Search

Entity resolution is an important data association task when fusing information from multiple sources. Oftentimes the information arrives continuously and the entity resolution algorithm needs to efficiently update its solution upon receiving new information. In this work, we introduce an incremental entity resolution algorithm based on a graph partitioning formulation. The developed algorithm is able to handle both incrementally arriving entity references, as well as incrementally arriving information which changes the pairwise similarity scores between the references. New information is handled in a way that allows the algorithm to reconsider past decisions when contradicting information arrives. Because the graph partitioning formulation used is NP-Hard, a heuristic algorithm is developed to produce good solutions, which is also compatible with a blocking technique to limit the number of required comparisons. The algorithm is tested on a variety of datasets (randomly generated and real) and it is shown that allowing the algorithm to consider revised scores and revisit prior decisions offers a substantial improvement to accuracy (approximately 30–40% better F-Score on a natural language dataset), compared to other greedy heuristics on the same set of coefficients. It is also shown that, on a test set with 100 references, the incremental algorithm is up to an order of magnitude faster than a batch algorithm approach that re-solves the entire problem.

Linear Assignment is one of the most fundamental problems in operations research.A creative parallelization of a Hungarian-like algorithm on GPU cluster.Efficient parallelization of the augmenting path search step.Large problems with 1.6 billion variables can be solved.It is probably the fastest LAP solver using a GPU. In this paper, we describe parallel versions of two different variants (classical and alternating tree) of the Hungarian algorithm for solving the Linear Assignment Problem (LAP). We have chosen Compute Unified Device Architecture (CUDA) enabled NVIDIA Graphics Processing Units (GPU) as the parallel programming architecture because of its ability to perform intense computations on arrays and matrices. The main contribution of this paper is an efficient parallelization of the augmenting path search phase of the Hungarian algorithm. Computational experiments on problems with up to 25 million variables reveal that the GPU-accelerated versions are extremely efficient in solving large problems, as compared to their CPU counterparts. Tremendous parallel speedups are achieved for problems with up to 400 million variables, which are solved within 13 seconds on average. We also tested multi-GPU versions of the two variants on up to 16 GPUs, which show decent scaling behavior for problems with up to 1.6 billion variables and dense cost matrix structure.

In this paper, we describe a GPU-accelerated parallel algorithm for the axial Multidimensional Assignment Problem with Decomposable Costs (MDADC), which is one of the most fundamental formulations for data association. MDADC is known to be NP-hard and is large-dimensioned in most realistic cases; hence, heuristic solutions with qualified optimality gaps is the best one can hope for, given the state-of-knowledge. The main contribution of this paper is an efficient parallelization of the Lagrangian subgradient search algorithm specifically targeted towards the Graphics Processing Units (GPUs) based on the NVIDIA Compute Unified Device Architecture (CUDA). A GPU-accelerated Linear Assignment Problem (LAP) solver is leveraged in concert with the Lagrangian scheme for further speed-up. We also implemented a multi-GPU variant of this algorithm which maintains a good speedup profile, when tested on problems with 31 billion variables, on up to 128 GPUs.

In this paper, we present an update to our previous submission from Graph Challenge 2017. This work describes and evaluates new software algorithm optimizations undertaken for our 2018 year submission on Collaborative CPU+GPU Algorithms for Triangle Counting and Truss Decomposition. First, we describe four major optimizations for the triangle counting which improved performance by up to 117x over our prior submission. Additionally, we show that our triangle-counting algorithm is on average 151.7x faster than NVIDIA's NVGraph library (max 476x) for SNAP datasets. Second, we propose a novel parallel k-truss decomposition algorithm that is time-efficient and is up to 13.9x faster than our previous submission. Third, we evaluate the effect of generational hardware improvements between the IBM “Minsky” (POWER8, P100, NVLink 1.0) and “Newell” (POWER9, V100, NVLink 2.0) platforms. Lastly, the software optimizations presented in this work and the hardware improvements in the Newell platform enable analytics and discovery on large graphs with millions of nodes and billions of edges in less than a minute. In sum, the new algorithmic implementations are significantly faster and can handle much larger “big” graphs.

In this paper, we propose an integer programming based model for tracking multiple maneuverable targets in a planar region. The objective function of this model uses both pairs and triplets of observations, which offer more accurate representation for constant velocity targets. Triplet scores in this model are calculated using a novel approach based on cubic spline interpolation, while the data association problem is solved using a specialized multi-dimensional assignment formulation. We show that the spline interpolation based scoring model provides more accurate reconstruction of trajectories, when compared to a naive model based on linear interpolation, on various randomly generated trajectories, at the expense of modest increase in computation time. The proposed multi-dimensional assignment formulation has nice structural properties and tight linear programming relaxation bound, which results in small computation times.

In this paper, we consider the problem of optimal placement of a single finite-size, rectangular facility in the presence of other rectangular facilities. There is a non-negative interaction between the new and existing facilities as well as pairs of existing facilities. All interactions are serviced through a finite number of input/output points located on the boundary of each facility. We assume that the travel occurs according to the rectilinear metric and the travel through the facilities is not permitted. It has been established in the previous works that the optimal placement of the new facility belongs to a finite set of candidate points, whose cardinality is polynomially bounded in the number of existing facilities. The optimal placement of the new facility can be found by evaluating the objective function value at each and every candidate point. This explicit enumeration guarantees the optimal solution, however it might become time consuming for a large number of existing facilities. We propose a new procedure based on the lower bounding technique, which can effectively cut down the number of candidate points that need to be evaluated, resulting in significant reduction in the computing time. The procedure was tested on a large number of randomly generated layouts with varying congestion factors (ratio of area occupied by the existing facilities to the total layout area). These extensive numerical tests reveal that, for a moderately congested layout, there is more than 70% reduction in both the number of evaluated candidates and the computing time.