Minimax Probability Machine (MPM) Jay Silver
Very High Level Diagram of Training a Pattern Classifier Augmented Testing a New Data Point
Finding a Function that Decides Decision If , choose class wy , choose class wx Assume Binary Non Parametric Parametric Support Vector Machine (SVM) Minimax Probability Machine (MPM) Gaussian
MPM SVM Non-Parametric Linear Decision Boundaries Maximal Margin Classifier Minimize Worst Future Error An SVM and MPM toolbox were used for implementation [1,4]. MPM figure borrowed from [2].
MPM Problem Statement Lower bound on test accuracy Upper bound of misclassifying future point with Mahalanobis Distance Equal Problem Statement s.t. Lower bound on test accuracy An SVM and MPM toolbox were used for implementation [1,4]. MPM figure borrowed from [2].
Expanding the Feature Space with Kernels Original Feature Space Expanded Feature Space XOR: {x1, x2} XOR: {x1, x2, x1x2} Not Linearly Separable Linearly Separable Kernel Examples Gaussian Kernel: Polynomial Kernel:
Take a Look at Some Linear Decision Boundaries Key
Results for the Distribution We Just Saw SVM Performs Best MPM Performs Well SVM Homogeneous Polynomial Fails to Converge
Alpha as an Underbound to Test Accuracy Compare Alpha to Test Accuracy Just Note Correlation Between Alpha and Test Accuracy Key
Testing on a Real Speech Task Deterding Data – 11 vowel sounds with 10 features Multiple classes – Use 1 vs. 1 voting to generalize binary classifiers Test Accuracy for the Gaussian Kernel MPM Peaks At 67.3% Key SVM Peaks At 68.4%
Summary of Deterding Results Distill Results Further Linear Nonlinear Classifier Accuracy Bayes 50.7% SVM 51.7% MPM 48.7% Classifier Accuracy Bayes 47.2% SVM 68.4% MPM 67.3%
Conclusions Alpha is an accurate lower bound for all cases but one. Alpha was reasonably well correlated with test accuracy. SVM homogeneous polynomial kernel outperformed MPM But MPM homo. poly. kernel was more consistent MPM Gaussian kernel performed 1% below SVM on Deterding MPM: Competitive, including realistic speech tasks Mathematically pleasing Room to grow Not quite as accurate as SVMs
References
Questions? The Rainbow Linear Discriminant Between CSTIT Students