Suppose one has a set of $n$ points in $\mathbb{R}^d$, where the points are represented as $p_1$ through to $p_n$. Define the weight matrix $W$ as follows: let $W_{i,j}$ be $e^{-||p_i - p_j ||^2}$, also sometimes known as a Gaussian weight matrix. My question is, is there a relation of the rank of $W$ to the underlying dimensionality $d$? If not, is there a rank $d$ matrix approximation of $W$? The reason I say this is that in a loose "information theoretic" sense, all the weights can be reconstructed from the $d$ real coordinates of each point.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.