(→TODO) |
m (→TODO: Typo fix.) |
||
| Line 11: | Line 11: | ||
* [[Naive Bayes_Old Kiwi]]-- What it is and why everyone should know about it. | * [[Naive Bayes_Old Kiwi]]-- What it is and why everyone should know about it. | ||
* [[Philosophies of Machine Learning_Old Kiwi]] -- A long article | * [[Philosophies of Machine Learning_Old Kiwi]] -- A long article | ||
| − | * [[Lower bound on performance of Bayes Classification_Old Kiwi]] is <math>\frac{1}{2}</math> when the number of classes is | + | * [[Lower bound on performance of Bayes Classification_Old Kiwi]] is <math>\frac{1}{2}</math> when the number of classes is 2 |
* [[Ideal performance of Bayes Classification_Old Kiwi]] when the two classes are Gaussian with the same variance and prior probability can be computed exactly, even when there is correlation between the dimensions | * [[Ideal performance of Bayes Classification_Old Kiwi]] when the two classes are Gaussian with the same variance and prior probability can be computed exactly, even when there is correlation between the dimensions | ||
* [[Amount of training data needed_Old Kiwi]] as a function of dimensions, covariance, etc. | * [[Amount of training data needed_Old Kiwi]] as a function of dimensions, covariance, etc. | ||
Revision as of 07:46, 17 April 2008
Hi! I'm Josiah Yoder, and I'm a big fan of Kiwis... and wikis.
My webpage is little out of date, but you can visit it anyway!
TODO
There are several articles I would like to write on the Kiwi when I get the time. If you would like to write them instead, please go for it, and let me know!
- Testing, Training, and Cross-Validation Data_Old Kiwi -- Everyone should know what each of these are!
- Using LibSVM effectively_Old Kiwi -- a brief review of what they already show in their documentation.
- Naive Bayes_Old Kiwi-- What it is and why everyone should know about it.
- Philosophies of Machine Learning_Old Kiwi -- A long article
- Lower bound on performance of Bayes Classification_Old Kiwi is $ \frac{1}{2} $ when the number of classes is 2
- Ideal performance of Bayes Classification_Old Kiwi when the two classes are Gaussian with the same variance and prior probability can be computed exactly, even when there is correlation between the dimensions
- Amount of training data needed_Old Kiwi as a function of dimensions, covariance, etc.
- Classification of data not in the Reals_Old Kiwi ($ \mathbb{R}^n $), such as text documents and graphs
- Fisher's Linear Discriminant_Old Kiwi -- Why it is ideal in the case of equal-variance Gaussians, a derivation that is less heuristic than the traditional development.
Administrative stuff to do:
- Copying stuff over from the old kiwi!
- Create a Lecture Template_Old Kiwi like someone has done manually at the bottom of every page.
