arXiv subjects

Siqiang Luo

Publications and source records attributed to Siqiang Luo.

51 records · Page 3Linked to original sources

arXiv2023

Learning to Optimize LSM-trees: Towards A Reinforcement Learning based Key-Value Store for Dynamic Workloads

LSM-trees are widely adopted as the storage backend of key-value stores. However, optimizing the system performance under dynamic workloads has not been sufficiently studied or evaluated in previous work. To fill the gap, we present RusKey, a key-value store with the following new features: (1) RusKey is a first attempt to orchestrate LSM-tree structures online to enable robust performance under the context of dynamic workloads; (2) RusKey is the first study to use Reinforcement Learning (RL) to guide LSM-tree transformations; (3) RusKey includes a new LSM-tree design, named FLSM-tree, for an efficient transition between different compaction policies -- the bottleneck of dynamic key-value stores. We justify the superiority of the new design with theoretical analysis; (4) RusKey requires no prior workload knowledge for system adjustment, in contrast to state-of-the-art techniques. Experiments show that RusKey exhibits strong performance robustness in diverse workloads, achieving up to 4x better end-to-end performance than the RocksDB system under various settings.

cs.DB↗

arXiv2023

A Survey on Machine Learning Solutions for Graph Pattern Extraction

A subgraph is constructed by using a subset of vertices and edges of a given graph. There exist many graph properties that are hereditary for subgraphs. Hence, researchers from different communities have paid a great deal of attention in studying numerous subgraph problems, on top of the ordinary graph problems. Many algorithms are proposed in studying subgraph problems, where one common approach is by extracting the patterns and structures of a given graph. Due to the complex structures of certain types of graphs and to improve overall performances of the existing frameworks, machine learning techniques have recently been employed in dealing with various subgraph problems. In this article, we present a comprehensive review on five well known subgraph problems that have been tackled by using machine learning methods. They are subgraph isomorphism (both counting and matching), maximum common subgraph, community detection and community search problems. We provide an outline of each proposed method, and examine its designs and performances. We also explore non-learning-based algorithms for each problem and a brief discussion is given. We then suggest some promising research directions in this area, hoping that relevant subgraph problems can be tackled by using a similar strategy. Since there is a huge growth in employing machine learning techniques in recent years, we believe that this survey will serve as a good reference point to relevant research communities.

math.CO↗

arXiv2023

River of No Return: Graph Percolation Embeddings for Efficient Knowledge Graph Reasoning

We study Graph Neural Networks (GNNs)-based embedding techniques for knowledge graph (KG) reasoning. For the first time, we link the path redundancy issue in the state-of-the-art KG reasoning models based on path encoding and message passing to the transformation error in model training, which brings us new theoretical insights into KG reasoning, as well as high efficacy in practice. On the theoretical side, we analyze the entropy of transformation error in KG paths and point out query-specific redundant paths causing entropy increases. These findings guide us to maintain the shortest paths and remove redundant paths for minimized-entropy message passing. To achieve this goal, on the practical side, we propose an efficient Graph Percolation Process motivated by the percolation model in Fluid Mechanics, and design a lightweight GNN-based KG reasoning framework called Graph Percolation Embeddings (GraPE). GraPE outperforms previous state-of-the-art methods in both transductive and inductive reasoning tasks while requiring fewer training parameters and less inference time.

cs.AI↗

arXiv2023

Massively Parallel Single-Source SimRanks in $o(\log n)$ Rounds

SimRank is one of the most fundamental measures that evaluate the structural similarity between two nodes in a graph and has been applied in a plethora of data management tasks. These tasks often involve single-source SimRank computation that evaluates the SimRank values between a source node $s$ and all other nodes. Due to its high computation complexity, single-source SimRank computation for large graphs is notoriously challenging, and hence recent studies resort to distributed processing. To our surprise, although SimRank has been widely adopted for two decades, theoretical aspects of distributed SimRanks with provable results have rarely been studied. In this paper, we conduct a theoretical study on single-source SimRank computation in the Massive Parallel Computation (MPC) model, which is the standard theoretical framework modeling distributed systems such as MapReduce, Hadoop, or Spark. Existing distributed SimRank algorithms enforce either $Ω(\log n)$ communication round complexity or $Ω(n)$ machine space for a graph of $n$ nodes. We overcome this barrier. Particularly, given a graph of $n$ nodes, for any query node $v$ and constant error $ε>\frac{3}{n}$, we show that using $O(\log^2 \log n)$ rounds of communication among machines is almost enough to compute single-source SimRank values with at most $ε$ absolute errors, while each machine only needs a space sub-linear to $n$. To the best of our knowledge, this is the first single-source SimRank algorithm in MPC that can overcome the $Θ(\log n)$ round complexity barrier with provable result accuracy.

cs.DC↗

arXiv2023

Less is More: Towards Lightweight Cost Estimator for Database Systems

We present FasCo, a simple yet effective learning-based estimator for the cost of executing a database query plan. FasCo uses significantly shorter training time and a lower inference cost than the state-of-the-art approaches, while achieving higher estimation accuracy. The effectiveness of FasCo comes from embedding abundant explicit execution-plan-based features and incorporating a novel technique called cardinality calibration. Extensive experimental results show that FasCo achieves orders of magnitude higher efficiency than the state-of-the-art methods: on the JOB-M benchmark dataset, it cuts off training plans by 98\%, reducing training time from more than two days to about eight minutes while entailing better accuracy. Furthermore, in dynamic environments, FasCo can maintain satisfactory accuracy even without retraining, narrowing the gap between learning-based estimators and real systems.

cs.DB↗

arXiv2022

Spiking Graph Convolutional Networks

Graph Convolutional Networks (GCNs) achieve an impressive performance due to the remarkable representation ability in learning the graph information. However, GCNs, when implemented on a deep network, require expensive computation power, making them difficult to be deployed on battery-powered devices. In contrast, Spiking Neural Networks (SNNs), which perform a bio-fidelity inference process, offer an energy-efficient neural architecture. In this work, we propose SpikingGCN, an end-to-end framework that aims to integrate the embedding of GCNs with the biofidelity characteristics of SNNs. The original graph data are encoded into spike trains based on the incorporation of graph convolution. We further model biological information processing by utilizing a fully connected layer combined with neuron nodes. In a wide range of scenarios (e.g. citation networks, image graph classification, and recommender systems), our experimental results show that the proposed method could gain competitive performance against state-of-the-art approaches. Furthermore, we show that SpikingGCN on a neuromorphic chip can bring a clear advantage of energy efficiency into graph data analysis, which demonstrates its great potential to construct environment-friendly machine learning models.

cs.LG↗

arXiv2022

SCARA: Scalable Graph Neural Networks with Feature-Oriented Optimization

Recent advances in data processing have stimulated the demand for learning graphs of very large scales. Graph Neural Networks (GNNs), being an emerging and powerful approach in solving graph learning tasks, are known to be difficult to scale up. Most scalable models apply node-based techniques in simplifying the expensive graph message-passing propagation procedure of GNN. However, we find such acceleration insufficient when applied to million- or even billion-scale graphs. In this work, we propose SCARA, a scalable GNN with feature-oriented optimization for graph computation. SCARA efficiently computes graph embedding from node features, and further selects and reuses feature computation results to reduce overhead. Theoretical analysis indicates that our model achieves sub-linear time complexity with a guaranteed precision in propagation process as well as GNN training and inference. We conduct extensive experiments on various datasets to evaluate the efficacy and efficiency of SCARA. Performance comparison with baselines shows that SCARA can reach up to 100x graph propagation acceleration than current state-of-the-art methods with fast convergence and comparable accuracy. Most notably, it is efficient to process precomputation on the largest available billion-scale GNN dataset Papers100M (111M nodes, 1.6B edges) in 100 seconds.

cs.LG↗

arXiv2022

Proteus: A Self-Designing Range Filter

We introduce Proteus, a novel self-designing approximate range filter, which configures itself based on sampled data in order to optimize its false positive rate (FPR) for a given space requirement. Proteus unifies the probabilistic and deterministic design spaces of state-of-the-art range filters to achieve robust performance across a larger variety of use cases. At the core of Proteus lies our Contextual Prefix FPR (CPFPR) model - a formal framework for the FPR of prefix-based filters across their design spaces. We empirically demonstrate the accuracy of our model and Proteus' ability to optimize over both synthetic workloads and real-world datasets. We further evaluate Proteus in RocksDB and show that it is able to improve end-to-end performance by as much as 5.3x over more brittle state-of-the-art methods such as SuRF and Rosetta. Our experiments also indicate that the cost of modeling is not significant compared to the end-to-end performance gains and that Proteus is robust to workload shifts.

cs.DB↗

arXiv2022

Finding Global Homophily in Graph Neural Networks When Meeting Heterophily

We investigate graph neural networks on graphs with heterophily. Some existing methods amplify a node's neighborhood with multi-hop neighbors to include more nodes with homophily. However, it is a significant challenge to set personalized neighborhood sizes for different nodes. Further, for other homophilous nodes excluded in the neighborhood, they are ignored for information aggregation. To address these problems, we propose two models GloGNN and GloGNN++, which generate a node's embedding by aggregating information from global nodes in the graph. In each layer, both models learn a coefficient matrix to capture the correlations between nodes, based on which neighborhood aggregation is performed. The coefficient matrix allows signed values and is derived from an optimization problem that has a closed-form solution. We further accelerate neighborhood aggregation and derive a linear time complexity. We theoretically explain the models' effectiveness by proving that both the coefficient matrix and the generated node embedding matrix have the desired grouping effect. We conduct extensive experiments to compare our models against 11 other competitors on 15 benchmark datasets in a wide range of domains, scales and graph heterophilies. Experimental results show that our methods achieve superior performance and are also very efficient.

cs.LG↗

arXiv2022

DMCS : Density Modularity based Community Search

Community Search, or finding a connected subgraph (known as a community) containing the given query nodes in a social network, is a fundamental problem. Most of the existing community search models only focus on the internal cohesiveness of a community. However, a high-quality community often has high modularity, which means dense connections inside communities and sparse connections to the nodes outside the community. In this paper, we conduct a pioneer study on searching a community with high modularity. We point out that while modularity has been popularly used in community detection (without query nodes), it has not been adopted for community search, surprisingly, and its application in community search (related to query nodes) brings in new challenges. We address these challenges by designing a new graph modularity function named Density Modularity. To the best of our knowledge, this is the first work on the community search problem using graph modularity. The community search based on the density modularity, termed as DMCS, is to find a community in a social network that contains all the query nodes and has high density-modularity. We prove that the DMCS problem is NP-hard. To efficiently address DMCS, we present new algorithms that run in log-linear time to the graph size. We conduct extensive experimental studies in real-world and synthetic networks, which offer insights into the efficiency and effectiveness of our algorithms. In particular, our algorithm achieves up to 8.5 times higher accuracy in terms of NMI than baseline algorithms.

cs.SI↗

arXiv2019

Semiclassical analysis of photoelectron interference in a synthesized two-color laser pulse

We measure the photoelectron energy spectra from strong-field ionization of Kr in a two-color laser pulse consisting of a strong 400-nm field and a weak 800-nm field. The intensities of the main above-threshold ionization (ATI) and sideband peaks in the photoelectron energy spectra oscillate roughly oppositely with respect to the relative phase between the two-color components. We study the photoelectron interferometry in strong-field ATI regime from the view of interference of different electron trajectories in order to extend RABBITT type analysis to the strong-field regime. Based on the strong-field approximation model, we obtain analytical expressions for the oscillations of both ATI and sideband peaks with the relative phase. A phase shift of π/4 with respect to the field maximum of the two-color laser pulse is revealed for the interference maximum in the main ATI peak without including the effect of the atomic potential.

physics.atom-ph↗

arXiv2019

Exit momentum and instantaneous ionization rate of nonadiabatic tunneling ionization in elliptically polarized laser fields

Based on the strong-field approximation, we obtain analytical expressions for the initial momentum at the tunnel exit and instantaneous ionization rate of tunneling ionization in elliptically polarized laser fields with arbitrary ellipticity. The tunneling electron reveals a nonzero offset of the initial momentum at the tunnel exit in the elliptically polarized laser field. We find that the transverse and longitudinal components of this momentum offset with respect to the instantaneous field direction are directly related to the time derivatives of the instantaneous laser electric field along the angular and radial directions, respectively. We further show that the nonzero initial momentum at the tunnel exit has a significant influence on the laser phase dependence of the instantaneous ionization rate in the nonadiabatic tunneling regime.

physics.atom-ph↗

arXiv2019

Photoelectron holographic interferometry to probe the longitudinal momentum offset at the tunnel exit

Laser-induced electron tunneling underlies numerous emerging spectroscopic techniques to probe attosecond electron dynamics in atoms and molecules. The improvement of those techniques requires an accurate knowledge of the exit momentum for the tunneling wave packet. Here we demonstrate a photoelectron interferometric scheme to probe the electron momentum longitudinal to the tunnel direction at the tunnel exit by measuring the photoelectron holographic pattern in an orthogonally polarized two-color laser pulse. In this scheme, we use a perturbative 400-nm laser field to modulate the photoelectron holographic fringes generated by a strong 800-nm pulse. The fringe shift offers a direct experimental access to the intermediate canonical momentum of the rescattering electron, allowing us to reconstruct the momentum offset at the tunnel exit with high accuracy. Our result unambiguously proves the existence of nonzero initial longitudinal momentum at the tunnel exit and provides fundamental insights into the non-quasi-static nature of the strong-field tunneling.

physics.atom-ph↗

arXiv2015

Selective enhancement of resonant multiphoton ionization with strong laser fields

High-resolution photoelectron momentum distributions of Xe atoms ionized by 800-nm linearly polarized laser fields have been traced at intensities from 1.1*1013 to 3.5*1013W/cm2 using velocity-map imaging techniques. At certain laser intensities, the momentum spectrum exhibits a distinct double-ring structure for low-order above-threshold ionization, which appears to be absent at lower or higher laser intensities. By investigating the intensity-resolved photoelectron energy spectrum, we find that this double-ring structure originates from resonant multiphoton ionization involving multiple Rydberg states of atoms. Varying the laser intensity, we can selectively enhance the resonant multiphoton ionization through certain atomic Rydberg states. The photoelectron angular distributions of multiphoton resonance are also investigated for the low-order above-threshold ionization.

physics.atom-ph↗

arXiv2013

Shortest Path and Distance Queries on Road Networks: Towards Bridging Theory and Practice

Given two locations $s$ and $t$ in a road network, a distance query returns the minimum network distance from $s$ to $t$, while a shortest path query computes the actual route that achieves the minimum distance. These two types of queries find important applications in practice, and a plethora of solutions have been proposed in past few decades. The existing solutions, however, are optimized for either practical or asymptotic performance, but not both. In particular, the techniques with enhanced practical efficiency are mostly heuristic-based, and they offer unattractive worst-case guarantees in terms of space and time. On the other hand, the methods that are worst-case efficient often entail prohibitive preprocessing or space overheads, which render them inapplicable for the large road networks (with millions of nodes) commonly used in modern map applications. This paper presents {\em Arterial Hierarchy (AH)}, an index structure that narrows the gap between theory and practice in answering shortest path and distance queries on road networks. On the theoretical side, we show that, under a realistic assumption, AH answers any distance query in $\tilde{O}(\log \r)$ time, where $\r = d_{max}/d_{min}$, and $d_{max}$ (resp.\ $d_{min}$) is the largest (resp.\ smallest) $L_\infty$ distance between any two nodes in the road network. In addition, any shortest path query can be answered in $\tilde{O}(k + \log \r)$ time, where $k$ is the number of nodes on the shortest path. On the practical side, we experimentally evaluate AH on a large set of real road networks with up to twenty million nodes, and we demonstrate that (i) AH outperforms the state of the art in terms of query time, and (ii) its space and pre-computation overheads are moderate.

cs.DS↗