List of Contributors |
Preface |
Foundations / I: |
Introduction / Samy Bengio ; Joseph Keshet1: |
The Traditional Approach to Speech Processing / 1.1: |
Potential Problems of the Probabilistic Approach / 1.2: |
Support Vector Machines for Binary Classification / 1.3: |
Outline / 1.4: |
References |
Theory and Practice of Support Vector Machines Optimization / Shai Shalev-Shwartz ; Nathan Srebo2: |
SVM and L2-regularized Linear Prediction / 2.1: |
Optimization Accuracy From a Machine Learning Perspective / 2.3: |
Stochastic Gradient Descent / 2.4: |
Dual Decomposition Methods / 2.5: |
Duality / 2.5.1: |
Summary / 2.6: |
From Binary Classification to Categorial Prediction / Koby Crammer3: |
Multi-category Problems / 3.1: |
Hypothesis Class / 3.2: |
Loss Functions / 3.3: |
Hinge Loss Functions / 3.4: |
A Generalized Perceptron Algorithm / 3.5: |
A Generalized Passive-Aggressive Algorithm / 3.6: |
A Batch Formulation / 3.7: |
Concluding Remarks / 3.8: |
Appendix / 3.9: |
Derivations of the Duals of the Passive-Aggressive Algorithm and the Batch Formulation |
Acoustic Modeling / II: |
A Large Margin Algorithm for Forced Alignment / Yoram Singer ; Dan Chazan4: |
Problem Setting / 4.1: |
Cost and Risk / 4.3: |
ALarge Margin Approach for Forced Alignment / 4.4: |
An Iterative Algorithm / 4.5: |
Efficient Evaluation of the Alignment Function / 4.6: |
Base Alignment Functions / 4.7: |
Experimental Results / 4.8: |
Discussion / 4.9: |
A Kernel Wrapper for Phoneme Sequence Recognition / 5: |
Frame-based Phoneme Classifier / 5.1: |
Kernel-based Iterative Algorithm for Phoneme Recognition / 5.4: |
Nonlinear Feature Functions / 5.5: |
Preliminary Experimental Results / 5.6: |
Discussion: Canwe Hope for Better Results? / 5.7: |
Augmented Statistical Models: Using Dynamic Kernels for Acoustic Models / Mark J. F. Gales6: |
Temporal Correlation Modeling / 6.1: |
Dynamic Kernels / 6.3: |
Augmented Statistical Models / 6.4: |
Conclusions / 6.5: |
Acknowledgements |
Large Margin Training of Continuous Density Hidden Markov Models / Fei Sha ; Lawrence K. Saul7: |
Background / 7.1: |
Large Margin Training / 7.3: |
Conclusion / 7.4: |
Language Modeling / III: |
A Survey of Discriminative Language Modeling Approaches for Large Vocabulary Continuous Speech Recognition / Brian Roark8: |
General Framework / 8.1: |
Further Developments / 8.3: |
Summary and Discussion / 8.4: |
Large Margin Methods for Part-of-Speech Tagging / Yasemin Altun9: |
<$$$> / 9.1: |
List of Contributors |
Preface |
Foundations / I: |
Introduction / Samy Bengio ; Joseph Keshet1: |
The Traditional Approach to Speech Processing / 1.1: |
Potential Problems of the Probabilistic Approach / 1.2: |