Abstract: One embodiment provides a system for facilitating a graph search engine. During operation, the system receives, by a server from a client computing device, a search request which includes a user-inputted graph. The system performs a search based on a structure of the user-inputted graph for a plurality of relevant graphs. The system orders the plurality of relevant graphs from a most relevant ranking to a least relevant ranking. The system returns, to the client computing device, the ordered plurality of relevant graphs for display on a user interface of the client computing device, thereby enhancing the search for relevant graphs by allowing the graph search engine to take as an input the user-inputted graph and return as an output the relevant graphs.

Abstract: A method of deep graph representation learning includes: calculating a plurality of base features from a graph and adding the plurality of base features to a feature matrix. The method further includes generating, by a processing device, a current feature layer from the feature matrix and a set of relational feature operators, wherein the current feature layer corresponds to a set of current features, evaluating feature pairs associated with the current feature layer, and selecting a subset of features from the set of current features based on the evaluated feature pairs. The method further includes adding the subset of features to the feature matrix to generate an updated feature matrix.

Abstract: System and methods for relational time-series learning are provided. Unlike traditional time series forecasting techniques, which assume either complete time series independence or complete dependence, the disclosed system and method allow time series forecasting that can be performed on multivariate time series represented as vertices in graphs with arbitrary structures and predicting a future classification for data items represented by one of nodes in the graph. The system and methods also utilize non-relational, relational, temporal data for classification, and allow using fast and parallel classification techniques with linear speedups. The system and methods are well-suited for processing data in a streaming or online setting and naturally handle training data with skewed or unbalanced class labels.

Abstract: A method of deep graph representation learning includes: calculating a plurality of base features from a graph and adding the plurality of base features to a feature matrix. The method further includes generating, by a processing device, a current feature layer from the feature matrix and a set of relational feature operators, wherein the current feature layer corresponds to a set of current features, evaluating feature pairs associated with the current feature layer, and selecting a subset of features from the set of current features based on the evaluated feature pairs. The method further includes adding the subset of features to the feature matrix to generate an updated feature matrix.

Abstract: Embodiments described herein provide a system for facilitating hybrid task management across a central processing unit (CPU) and a graphics processing unit (GPU) of a computer. During operation, the system determines a set of tasks for performing data mining on a data set and storing the set of tasks in a data structure in an ascending order of uniformity associated with a respective task. The uniformity of a task indicates how uneven and skewed the task is compared to other tasks in the set of tasks. The system then allocates a subset of tasks to a core of the CPU from a front of the data structure and a subset of tasks to a core of the GPU from a back of the data structure.

Abstract: A system and a method perform matrix factorization. According to the system and the method, at least one matrix is received. The at least one matrix is to be factorized into a plurality of lower-dimension matrices defining a latent feature model. After receipt of the at least one matrix, the latent feature model is updated to approximate the at least one matrix. The latent feature model includes a plurality of latent features. Further, the update performed by cycling through the plurality of latent features at least once and alternatingly updating the plurality of lower-dimension matrices during each cycle.

Abstract: Embodiments of the present invention provide a system for fast parallel graph compression based on identifying a set of large cliques, which is used to encode the graph. The system provides both permanently-stored and in-memory graph encoding and reduces the space needed to represent and store a graph, the I/O traffic to use the graph, and the computation needed to perform algorithms involving the graph. The system thereby improves computing technology and graph computation. During operation, the system obtains data indicating vertices and edges of a graph. The system executes a clique-finding method to identify a maximum clique in the graph. The system then removes the clique from the graph, adds the clique to a set of found cliques, and generates a compressed representation of the graph based on the set of found cliques.

Abstract: Embodiments described herein provide a system for facilitating hybrid task management across a central processing unit (CPU) and a graphics processing unit (GPU) of a computer. During operation, the system determines a set of tasks for performing data mining on a data set and storing the set of tasks in a data structure in an ascending order of uniformity associated with a respective task. The uniformity of a task indicates how uneven and skewed the task is compared to other tasks in the set of tasks. The system then allocates a subset of tasks to a core of the CPU from a front of the data structure and a subset of tasks to a core of the GPU from a back of the data structure.

Abstract: Embodiments of the present invention provide a system for fast parallel graph compression based on identifying a set of large cliques, which is used to encode the graph. The system provides both permanently-stored and in-memory graph encoding and reduces the space needed to represent and store a graph, the I/O traffic to use the graph, and the computation needed to perform algorithms involving the graph. The system thereby improves computing technology and graph computation. During operation, the system obtains data indicating vertices and edges of a graph. The system executes a clique-finding method to identify a maximum clique in the graph. The system then removes the clique from the graph, adds the clique to a set of found cliques, and generates a compressed representation of the graph based on the set of found cliques.

Abstract: Embodiments of the present invention provide a system for fast, accurate, and scalable unbiased graphlet estimation. The system utilizes neighborhood sampling and combinatorial relations to estimate graphlet counts, statistics, and frequency distributions in a small fraction of the computing time of existing systems. The obtained unbiased estimates are highly accurate, and have applications in the analysis, mining, and predictive modeling of massive real-world networks. During operation, the system obtains data indicating vertices and edges of a graph. The system samples a portion of the graph and counts a number of graph features in the sampled portion of the graph. The system then computes an occurrence frequency of a graphlet pattern and a total number of graphlets associated with the graphlet pattern in the graph.

Abstract: Embodiments of the present invention provide a system for querying a graph based on applying filters to a visual representation of the graph. The system allows complicated graph query operations to be performed with ease visually. During operation, the system obtains data indicating vertices and edges of a graph. The system displays a visual representation of the graph for a user. The system receives, from the user, a command defining a local graph filter comprising a region in the visual representation. The system then filters a representation of the graph, and stores the filtered representation.

Abstract: System and methods for relational time-series learning are provided. Unlike traditional time series forecasting techniques, which assume either complete time series independence or complete dependence, the disclosed system and method allow time series forecasting that can be performed on multivariate time series represented as vertices in graphs with arbitrary structures and predicting a future classification for data items represented by one of nodes in the graph. The system and methods also utilize non-relational, relational, temporal data for classification, and allow using fast and parallel classification techniques with linear speedups. The system and methods are well-suited for processing data in a streaming or online setting and naturally handle training data with skewed or unbalanced class labels.

Abstract: A system and a method perform matrix factorization. According to the system and the method, at least one matrix is received. The at least one matrix is to be factorized into a plurality of lower-dimension matrices defining a latent feature model. After receipt of the at least one matrix, the latent feature model is updated to approximate the at least one matrix. The latent feature model includes a plurality of latent features. Further, the update performed by cycling through the plurality of latent features at least once and alternatingly updating the plurality of lower-dimension matrices during each cycle.