PAC-Bayesian Compression Bounds on the Prediction Error of Learning Algorithms for Classification.
Thore GraepelRalf HerbrichJohn Shawe-TaylorPublished in: Mach. Learn. (2005)
Keyphrases
- prediction error
- pac bayesian
- learning algorithm
- machine learning
- machine learning algorithms
- supervised learning
- generalization ability
- feature vectors
- upper bound
- generalization bounds
- support vector
- feature extraction
- lower bound
- learning machines
- decision trees
- training samples
- bit rate
- support vector machine
- generalization error
- image classification
- computer vision
- multiresolution
- rademacher complexity
- distribution free
- active learning
- data dependent
- error bounds
- learning tasks
- cross validation
- svm classifier
- back propagation
- image compression
- support vector machine svm