Learning interpretable SVMs for biological sequence classification

Posted in Science on August 17, 2008


Learning interpretable SVMs for biological sequence classification

Lecture slides:

  • Roadmap:
  • Biology: Detection of Splice Sites
  • Approach: String Kernel + SVM
  • Success
  • Gain
  • Reformulation: Multiple Kernel Learning
  • Biology: Detection of Splice Sites
  • Approach: String Kernel + SVM
  • Multiple Kernel Learning
  • Constraining the weights
  • Standard SVM Optimization Problem
  • MKL Optimization Problem
  • The Semi-Infinite Linear Program
  • Solving the SILP: Column Generation
  • Solving the SILP: Boosting
  • Stability of the solution ?
  • Toy Dataset
  • Application to Acceptor splice sites
  • Conclusion

Author: Sören Sonnenburg, Fraunhofer First

Watch Video

Tags: Science, Lectures, Computer Science, Machine Learning, VideoLectures.Net, Kernel Methods