# Eric Gregor

less info
reputation
617
bio website location age member for 2 years, 11 months seen Dec 17 at 17:01 profile views 234

 Dec9 comment Norm of covariance and precision matrices: is there any meaning? Good answer, but I am actually more interested in the precision matrix case. Dec6 comment Maximum entry of small-normed matrix And I was referring to the spectral norm. Dec4 comment Are roots of a sparse matrix are sparse? I didn't say anything about diagonal matrices. Dec2 comment Norm inequality question I am not saying that we should have an equality, just some reasonable perturbation of this $\epsilon$, perhaps involving the norm of $T$. Dec1 comment “max” entry norm inequality? For the 2 norm I think you need a factor of $\sqrt{n}$ in front of it, by the relationship of the 2 norm and infinity norm. Dec1 comment “max” entry norm inequality? @AlgebraicPavel, the spectral norm. But perhaps a more sensible bound is a multiple of the infinity norm? Nov7 comment Covariance of a mixture of Gaussians I mean can we say this is elliptical or something? Nov7 comment Covariance of a mixture of Gaussians Thanks! Does this have a nice distribution, or is this all we can say? Oct8 comment Probability/Decision- infimum over set of expectations (can be interpreted as decision problem) @whuber Btw, how is the second limit being zero implied by the assumption on the first moment of F being finite? Oct8 comment Solving integral related to second moment of normal distribution restricted to interval. CAS answer acceptable That's not consistent with what you have about. Notice what happens if m is zero in your output. Oct7 comment Solving integral related to second moment of normal distribution restricted to interval. CAS answer acceptable If m=0, everything cancels out and you get zero. Am I mistaken? Oct7 comment Solving integral related to second moment of normal distribution restricted to interval. CAS answer acceptable If m is the mean, your answer says you have zero when m=0. That can't be. Oct7 comment Solving integral related to second moment of normal distribution restricted to interval. CAS answer acceptable are you still in process? I see that both the last and second integrals are zero when the mean is zero. And then you are left with an integral which is the same as the first, in the case where the mean is zero. Oct7 comment Solving integral related to second moment of normal distribution restricted to interval. CAS answer acceptable What are you calling m? Oct6 comment Probability/Decision- infimum over set of expectations (can be interpreted as decision problem) Yes. It can be simplified a bit. But how to show that the right side reduces to the left... Oct6 comment Probability/Decision- infimum over set of expectations (can be interpreted as decision problem) After what line? May22 comment Computing bottom $k$-eigenspace of a matrix via top $k$-eigenspace of another matrix Yes, Will, I am indeed considering the absolute values of the eigenvalues. The problem with the inverse is that it is about as expensive as the full eigendecomposition. May22 comment Probability that one folded normal is bigger than another? I would love to see your sketch. May22 comment Computing bottom $k$-eigenspace of a matrix via top $k$-eigenspace of another matrix Why should the bottom eigenvectors of $R$ be the same here as the bottom eigenvectors of $S$? May22 comment Computing bottom $k$-eigenspace of a matrix via top $k$-eigenspace of another matrix Thank you for your comment. You are right. I have modified the question to reflect my real wish, which is a matrix that is less expensive to compute than the full eigendecomposition or inverse.