Some tips about uncertainty visualization and associated techniques.

### Key ideas

The data associated with each point is a distribution instead of a fixed value

### Source of the uncertainty

looking at related ensmeble paper, compression, noise, and ensmeble simulation.

Typically, the uncertainty can come from noise, the down-sampling, and ensmeble members generated by simulation. The

### monte carlo sampling

using to distribution to sample from dedicated one

using sampled number dividid by whole number, using semi monte-carlo way, sample, then compute the probability

### density estimation

put the para non-para table here

the covaraince matrix.

The idea is simple, given an sampled array, how to derive its original distribution.

All kinds of ideas are summaries in the section of density estimation for the blog with title “Many aspects of MG, GMM and implementing linear algorithm”

### Combine with the ml

gaussian splitting

Discriminative network