Matrix factorization with the Alternative Least Squares (ALS) algorithm. This code is based on GraphLab's implementation of ALS by Joey Gonzalez and D

Matrix factorization with the Alternative Least Squares (ALS) algorithm. This code is based on GraphLab's implementation of ALS by Joey Gonzalez and D

Scheduler implementation for "Selective Scheduling". Each vertex in the graph has a bit which is 1 if the vertex should be updated, and 0 otherwise. T

GraphChi stores vertex adn edge values in compact byte format. To convert the bytes to Java objects, one needs to define converter objects. Note that

Represents an edge. @param <EdgeData>

Copyright [2012] [Aapo Kyrola, Guy Blelloch, Carlos Guestrin / Carnegie Mellon University] Licensed under the Apache License, Version 2.0 (the "Licen

Wrapper for Java logging. Use ChiLogger.getLogger("object-name") to get a logger object.

Internal representation for referring to a chunk of bytes. Emulates C/C++ pointers.

Represents a vertex. Vertex contains a value and a set of in- and out-edges. @param <VertexValue> @param <EdgeValue>

Zlib-compressed I/O utilities. @author akyrola

Example application for computing the weakly connected components of a graph. The algorithm uses label exchange: each vertex first chooses a label equ

Creates a cassovary adjacency @author Aapo Kyrola

Manages large chunks of data which are accessed using ChiPointers. Used internally by GraphChi. @author akyrola

GraphChi keeps track of the degree of each vertex (count of in- and out-edges). This class allows accessing the vertex degrees efficiently by loading

@author Aapo Kyrola

Presents a map from integers to frequencies. Special distributions for avoidance can be used to exclude certain ids from the distributions in merges.

Client for querying the DrunkardCompanion.

DrunkardCompanion is a remote (or local) service that receives walks from the DrunkardEngine and maintains a distribution of visits from each source.

@author Aapo Kyrola

Class to encapsulate the graphchi program running the show. Due to several optimizations, it is quite complicated!

Encapsulates a random walk computation.

Class for running DrunkardMob random walk applications. This can run multiple distinct random walk computations. They are executed simultaneously when

@author Aapo Kyrola

@author Aapo Kyrola, akyrola@cs.cmu.edu, akyrola@twitter.com

Interface for objects that translate edge values from string to the value type. @param <ValueType>

Placeholder for vertex and edge data types that are 'empty' @author Aapo Kyrola

New version of sharder that requires predefined number of shards and translates the vertex ids in order to randomize the order, thus requiring no addi

@author Aapo Kyrola

Converts byte[4] array to Float and Float to byte[4]. @author Aapo Kyrola

Represents a 2-tuple of floats. Access the tuple elements by pair.first, pair.second. @author Aapo Kyrola

Converts byte[8] arrays to FloatPair's and vice versa @see edu.cmu.graphchi.datablocks.FloatPair @author Aapo Kyrola, akyrola@cs.cmu.edu

Callback object used when iterating vertex-values with VertexAggregator @author akyrola @see edu.cmu.graphchi.vertexdata.VertexAggregator

Demonstration of the queryAndCombine capabilities of GraphChi. With this app, after you have computed PageRank for each vertex, you can make simple fr

GraphChiContext represents the current state of the computation. This is passed to the update-function. @see edu.cmu.graphchi.GraphChiProgram

Multiplex for DrunkardDrivers.

The engine responsible for executing a GraphChi computation. @param <VertexDataType> type of vertex-data @param <EdgeDataType> type of edge-data

All GraphChi applications must extend this class. @param <VertexDataType> @param <EdgeDataType>

@author Aapo Kyrola, akyrola@cs.cmu.edu, akyrola@twitter.com

Converts an indirected input graph into a directed by checking that each edge has a complimentary edge in the opposite direction, and adding those com

Loads a graph from HDFS edge by edge and calls a callback for each edge. Used by the Pig-integration: @see edu.cmu.graphchi.hadoop.PigGraphChiBase

Version of HITS that uses just a little memory (values propagated via edges), and can be run under Pig. On each iteration either left or right side i

A huge dense matrix, which internally splits to many sub-blocks Row-directed storage, so scanning row by row is efficient. This is useful in keeping

A huge dense matrix, which internally splits to many sub-blocks Row-directed storage, so scanning row by row is efficient. This is useful in keeping

A huge dense matrix, which internally splits to many sub-blocks Row-directed storage, so scanning row by row is efficient. This is useful in keeping

Tuple presenting a vertex-id and a count. @author akyrola Date: 7/15/12

Container class carrying a vertex id and a float value. @author Aapo Kyrola

Tuple containg a vertex-id and an integer number.

Converts byte[4] ot Integer and vice versa. @author Aapo Kyrola

@author Aapo Kyrola

Class to encapsulate the graphchi program running the show. Due to several optimizations, it is quite complicated!

Manager for random walks Done partially during authors internship at Twitter, Fall 2012. @author Aapo Kyrola, akyrola@cs.cmu.edu

An integer array buffer. Not thread-safe!

K-core decomposition algorithm Outputs: a file containing key-value pairs: vertexId, coreness How does it work ? 1 - Initializes vertex values to th

Utility for counting the number of different labels in the vertex-data. Vertices which have same id as their vertex-id are not calculated. This is use

Converts byte[8] to Long and vice versa. @author Aapo Kyrola

@author Aapo Kyrola

Class to encapsulate the graphchi program running the show. Due to several optimizations, it is quite complicated!

Manager for random walks Done partially during authors internship at Twitter, Fall 2012. @author Aapo Kyrola, akyrola@cs.cmu.edu

Used only internally - do not modify. To understand Memory shards, see http://code.google.com/p/graphchi/wiki/IntroductionToGraphChi @param <EdgeDataT

Uses Netflix ratings data (or equivalent data) to compute recommendations for users. Computes in three steps: 1. Runs Alternating Least Squares (ALS

Samples values from multinomial distribution. @author Aapo Kyrola

Example application: PageRank (http://en.wikipedia.org/wiki/Pagerank) Iteratively computes a pagerank for each vertex by averaging the pageranks of in

Computes estimate of personalized pagerank using the DrunkardMobEngine. <b>Note:</b> this version omits walks to adjacent vertices, and thus could be

<b>Version for Pig http://pig.apache.org</b> Matrix factorization with the Alternative Least Squares (ALS) algorithm. This code is based on GraphLab'

Special PIG loader that wraps a graphchi application. This allows execution of GraphChi programs under Hadoop/PIG. Generally, if you have an GraphChi

Example application: PageRank (http://en.wikipedia.org/wiki/Pagerank) Iteratively computes a pagerank for each vertex by averaging the pageranks of in

Example application: WeightedPageRank (http://en.wikipedia.org/wiki/Pagerank) Iteratively computes a pagerank for each vertex by averaging the pagera