machine learning - Projecting from 60D (shape context) space to 2D for visual analysis -

I have a set of 60D shape reference vectors, they have 5 radial compartments and 12 angular coaches (Thus, I have 60 D is the 400-size reference vectors) using a silhouette using 400 edge point samples.

How I would like to analyze these vectors are representing the overall shape of the underlying silhouette. To do this, I would like to project 60D size reference vectors back in 2D space and as a result, I will inspect the result - which I expect to see is a set of digits which is roughly the same as the original silhouette size is.

An approach to doing this is projecting the first two major components (PCA). Based on my implementation, the estimated point of the silhouette was not the same. I can see two main reasons for this (assuming for the time being that my implementation is correct): (1) Size references either are not suitable as a deliverator, or its parameters should be better tuned. (2) This analysis method is faulty / not valid.

My question is, is this the correct approach for analyzing the descriptivity of size references in relation to my silhouette shape? If not, can someone please explain why and offer an alternative method?

Thanks,

Josh

Good to check this The way that the features are descriptive or not, some classifier (SVM / Byz / Tree / Anyone) has to check them and check it for cross-vested precision / memory etc. Like feature selectors, they can filter their feature vector such as chi / infographic.

In addition to PCA, you can visualize your data with SOM or by clustering.