Variance needed for both axes about rtsne

[jkrijthe]

The costs variable contains some measure of how well each object is represented, but in general, I do not think there is a straightforward measure of how much each dimension "adds" to the representation of the objects. Note that t-sne does not try to explain the total variance of the data (of which you could explain a certain percentage, like the principal components), but rather the local configuration of objects. See the help files and the original t-sne papers for more details.