-volume random accesses, SPLAG employs a customized vertex cache (CVC) to reduce off-chip memory access and improve the throughput to read and update vertex data. Experimental results on various synthetic and real-world datasets show up to a 4.9× speedup over state-of-the-art SSSP accelerators, a 2.6× speedup over 32-thread CPU running at 4.4 GHz, and a 0.9× speedup over an A100 GPU that has 4.1× power budget and 3.4× HBM bandwidth. Such a high performance would place SPLAG in the 14th position of the Graph 500 benchmark for data intensive applications (the highest using a single FPGA) with only a 45 W power budget. SPLAG is written in high-level synthesis C++ and is fully parameterized, which means it can be easily ported to various different FPGAs with different configurations. SPLAG is open-source at https://github.com/UCLA-VAST/splag.

Autonomous navigation of ships by combining optimal trajectory planning with informed graph search

ABSTRACT Autonomous trajectory generation plays an essential role in the navigation of vehicles in space as well as in terrestrial scenarios, i.e. in the air, on solid ground, or water. For the latter, the navigation of ships in ports has specific challenges since ship dynamics are highly nonlinear with limited agility, while the manoeuvre space in ports is limited. Nevertheless, for providing support to humanly designed control strategies, autonomously generated trajectories have not only to be feasible, i.e. collision-free but shall also be optimal with respect to manoeuvre time and control effort. This article presents a novel approach to autonomous trajectory planning on the basis of precomputed and connectable trajectory segments, the so-called motion primitives, and an A*-search algorithm. Sequences of motion primitives provide an initial guess for a subsequent optimization by which optimal trajectories are found even in terrains with many obstacles. We illustrate the approach with different navigation scenarios.

HiveMind: Towards Cellular Native Machine Learning Model Splitting

The increasing processing load of today’s mobile machine learning (ML) application challenges the stringent computation budget of mobile user equipment (UE). With the wide deployment of 5G edge-cloud, a new ML offloading scheme called split ML is provisioned to enable computation-intensive mobile ML applications by splitting an ML model across mobile UE, edge, and cloud. However, the complex split assignment problems pose new challenges for split ML system design. In this paper, we introduce HiveMind, the first practical multi-split ML system tailored for 5G cellular networks. HiveMind reformulates the complicated multi-split problem to a min-cost graph search and optimizes the distributed algorithm to drastically reduce the signaling overhead. Benefit from its low overhead property, HiveMind makes the optimal split decision on multiple computing nodes in real-time and adapts the split decisions to the instantaneous network dynamics. HiveMind also incorporates a multi-objective mechanism that accommodates heterogeneous objectives for a single ML task. HiveMind adapts to a wide range of ML frameworks, including non-linear models like Recurrent Neural Network (RNN), Federated Learning (FL), and Multi-agent Reinforcement Learning (MARL). We evaluate HiveMind on 5G MEC network simulators with realistic traffic patterns and real-life MEC computation/communication profiles. Our experiments demonstrate that HiveMind achieves the optimal efficiency comparing to state-of-art split ML designs.

An Efficient Full Hardware Implementation of Extended Merkle Signature Scheme

This paper presents a full hardware implementation of the eXtended Merkle Signature Scheme (XMSS), a NIST approved and IETF RFC specified post-quantum cryptography (PQC) algorithm. An optimized node traversal is proposed to enable efficient memory utilization without compromising the computational latency of the L-tree and Merkle tree construction, which are two key components used for the compression of the Winternitz One-Time Signature (WOTS) public key in XMSS. The computation of the authentication path during signature generation has also been significantly sped up by our proposed hardware implementation of the Buchmann, Dahmen, and Schneider (BDS) algorithm. Our implementation has completely avoided the use of block random-access memory, which is known to be vulnerable to side-channel attacks. The memory requirement has been highly optimized for implementation with small flip-flop chains and register counters as pointers for fast data access. To the best of our knowledge, this is the first full hardware implementation of all three <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">key generation</i> , <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">signing</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">verification</i> operations of XMSS. The design has been prototyped and evaluated on a 28 nm FPGA platform to demonstrate its performance improvements over the most efficient software and hardware/software co-design methods reported to date. Specifically, it increases the computational efficiency of the best reported XMSS implementation for <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">key generation</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">signature generation</i> by about 20% and 50%, respectively. It can also run at 10% higher clock speed than the fastest hardware implementation of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">signature verification</i> in FPGA with 8% lower hardware resource utilization.

Version‐based and risk‐enabled testing, monitoring, and visualization of microservice systems

AbstractDespite the growing importance of the microservice architecture (MSA), interactions among the various elements (e.g., services, endpoints, and versions) remain difficult to manage. This research devised a system, called version‐based microservice analysis, monitoring, and visualization (VMAMV), including multiple proposed methods to facilitate the testing, monitoring, and visualization of microservice systems by considering service versions and risks. VMAMV can generate version‐based service dependency graphs, provide graph search services, perform consumer‐driven contract testing, and conduct a risk analysis. Besides, VMAMV can also notify users of existing anomalies and potential risks to facilitate the development and operation of microservice systems. Experiments were conducted to demonstrate the efficacy of the VMAMV system in the detection of service anomalies, old patch versions, low‐usage service versions, the presentation of contract test results and service error chains, and the validity of risk analysis.

Оптимизация выполнения целочисленных арифметических операций на основе нейронных сетей

Nowadays, the main methods of improving the performance of arithmetic operations on fixed and floating numbers are performed by multicore computer systems (CS) are the improvement of the microarchitecture of systems and algorithms implemented in compilers and libraries. Separately, it is necessary to highlight the complex tasks of implementing operations with long numbers that exceed the size of a machine word. This is due to the limited size of the processor registers and the data bus. As a result, the implementation of traditional operations, such as addition, multiplication, division, become more difficult to implement in modern computer systems. To solve this problem, there are currently several algorithms. However, the available algorithms can be quite slow, since they are sequential and implement bitwise operations on the bits of a number.In order to improve the performance of this class of calculations, a model based on a neural network is proposed. The purpose of this model is to reduce the execution time and CPU usage. A neural network model for performing arithmetic operations has been developed. A method of generating a data set based on graph traversal is proposed and tested experimentally. Algorithms for performing arithmetic operations based on a pretrained neural network are constructed.

Automatic 3-D segmentation and volumetric light fluence correction for photoacoustic tomography based on optimal 3-D graph search.

Preclinical imaging with photoacoustic tomography (PAT) has attracted wide attention in recent years since it is capable of providing molecular contrast with deep imaging depth. The automatic extraction and segmentation of the animal in PAT images is crucial for improving image analysis efficiency and enabling advanced image post-processing, such as light fluence (LF) correction for quantitative PAT imaging. Previous automatic segmentation methods are mostly two-dimensional approaches, which failed to conserve the 3-D surface continuity because the image slices were processed separately. This discontinuity problem further hampers LF correction, which, ideally, should be carried out in 3-D due to spatially diffused illumination. Here, to solve these problems, we propose a volumetric auto-segmentation method for small animal PAT imaging based on the 3-D optimal graph search (3-D GS) algorithm. The 3-D GS algorithm takes into account the relation among image slices by constructing a 3-D node-weighted directed graph, and thus ensures surface continuity. In view of the characteristics of PAT images, we improve the original 3-D GS algorithm on graph construction, solution guidance and cost assignment, such that the accuracy and smoothness of the segmented animal surface were guaranteed. We tested the performance of the proposed method by conducting in vivo nude mice imaging experiments with a commercial preclinical cross-sectional PAT system. The results showed that our method successfully retained the continuous global surface structure of the whole 3-D animal body, as well as smooth local subcutaneous tumor boundaries at different development stages. Moreover, based on the 3-D segmentation result, we were able to simulate volumetric LF distribution of the entire animal body and obtained LF corrected PAT images with enhanced structural visibility and uniform image intensity.

Analyzing GCN Aggregation on GPU

Graph convolutional neural networks (GCNs) are emerging neural networks for graph structures that include large features associated with each vertex. The operations of GCN can be divided into two phases - aggregation and combination. While the combination just performs matrix multiplications using trained weights and aggregated features, the aggregation phase requires graph traversal to collect features from adjacent vertices. Even though neural network applications rely on GPU’s massively parallel processing, GCN aggregation kernels exhibit rather low performance since graph processing using compressed graph structures provokes frequent irregular accesses in GPUs. In order to investigate the performance hurdles of GCN aggregation on GPU, we perform an in-depth analysis of the aggregation kernels using real GPU hardware and a cycle-accurate GPU simulator. We first analyze the characteristics of the popular graph datasets used for GCN studies. We reveal the fractions of non-zero elements in feature vectors are diverse among datasets. Based on the observation, we build two types of aggregation kernels that handle uncompressed and compressed feature vectors. Our evaluation exhibits the performance of aggregation can be significantly influenced by kernel design approaches and feature density. We also analyze the individual loads that access the data arrays of the aggregation kernels to specify critical loads. Our analysis reveals the performance of GPU memory hierarchy is influenced by access patterns and feature size of graph datasets. Based on our observations we discuss possible kernel design approaches and architectural ideas that can improve the performance of GCN aggregation.

The concept of an intelligent data lake management system: machine consciousness and a universal data model

The concept of a data lake is proposed, the management system of which has a machine consciousness to control the parameters of the data lake, resolve independently from emergency situations and interact with administrators. For data storage and processing, a universal data model is used that allows processing data in any form: tables, multidimensional cubes, graphs, complex graphs (hypergraphs, metagraphs, etc.), what provides additional possibilities for information search and transformation. The specially configured relationships in the data lake allows to see the history of data transformation. The universal data model foresees usage of a family of languages that implement the capabilities of working with metagraphs, including declarative and navigational query and data manipulation languages, which will have built-in SQL and MDX for working with tables and cubes. The subconscious performs regular standard procedures, such as the data lake monitoring, identifying trends and abnormal situations. The subconscious uses the AutoML system to predict the characteristics of the data lake. The data lake management system uses special agents that are looking for missing information in the data lake in external data sources (the Internet, data bases and data warehouses), and special bots for interacting with administrators and clients in natural language (they must be multimodal in order to be able to work with diagrams and speech). To ensure integration with data manipulation languages and effective fulfillment of user's requests, the data lake management system includes a library of classical mathematical algorithms for working with graphs adapted for use on metagraphs (for example, shortest path algorithms, graph traversal algorithms, etc.). The potential implementation of the data lake management system is wider than just providing analytics. The data lake under management of the proposed system can be used for transactional data processing. Therefore, the data lake may become the only data storage system in organization, on the top of which all applications of the corporate management system are working. The implementation of the proposed concept will eliminate the backlog in the development of the Data Lake architecture from the architecture of enterprise management systems.

Toward an Adaptive Skip-Gram Model for Network Representation Learning

The random walk process on network data is a widely-used approach for network representation learning. However, we argue that the sampling of node sequences and the subsampling for the Skip-gram’s contexts have two drawbacks. One is less possible to precisely find the most correlated context nodes for every central node with only uniform graph search. The other is not easily controlled due to the expensive cost of hyperparameter tuning. Such two drawbacks lead to higher training cost and lower accuracy due to abundant and irrelevant samples. To solve these problems, we compute the adaptive probability of random walk based on Personalized PageRank (PPR), and propose an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Adaptive SKip-gram</i> (ASK) model without using complicated sampling process and negative sampling. We utilize <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> -most important neighbors for positive samples selection, and attach their corresponding PPR probability into the objective function. Based on benchmark datasets with three citation networks and three social networks, we demonstrate the improvement of our ASK model for network representation learning in tasks of link prediction, node classification, and embedding visualization. The results achieve more effective performance and efficient learning time.

Two-Stage Vehicle Routing Optimization for Logistics Distribution Based on HSA-HGBS Algorithm

Aiming at the problems of complex urban road network, low efficiency of logistics distribution, and the difficulty of large-scale logistics distribution area division and routing planning, this paper proposes a two-stage logistics distribution vehicle routing optimization (VRP) method based on the establishment of a multi-factor complex road network constrained logistics distribution mathematical model. Considering the complex traffic elements and road network topological structure in logistics and distribution, in the first stage, a heuristic simulated annealing (HSA) distribution region partitioning algorithm is proposed with the objective of balancing vehicle task load to divide the urban logistics distribution network under complex road networks, so as to reduce the area scale and path search cost. In the second stage of route decision making, aiming at minimizing the total cost of logistics distribution, combining the VRP problem with complex road network conditions, a heuristic path search method combined with complex road network model constraints is proposed. In this stage, a hybrid genetic beam search(HGBS) algorithm is used to plan the path nodes, reduce the randomness of the model in the initial search for paths by heuristic genetic algorithms, then combine with Beam graph search methods to reduce the space and time used for the search, and use optimization algorithms to improve the accuracy of independent sub-region routing optimization and the rationality of overall physical distribution route selection. Finally, the proposed method is validated in this paper with two practical cases. The experimental results show that the two-stage decision-making algorithm proposed in this paper has certain advantages in partitioning schemes, minimizing total cost and iteration times. Through comparison, the optimization ability of this method for logistics distribution networks is proved.

Generalized Path Planning for UTM Systems With a Space-Time Graph

Motivated by the increased use of UAS in commercial applications, in this paper we tackle the problem of path planning when requests are submitted by UAS managed by different operators. We propose the new problem of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">generalized path planning</i> for UAS Traffic Management, where the UAS path is described by operators with a sequence of waypoint groups and a solution trajectory must pass through a waypoint in each group. This problem is typical for applications where multiple charging stations and pickup/drop-off locations are distributed in a flight area. Our solution builds upon prior work on discretized space-time graph path planning and proposes a novel multi-source/multi-destination graph search algorithm that generates collision-free trajectories for pre-flight CDR. Our efficient algorithm has runtime proportional to the number of groups and avoids combinatorial explosion. We apply our mechanism to the energy-constrained UAS package delivery problem with multiple warehouses and battery charging stations. Simulation results show that our algorithm is efficient and scalable with the number of requests and graph size. The addition of charging stations and the option for multiple warehouses increases the request admission ratio and reduces the overall trajectory duration, effectively improving both the planner’s quality of service and the efficiency of airspace usage.

Building and Applying Knowledge Graph in Edge Analytics Environment

For the scenario of limited hardware resources and restricted software environment in edge computing architecture, the method of building and applying knowledge graph in edge analytics environment is proposed. The main process includes: building knowledge graph in the cloud, storing knowledge base with RDF format at the edge through customization and transformation, and performing query and analytics at the edge with SPARQL graph search language. The method is simulated in a communication anti-jamming test environment, and the results show that the relevant technical solutions can better meet the requirements of construction of knowledge graph and data analysis in a restricted edge analytics environment.

Interactive Visual Pattern Search on Graph Data via Graph Representation Learning.

Graphs are a ubiquitous data structure to model processes and relations in a wide range of domains. Examples include control-flow graphs in programs and semantic scene graphs in images. Identifying subgraph patterns in graphs is an important approach to understand their structural properties. We propose a visual analytics system GraphQ to support human-in-the-loop, example-based, subgraph pattern search in a database containing many individual graphs. To support fast, interactive queries, we use graph neural networks (GNNs) to encode a graph as fixed-length latent vector representation, and perform subgraph matching in the latent space. Due to the complexity of the problem, it is still difficult to obtain accurate one-to-one node correspondences in the matching results that are crucial for visualization and interpretation. We, therefore, propose a novel GNN for node-alignment called NeuroAlign, to facilitate easy validation and interpretation of the query results. GraphQ provides a visual query interface with a query editor and a multi-scale visualization of the results, as well as a user feedback mechanism for refining the results with additional constraints. We demonstrate GraphQ through two example usage scenarios: analyzing reusable subroutines in program workflows and semantic scene graph search in images. Quantitative experiments show that NeuroAlign achieves 19%-29% improvement in node-alignment accuracy compared to baseline GNN and provides up to 100× speedup compared to combinatorial algorithms. Our qualitative study with domain experts confirms the effectiveness for both usage scenarios.