| Line 13: | Line 13: | ||
This slecture to be reviewed by Ben Foster | This slecture to be reviewed by Ben Foster | ||
| + | |||
| + | This slecture provides a brief overview of the KNN classification method and transitions from KNN into a brief description of the nearest neighbor classification method. A few comments/suggestions: | ||
| + | |||
| + | *It is alright that there is not an abundance of theoretical background for KNN presented in this slecture, but it does seem like there could be a bit more. This is especially true since the idea of the slecture is to set the stage for the nearest neighbor method with KNN as a support. With this in mind, the author could use to make a more structured transition from KNN to the nearest neighbor method. | ||
| + | |||
| + | *There could also be some more discussion of the unbiased nature of KNN as an estimator for the local density <math>\rho(\vec{X_0})</math>. Why is this a point of interest? Are other local estimators of <math>\rho(\vec{X_0})</math> biased? Those that have taken the class are likely familiar with the answers to these questions, but this may be a point of confusion for a reader that is just becoming familiar with the material. | ||
| + | |||
| + | *The discussion of distance metrics is similarly sparse. Perhaps the author could add links/references to sites where more information on these topics can be found. | ||
---- | ---- | ||
Revision as of 07:52, 7 May 2014
Review on KNN to Nearest Neighbor Slecture by Jonathan Manring
A slecture by graduate student Jonathan Manring
Comments/Feedback
This slecture to be reviewed by Ben Foster
This slecture provides a brief overview of the KNN classification method and transitions from KNN into a brief description of the nearest neighbor classification method. A few comments/suggestions:
- It is alright that there is not an abundance of theoretical background for KNN presented in this slecture, but it does seem like there could be a bit more. This is especially true since the idea of the slecture is to set the stage for the nearest neighbor method with KNN as a support. With this in mind, the author could use to make a more structured transition from KNN to the nearest neighbor method.
- There could also be some more discussion of the unbiased nature of KNN as an estimator for the local density $ \rho(\vec{X_0}) $. Why is this a point of interest? Are other local estimators of $ \rho(\vec{X_0}) $ biased? Those that have taken the class are likely familiar with the answers to these questions, but this may be a point of confusion for a reader that is just becoming familiar with the material.
- The discussion of distance metrics is similarly sparse. Perhaps the author could add links/references to sites where more information on these topics can be found.
