GraphLab: A New Parallel Framework for Machine Learning

By Yucheng Low, Joseph E. Gonzalez, Aapo Kyrola, Danny Bickson, Carlos Guestrin, Joseph M Hellerstein. 2010

471 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Kernel Belief Propagation (2011, methodology): “The first two were image denoising and depth prediction problems, where we show that kernel BP is superior to discrete, Gaussian mixture and particle BP in both speed and accuracy, using a GraphLab implementation of each (Low et al., 2010).”
Supporting Very Large Models using Automatic Dataflow Graph Partitioning (2018, background): “They can be partitioned in a “model parallel” manner by distributed graph computation frameworks [47, 48, 49, 50, 51, 52].”
Distributed GraphLab: A Framework for Machine Learning in the Cloud (2012, methodology): “In [24] we call this the full consistency model (see Fig.”
CloudCV: Large-Scale Distributed Computer Vision as a Cloud Service (2015, background): “GraphLab [39] is a high-level abstraction for distributed computing that efficiently and intuitively expresses data and computational dependencies with a sparse datagraph.”
A lightweight infrastructure for graph analytics (2013, background): “GraphLab [18] is a shared-memory programming model for topology or data-driven computations with autonomous or coordinated scheduling, but it is restricted to vertex programs.”
WTF: the who to follow service at Twitter (2013, methodology): “Finally, GraphLab [23] and its distributed variant [24] implement an alternative graph processing framework where computations can be performed either through an update function which defines local computations or through a sync mechanism which defines global aggregation in the context of different…”
AD3: alternating directions dual decomposition for MAP inference in graphical models (2015, unknown): “This has been shown to be very useful for large-scale message-passing inference in graphical models (Felzenszwalb and Huttenlocher, 2006; Low et al., 2010; Schwing et al., 2011).”
Green-Marl: a DSL for easy and efficient graph analysis (2012, methodology): “GraphLab [25] is a framework for machine learning type graph algorithms.”
GraphChi: Large-Scale Graph Computation on Just a PC (2012, unknown): “While we expect the memory capacity of personal computers to grow in the future, the datasets are expected to grow quickly as well.”
Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers (2011, unknown): “Specific frameworks directly based on or inspired by the BSP model include the Parallel BGL [91], GraphLab [114], and Pregel [119], among others.”
Distributed GraphLab: A Framework for Machine Learning in the Cloud (2012, methodology): “ resulting transformations can be challenging and may introduce substantial inefciency. As a consequence, there has been a recent trend toward graphparallel abstractions like Pregel [25] and GraphLab [24] which naturally express computational dependencies. These abstractions adopt a vertex-centric m”
GraphX: Graph Processing in a Distributed Dataflow Framework (2014, background): “While some of these systems [19, 13, 34] are capable of operating on multiple graphs independently, they do not expose an API or present optimizations for operations spanning graphs (or tables).”
Learning Graph Representations with Embedding Propagation (2017, background): “Whenever a computational process can be mapped to message exchanges between nodes, it is implementable in graph processing frameworks such as Pregel [29], GraphLab [23], and GraphX [44].”
Solving the Straggler Problem with Bounded Staleness (2013, methodology): “BSP provides a simple and easy-to-reason-about model for parallel computation, and can be easily applied to most iterative convergent algorithms.”
MapReduce for Parallel Reinforcement Learning (2011, background): “It would also be interesting to study alternative parallel frameworks, e.g., GraphLab [12], which address potential inefficiency with MapReduce, e.g., for problems with data-dpendancy due to MapReduce's share-nothing feature.”