TY - GEN
T1 - Cooperative training for attribute-distributed data
T2 - 2009 12th International Conference on Information Fusion, FUSION 2009
AU - Zheng, Haipeng
AU - Kulkarni, Sanjeev R.
AU - Poor, H. Vincent
PY - 2009
Y1 - 2009
N2 - This paper introduces a modeling framework for distributed regression with agents/experts observing attribute-distributed data (heterogeneous data). Under this model, a new algorithm, the iterative covariance optimization algorithm (ICOA), is designed to reshape the covariance matrix of the training residuals of individual agents so that the linear combination of the individual estimators minimizes the ensemble training error. Moreover, a scheme (Minimax Protection) is designed to provide a trade-off between the number of data instances transmitted among the agents and the performance of the ensemble estimator without undermining the convergence of the algorithm. This scheme also provides an upper bound (with high probability) on the test error of the ensemble estimator. The efficacy of ICOA combined with Minimax Protection and the comparison between the upper bound and actual performance are both demonstrated by simulations.
AB - This paper introduces a modeling framework for distributed regression with agents/experts observing attribute-distributed data (heterogeneous data). Under this model, a new algorithm, the iterative covariance optimization algorithm (ICOA), is designed to reshape the covariance matrix of the training residuals of individual agents so that the linear combination of the individual estimators minimizes the ensemble training error. Moreover, a scheme (Minimax Protection) is designed to provide a trade-off between the number of data instances transmitted among the agents and the performance of the ensemble estimator without undermining the convergence of the algorithm. This scheme also provides an upper bound (with high probability) on the test error of the ensemble estimator. The efficacy of ICOA combined with Minimax Protection and the comparison between the upper bound and actual performance are both demonstrated by simulations.
KW - Cooperative training
KW - Distributed learning
KW - Heterogeneous data
UR - http://www.scopus.com/inward/record.url?scp=70449368209&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=70449368209&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:70449368209
SN - 9780982443804
T3 - 2009 12th International Conference on Information Fusion, FUSION 2009
SP - 664
EP - 671
BT - 2009 12th International Conference on Information Fusion, FUSION 2009
Y2 - 6 July 2009 through 9 July 2009
ER -