Mahalanobis Distance
HTML-код
- Опубликовано: 14 дек 2021
- What is the Mahalanobis Distance?
This video starts from regular Euclidean distance and builds up to linear transformations. The Mahalanobis distance is actually a form of this linear transformation, when taking the transformation to be the inverse covariance matrix of some distribution - hence the Mahalanobis distance is not simply a distance in transformed space, it's a distance w.r.t. a specific distribution.
Become a member!
meerkatstatistics.com/courses/
* 🎉 Special RUclips 60% Discount on Yearly Plan - valid for the 1st 100 subscribers; Voucher code: First100 🎉 *
Why become a member?
* All video content
* Extra material on complete-courses (notebooks)
* Access to code and notes
* Community Discussion
* No Ads
* Support the Creator ❤️
If you’re looking for statistical consultation, work on interesting projects, or training workshop, visit my website meerkatstatistics.com/ or contact me directly at david@meerkatstatistics.com
~~~~~ SUPPORT ~~~~~
Paypal me: paypal.me/MeerkatStatistics
~~~~~~~~~~~~~~~~~
Intro/Outro Music: Dreamer - by Johny Grimes
• Johny Grimes - Dreamer
Probably the best explanation of the topic, thank you!
Hi, thanks for the helpful video. Could you please refer me to a paper where it mentioned for mahalanobis distance the A=sigma^(-0.5).
Don't have a paper in mind. Notice that my derivation is of A'A which comes out sigma^(-1) which coincide with the definition.
Hello, really a nice video. My question is "What does it mean lengths = 1/eigen values"? Which length does it refer to?
Thanks. The lengths from the origin to the "unit distance". I show the lengths in the video at 8:30 (look at my correction as well). A bit hard to explain with words, but: For the ellipsoid, the eigenvalue 1/4 corresponds to the length sqrt(4)=2 from the origin to the "unit" distance of the (-1,1) eigen-vector, and the 9/4 corresponds to the sqrt(4/9) from the origin to the "unit" distance of the (1,1) vector. Hope it helps.