Comments on: The Curse of Dimensionality in classification

By: Sara

Sara — Mon, 03 Jul 2017 05:57:37 +0000

Hi, I am a student who follows data analytics course module I my university. I was so confused with all the theories and awful explanations by the lecturer. Thank god I found this article!

By: Kevin George

Kevin George — Thu, 15 Jun 2017 05:31:26 +0000

Hi Vincent, first of all I would like to thank you for the wonder blog post and the work you are doing on visiondummy.

I had one doubt about equation 2. Shouldn’t there be only one distance from a sample point to a centroid? How can there be two distances i.e minimum and maximum distances?

By: Nikolaos Mparoutis

Nikolaos Mparoutis — Sat, 03 Jun 2017 10:38:03 +0000

Well written topic Vincent,good job.
However can you explain briefly (or confirm if it is true): Is the unitary hypershpere realated to clustering because clastering uses the Euclidean distance from the center?
Thank you for your time

By: Tahir

Tahir — Wed, 21 Dec 2016 13:54:40 +0000

Hi Vincent,

Thanks a lot for this useful Article. It’s a great explanation of (curse of Dimensionality).
I am interested in the first figure which shows the relation between the number of dimensions and classifier performance.
Is it possible to provide me some references that provide this figure with more details and experimental result, or any reference that contain this figure.

I suggest, it would be useful if you show the references that you used in writing your articles on this site.

Thanks in advance for answering or providing any reference.

By: jeremy

jeremy — Mon, 19 Sep 2016 14:43:42 +0000

Wow, this is super interesting! Just came upon your article from a stackexchange post, and I love your intuitive explanation of the curse of dimensionality; great work!

By: chanansh

chanansh — Sun, 17 Jan 2016 08:56:33 +0000

Hi, can you please explain what does the ratio dmax-dmin/dmin represents, why does it goes to zero and what does it mean? why is the distance from the center related to distance to neighbours if one would consider a KNN classifier in high dimension?

By: Anam

Anam — Wed, 06 Jan 2016 02:30:20 +0000

Article is informative but I am confused how to find optimal number of dimensions for dimension reduction techniques?

By: Richard

Richard — Tue, 22 Dec 2015 15:47:27 +0000

I’ve searched and searched and this was THE best explanation I found online. Thanks so much for this!

By: Bill Hsu

Bill Hsu — Tue, 08 Sep 2015 08:49:03 +0000

Can we have pdf version of these pages as well? Really helpful

By: Zone

Zone — Wed, 19 Aug 2015 12:51:09 +0000

I’m confused about the illustration of Figure 9:Training samples that fall outside the unit circle are in the corners of the feature space and are more difficult to classify than samples near the center of the feature space.
Why are the samples near the center easier to classify?
And why are the samples in the corners more difficult to classify?