0

I'm working on a personal project of mine thought it would be interesting if I could compare two graphs and determine just how similar they are, or given a certain portion of the graphs determine their similarity in the same manner.

For example, if I was to take two random graphs, is there any realistic way of deciding how similar the two are without doing some pixel-by-pixel comparison? As I am more inclined to the actual pattern/shape of the graphs rather than if they are identical.

I'm thinking this idea actually probably lends itself to some sort of machine learning as well, now, since the previous topic I found myself looking at said the same thing (posted 9 years ago).

Similar Previous Posting 9 Years Ago

Jacob Macallan
  • 959
  • 2
  • 8
  • 29

1 Answers1

0

Plot is graphical representation of data.

  1. If you have the data, you can compare them by some metrics (Mean squared error, Maximum value, etc.).
  2. If you dont have the data, you can get them by machine vision techniques.

I was to take two random graphs, is there any realistic way of deciding how similar the two are without doing some pixel-by-pixel comparison?

Yes:

  1. Find pixels of plot (e.g. by thresholding)
  2. Find function that approximates these pixels (e.g. polynomial function of second order by linear regression)
  3. Do step 1,2 for two images and compare coefficients.
Albert Myšák
  • 194
  • 1
  • 1
  • 13