4

One of my colleagues has results from an evolutionary algorithm that he wants to plot. The plot will be included in a paper he's writing and asked for advice. This is what he had in matlab:

enter image description here

I gave him the following advice:

  • The stroke length of the asterisks is about the same as that of the line dashes. Especially in the lower left and upper right corner of the graph this looks really chaotic, so I told him to remove the asterisks and use solid, thin lines instead.

  • intermediate yellow-ish plots are barely visible, so the yellow must go

  • Both low and high fitness values (plot color) have very similar gray values and couldn't be reproduced by print in grayscale. My advice was to use a different coloring, preferably gray or one that gracefully reduces to a meaningful outcome when printed.

The result:

enter image description here

How can we further improve this? The overall appearance is much less chaotic, but the message could be clearer: The darker plots depict better results, but they don't really stick out. Also regions where many plots are close to each other appear darker than the individual plots.

Of course, the color bar needs some annotation to clarify the meaning of dark and bright values (low, dark values mean a better result).

I'm not sure if the grid is helpful in this situation or not. It doesn't seems to obfuscate information.

  • I'll just comment instead of answering because I don't understand what I'm looking at, but I would say the best thing you could do is extend the X axis, maybe the Y axis too so there's more space between each points. – Ryan Nov 20 '15 at 13:04
3

This is (more or less) what we ended up with:

enter image description here

The grid and the "individuals" plots have the same gray value so the individuals will need to be a bit darker in the final version, but I think it's already reasonably readable.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.