摘要: Extending linear classifiers from feature vectors to attributed graphs results in sublinear classifiers. In contrast models, the classification performance of models depends on our choice as which class we label positive and negative. We prove that expected accuracy may differ for different labelings. Experiments confirm this finding empirical accuracies small samples. These give rise flip-flop consider both labelings during training select model prediction better fits data.
springer.com
sci-hub.st 参考文章(20)
Graph-Based Representations in Pattern Recognition Lecture Notes in Computer Science. ,vol. 5534, ,(1998) , 10.1007/978-3-642-20844-7
Brijnesh Johannes Jain, Structural Neural Learning Machines ,(2005) , 10.14279/DEPOSITONCE-1357
Robert Tibshirani, Trevor Hastie, Jerome H. Friedman, The Elements of Statistical Learning ,(2001)
Structural, syntactic, and statistical pattern recognition Lecture Notes in Computer Science. ,vol. 6218, ,(2002) , 10.1007/978-3-642-14980-1
Kaspar Riesen, Horst Bunke, IAM Graph Database Repository for Graph Based Pattern Recognition and Machine Learning SSPR & SPR '08 Proceedings of the 2008 Joint IAPR International Workshop on Structural, Syntactic, and Statistical Pattern Recognition. pp. 287- 297 ,(2008) , 10.1007/978-3-540-89689-0_33
Brijnesh J. Jain, Fritz Wysotzki, Structural Perceptrons for Attributed Graphs Lecture Notes in Computer Science. pp. 85- 94 ,(2004) , 10.1007/978-3-540-27868-9_8
Brijnesh J. Jain, Klaus Obermayer, Generalized Learning Graph Quantization Graph-Based Representations in Pattern Recognition. pp. 122- 131 ,(2011) , 10.1007/978-3-642-20844-7_13
B.J. Jain, F. Wysotzki, Perceptron learning in the domain of graphs international joint conference on neural network. ,vol. 3, pp. 1993- 1998 ,(2003) , 10.1109/IJCNN.2003.1223713
Brijnesh J. Jain, Klaus Obermayer, Maximum Likelihood for Gaussians on Graphs Graph-Based Representations in Pattern Recognition. pp. 62- 71 ,(2011) , 10.1007/978-3-642-20844-7_7
Brijnesh J. Jain, Sublinear Models for Graphs. arXiv: Learning. ,(2014)