PCA with standardization in Matlab

By running [coeff,score,latent] = pca(X, ‘Centered’, true, ‘VariableWeights’, ‘variance’);, we can get the standardized PCA results. (The transformed observations are in score.)

When running just [coeff,score,latent] = pca(X);, only centering is applied.

Latex figure positioning

http://ttic.uchicago.edu/~dmcallester/ttic101-07/lectures/em/em.pdf

An EM algorithm paper.

For better understanding of hard EM and soft EM.

wav file structure (in Korean)

On Linux servers, when there are many processes requiring memory spaces larger than the given memory spaces, paging occurs a lot and the overall performance is significantly degraded. To evade that situation, I suspended some processes by “kill -STOP <pid>”. Then, the suspended processes are very gradually swapped out than I expected. That means, suspending existing processes doesn’t help immediately reduce the paging :(.

Bayesian Information Criterion

Can also  be used for estimating # clusters.

Knowledge is not stored as complete representations but rather is subdivided into distinct categories and stored separately.

-p. 16, Principles of Neural Science, Eric R. Kandel

From Sangyoon Yi’s post

core is dumped at the current directory. If you can’t find it, try the following.

ulimit -c unlimited  # Setting a core dump file can have an arbitrary size.

ulimit -a # Check the changed setting

Then, the core will be generated in the current directory.

log(x+y) = log(1/M*exp(log(Mx)) + exp(log(My))), where M=max{1/x, 1/y}