dhetting / math561_final_project Goto Github PK

Use average/mean/range, etc of TSI

We could calculate what percentage of the values for each variable are the same. Variables where the value is the same every time can probably be compressed more than variables with lots of variation

I will work on a more intelligent way to calculate the duplicates using the mean value, or casting it to a particular range of numbers.

Calculate sill, nugget, range for each image

e.g., Gini index, see https://arxiv.org/pdf/0811.4706.pdf

Not 100% sure if this makes sense, but we could try to determine if each image has a lot of entries that are around the same magnitude/value. It might also be useful to define a number of points similar to one another that constitute a "high density" part of the image, and then count how many times these high density regions occur.

A possible approach could be to calculate a percentile line (such as 85th percentile) for each image and then count how many times values from the image cross that percentile line (each crossing is defined as a peak).