Microsoft Technology Licensing, LLC

Other

Renato F Werneck

Daniel R Delling

Andrew V Goldberg

Thomas Pajor

Edith Cohen

Ilya Razenshteyn

Andreas Georg Nowatzyk

Moises Goldszmidt

Xiang Zhao

Mihai Budiu

Haim Kaplan

Christopher John Leggetter

Ittai Abraham

Xia Sharon Wan

Darshan Vishwanath Kantak

Moritz Helge Kobitzsch

Yunhong Zhou

Naiping Liu

Fabian Fuchs

Aspects of the subject disclosure are directed towards managing sponsored online content based upon advertiser behavior. Defining mini-markets to represent such advertiser behavior may be accomplished by clustering queries that generate revenue from one or more campaigns. Query revenue data between queries and a set of campaigns may be used to determine such mini-markets. To illustrate, a query whose highest revenue is attributed to a campaign may be selected for that campaign's mini-market. When that query is entered as a search term, the campaign's mini-market helps allocate space for advertisements.

Estimating similarity of nodes using all-distances sketches

Customizable route planning is a technique for computing point-to-point shortest paths in road networks. It includes three phases: preprocessing, metric customization, and queries. A graphics processing unit may be used, e.g., in the metric customization phase, to make customization even faster, enabling a wide range of applications including highly dynamic applications and on-line personalized cost functions.

Customizable route planning using graphics processing unit

Distance query techniques are provided that are robust to network structure, scale to large and massive networks, and are fast, straightforward, and efficient. A hierarchical hub labeling (HHL) technique is described to determine a distance between two nodes or vertices on a network. The HHL technique provides indexing by ordering vertices by importance, then transforming the ordering into an index, which enables fast exact shortest-path distance queries. The index may be compressed without sacrificing its correctness.

Distance queries on massive networks

A graph that includes multiple nodes and edges is received. Multiple instances of the graph are generated by randomly instantiating the edges according to either a binary independent cascade model or a randomized edge length independent cascade model. Where the binary independent cascade model is used, combined reachability sketches are generated for each node across all instances of the graph. Where the randomized edge length independent cascade model is used, combined all-distances sketches are generated for each node across all instances of the graph. Depending on which model is used, the combined reachability or all-distances sketches are used to estimate the influence of nodes in the graph or to estimate a subset of nodes from a graph of a specified size with a maximum influence using a greedy algorithm.

Estimating influence using sketches

A point-to-point shortest path technique supports real-time queries and fast metric update or replacement (metric customization). Determining a shortest path between two locations uses three stages: a preprocessing stage, a metric customization stage, and a query stage. Extensions to the customizable route planning (CRP) technique for routing are provided. These extensions include, for example, the computation of alternative routes, faster techniques for unpacking shortcuts, efficient query techniques for batched shortest path (one-to-many, many-to-many, and points of interest) determinations, and determining routes and alternative routes using traffic information.

Query scenarios for customizable route planning

A distributed data-parallel execution (DDPE) system splits a computational problem into a plurality of sub-problems using a branch-and-bound algorithm, designates a synchronous stop time for a 'plurality of processors' (for example, a cluster) for each round of execution, processes the search tree by recursively using a branch-and-bound algorithm in multiple rounds (without inter-processor communications), determines if further processing is required based on the processing round state data, and terminates processing on the processors when processing is completed.

Distributed data-parallel execution engines for user-defined serial problems using branch-and-bound algorithm

Techniques are described for graph partitioning, and in particular, graph bisection. A lower bound is provided that is computed in near-linear time. These bounds may be used to determine optimum solutions to real-world graphs with many vertices (e.g., more than a million for road networks, or tens of thousands for VLSI and mesh instances). A packing lower bound technique determines lower bounds in a branch-and-bound tree, reducing the number of tree nodes. Techniques are employed to assign vertices without branching on them, again reducing the size of the tree. Decomposition is provided to translate an input graph into less complex subproblems. The decomposition boosts performance and determines the optimum solution to an input by solving subproblems independently. The subproblems can be solved independently using a branch-and-bound approach to determine the optimum bisection.

Determining optimum graph bisection using a tree and pruning the tree using a packing lower bound

Graph partitioning techniques are based on the notion of natural cuts. A filtering phase performs a series of minimum cut computations to identify and contract dense regions of the graph. This reduces the graph size significantly, but preserves its general structure. An assembly phase uses a combination of greedy and local search heuristics to assemble the final partition. The techniques may be used on road networks, which have an abundance of natural cuts (such as bridges, mountain passes, and ferries).

Graph partitioning with natural cuts

Alternative routes to an optimal route may be determined and presented to a user via a computing device. Alternative routes are selected from candidate routes that meet admissibility criteria. In an implementation, admissibility of a candidate route (in order for it to be considered an alternative route) may be determined based on three criteria: 'limited sharing', 'local optimality', and &#x201c;stretch&#x201d; such as &#x201c;uniformly bounded stretch&#x201d;. Limited sharing refers to the amount of difference between the alternative route and the optimal route, local optimality refers to lack of unnecessary detours, and uniformly bounded stretch refers to a length of the shortest path to travel between two points on the alternative route.

Determining alternative routes

Optimum journeys in public transportation networks are determined. The determination of Pareto optimal journeys from one stop to another stop in a public transportation network uses the criteria travel time and minimum transfers. A technique for bi-criteria journey planning using the aforementioned criteria in public transportation networks operates in rounds (K rounds at most), where after round k (k&#x2266;K), arrival times are computed for the stops that can be reached with up to k trips.

Journey planning in public transportation networks

The non-negative single-source shortest path (NSSP) problem is solved on a graph by using a preprocessing phase and then, in a query phase, computing the distances from a given source in the graph with a linear sweep over all the vertices. Contraction hierarchies may be used in the preprocessing phase and in the query phase. Optimizations may include reordering the vertices in advance to exploit locality, performing multiple NSSP computations simultaneously, marking vertices during initialization, and using parallelism. Techniques may be performed on a graphics processing unit (GPU). This makes applications based on all-pairs shortest-paths practical for continental-sized road networks. The applications include, for example, computing graph diameter, exact arc flags, and centrality measures such as exact reaches or betweenness.

Hardware accelerated shortest path computation

An algorithm referred to as REAL for the point-to-point shortest path problem combines A* search with landmark-based lower bounds and reach-based pruning. A symbiosis of these techniques is described, which gives a range of time and space tradeoffs, including those that improve both of these complexity measures. Locality is improved and exact reach computation is described.

Growing community of inventors

San Francisco, CA, United States of America

Renato F Werneck