Overcoming Relational Learning Biases to Accurately Predict Preferences in Large Scale Networks
Published 18 May 2015
Joseph J. Pfeiffer, Jennifer Neville, Paul N. Bennett
Citations24
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
This work analyzes the effect of full semi-supervised RML and finds that collective inference methods can introduce considerable bias into predictions, and outlines a massively scalable variational inference algorithm for large scale relational network domains.
Abstract
Many individuals on social networking sites provide traits about themselves, such as interests or demographics. Social networking sites can use this information to provide better content to match their users' interests, such as recommending scheduled events or various relevant products. These tasks require accurate probability estimates to determine the correct answer to return. Relational machine learning (RML) is an excellent framework for these problems as it jointly models the user labels given their attributes and the relational structure. Further, semi-supervised learning methods could enable RML methods to exploit the large amount of unlabeled data in networks.
Keywords
Computer Science
Journal of DocumentationA STATISTICAL INTERPRETATION OF TERM SPECIFICITY AND ITS APPLICATION IN RETRIEVAL
4,442 Citations1972Karen Spärck Jones
It is argued that terms should be weighted according to collection frequency, so that matches on less frequent, more specific, terms are of greater value than matches on frequent terms.
National Bureau of Economic ResearchThe NBER Patent Citation Data File: Lessons, Insights and Methodological Tools
3,620 Citations2001Bronwyn H. Hall, Adam B. Jaffe +1 more
Semi-supervised learning using Gaussian fields and harmonic functions
3,377 Citations2003Xiaojin Zhu, Zoubin Ghahramani +1 more
Machine LearningText Classification from Labeled and Unlabeled Documents using EM
2,749 Citations2000Kamal Nigam, Andrew Kachites McCallum +2 more
This paper shows that the accuracy of learned text classifiers can be improved by augmenting a small number of labeled training documents with a large pool of unlabeled documents, and presents two extensions to the algorithm that improve classification accuracy under these conditions.
Machine LearningMarkov logic networks
2,678 Citations2006Matthew Richardson, Pedro Domingos
Experiments with a real-world database and knowledge base in a university domain illustrate the promise of this approach to combining first-order logic and probabilistic graphical models in a single representation.
Graphs over time
2,264 Citations2005Jure Leskovec, Jon Kleinberg +1 more
A new graph generator is provided, based on a "forest fire" spreading process, that has a simple, intuitive justification, requires very few parameters (like the "flammability" of nodes), and produces graphs exhibiting the full range of properties observed both in prior work and in the present study.
ACM Transactions on the WebThe dynamics of viral marketing
2,069 Citations2007Jure Leskovec, Lada A. Adamic +1 more
Discriminative probabilistic models for relational data
637 Citations2002Ben Taskar, Pieter Abbeel +1 more
Approximate medians and other quantiles in one pass and with limited memory
266 Citations1998Gurmeet Singh Manku, Sridhar Rajagopalan +1 more
New algorithms for computing approximate quantiles of large datasets in a single pass are presented, and the main memory requirements are smaller than those reported by an order of magnitude.
A Simple Relational Classifier
246 Citations2003Sofus A. Macskassy, Foster Provost
It is argued that a simple relational predictive model the predicts only based on class labels of related neighbors, using no learning and no inherent attributes should be used as a baseline to assess the performance of relational learners.
arXiv (Cornell University)Streaming Variational Bayes
122 Citations2013Tamara Broderick, Nicholas Boyd +3 more
SDA-Bayes is presented, a framework for streaming updates to the estimated posterior of a Bayesian posterior, with variational Bayes (VB) as the primitive, and the usefulness of the framework is demonstrated by fitting the latent Dirichlet allocation model to two large-scale document collections.
Tall and skinny QR factorizations in MapReduce architectures
52 Citations2011Paul G. Constantine, David F. Gleich
An implementation of the tall and skinny QR (TSQR) factorization in the MapReduce framework is presented, and computational results for nearly terabyte-sized datasets are provided.
Pseudolikelihood EM for Within-network Relational Learning
42 Citations2008Rongjing Xiang, Jennifer Neville
This paper develops a novel pseudolikelihood EM method that facilitates more general collective learning and collective inference on partially labeled relational networks and compares this method to competing methods from the other categories on both synthetic and real-world data.
Training a naive bayes classifier via the EM algorithm with a class distribution constraint
32 Citations2003Yoshimasa Tsuruoka, Jun’ichi Tsujii
The proposed method for using unlabeled data considerably improves classification performance when the amount of labeled data is small, and a class distribution constraint is introduced into the iteration process of the EM algorithm.
arXiv (Cornell University)Semi-Supervised Collective Classification via Hybrid Label\n Regularization
30 Citations2012Luke K. McDowell, David W. Aha
arXiv (Cornell University)Semi-Supervised Collective Classification via Hybrid Label Regularization
16 Citations2012Luke K. McDowell, David M. J. Tax
This work describes how to use novel combinations of classifiers to exploit the different characteristics of the relational features vs. the non-relational features and extends the ideas of label regularization to such hybrid classifiers, enabling them to leverage the unlabeled data to bias the learning process.
Understanding Propagation Error and Its Effect on Collective Classification
12 Citations2011Rongjing Xiang, Jennifer Neville
This paper characterize the high propagation error of collective inference models that are estimated with maximum pseudolikelihood estimation (MPLE), and shows how this affects performance across the spectrum of label availability when compared to MLE, which has low propagation error.
Composite Likelihood Data Augmentation for Within-Network Statistical Relational Learning
6 Citations2014Joseph J. Pfeiffer, Jennifer Neville +1 more
This work shows the fixed-point methods that R-EM uses for approximate learning and inference result in errors that prevent convergence in sparsely labeled networks, and develops a Relational Stochastic EM (R-SEM) method, which uses stochastic parameters that are not as susceptible to approximation errors.