Matlab Code
Matlab File Exchange
My Matlab code comprises two pedagogical software demos - one for linear SVMs and one for Adaboost. The Adaboost demo was developed as part of a lecture that I delivered for an MSc Machine Learning course (Machine Learning and Data Mining).
There is also a progress bar tool that I wrote. It was chosen as the Matlab File Exchange "Pick of the Week"
Matlab File Exchange
My Matlab code comprises two pedagogical software demos - one for linear SVMs and one for Adaboost. The Adaboost demo was developed as part of a lecture that I delivered for an MSc Machine Learning course (Machine Learning and Data Mining).
There is also a progress bar tool that I wrote. It was chosen as the Matlab File Exchange "Pick of the Week"
No-Free-Lunch Theorem Talk
I gave this talk to the Machester MLO group. It gives an introduction to No-Free-Lunch theorems, shows how they apply to Bayesian Inference and Structural Risk Minimisation, and then describes a relationship with epistemic philosophy.
I have also presented a similar talk to the Random Walks on Physics lecture series in the University of Manchester School of Physics and Astronomy. In this version of the talk, I omitted discussion of SRM, and instead presented some instances of NFL-like issues in Aumann's Agreement Theorem, Compression and Regression.
I gave this talk to the Machester MLO group. It gives an introduction to No-Free-Lunch theorems, shows how they apply to Bayesian Inference and Structural Risk Minimisation, and then describes a relationship with epistemic philosophy.
I have also presented a similar talk to the Random Walks on Physics lecture series in the University of Manchester School of Physics and Astronomy. In this version of the talk, I omitted discussion of SRM, and instead presented some instances of NFL-like issues in Aumann's Agreement Theorem, Compression and Regression.
Boosting Lecture
I delivered this lecture on Adaboost to MSc students on the Machine Learning and Data Mining course. It introduces the algorithm and delves into one of its facinating theoretical properties.
I delivered this lecture on Adaboost to MSc students on the Machine Learning and Data Mining course. It introduces the algorithm and delves into one of its facinating theoretical properties.
CIDUE Paper
Theoretical and Empirical Analysis of Diversity in Non-Stationary Learning, R. Stapenhurst and G. Brown, 2011 IEEE Symposium on Computational Intelligence in Dynamic and Uncertain Environments.
Available here.
I presented the paper at CIDUE in Paris, April 2011. The presentation slides are available here.
Theoretical and Empirical Analysis of Diversity in Non-Stationary Learning, R. Stapenhurst and G. Brown, 2011 IEEE Symposium on Computational Intelligence in Dynamic and Uncertain Environments.
Available here.
I presented the paper at CIDUE in Paris, April 2011. The presentation slides are available here.
Technical Report, March 2012
On the Relationship between Ensemble Diversity and Margin Theory, R. Stapenhurst, MLO Group, School of Computer Science, University of Manchester, 2012.
Available here
On the Relationship between Ensemble Diversity and Margin Theory, R. Stapenhurst, MLO Group, School of Computer Science, University of Manchester, 2012.
Available here
1st year Continuation Report
www.cs.man.ac.uk/~stapenr5/report.pdf
My report from the end of first year. Mostly irrelevant now; concerning negative correlation learning, boosting, and applying online learning techniques to reinforcement learning problems.
www.cs.man.ac.uk/~stapenr5/report.pdf
My report from the end of first year. Mostly irrelevant now; concerning negative correlation learning, boosting, and applying online learning techniques to reinforcement learning problems.