ryt 294 Report post Posted December 22, 2013 Is it possible to lerp between functions ? The general formula that can be used for lerp between vectors is (1-t)v + tv1. But what if instead of a vector we put in a function like (1-t)x + v*x^2 where f(x) = x is the first function and g(x) = x^2 is second one. Will this work ? 0 Share this post Link to post Share on other sites
Paradigm Shifter 5832 Report post Posted December 22, 2013 Yes that would just be pointwise lerp between the functions evaluated at a specific point. A function can often be considered to be a vector anyway (e.g. the vector space C[0,1] is the space of all continuous functions defined in the interval [0,1]). 1 Share this post Link to post Share on other sites
alvaro 21246 Report post Posted December 22, 2013 If we want to be completely pedantic, LERP is an operation that is defined in affine spaces. So for instance you can linearly interpolate between two positions, even though positions are not vectors. But every vector space can be seen as an affine space as well (some people think of this as "forgetting where the origin is"), so you can use LERP in any vector space, including spaces of functions. 1 Share this post Link to post Share on other sites
ryt 294 Report post Posted December 23, 2013 I am not sure if this is right but if we take some other functions that are not linear like quadratic, will this still than hold ? Will the space than again be affine ? 0 Share this post Link to post Share on other sites
alvaro 21246 Report post Posted December 23, 2013 Quadratic in t? As long as all you are doing is taking a weighted sum of things with weights that add up to one, you are in good shape with any affine space. 0 Share this post Link to post Share on other sites
ryt 294 Report post Posted December 23, 2013 (edited) Well I meant in x, but it could be in t also. Why it needs to add up to one ? Edited December 23, 2013 by ryt 0 Share this post Link to post Share on other sites
alvaro 21246 Report post Posted December 23, 2013 No, no. Nobody is restricting the discussion to only linear functions. The "linear" in "linear interpolation" (what "LERP" stands for) refers to the dependence on t. 0 Share this post Link to post Share on other sites
ryt 294 Report post Posted December 23, 2013 I am a little confused. But anyway I can take any function I want for f(x) and g(x) so that (1-t)f(x) + tg(x) ? What would be the cases where I could not take any function ? 0 Share this post Link to post Share on other sites
alvaro 21246 Report post Posted December 23, 2013 But anyway I can take any function I want for f(x) and g(x) so that (1-t)f(x) + tg(x) ? Yes. What would be the cases where I could not take any function ? I don't understand that question. If you are asking about situations where it is not OK to take linear combinations like that, think of two points on the surface of a doughnut, two 3D rotations or even two points on a circle. There might be ways to make something that behaves similarly to LERP in those cases, but it will be much less straight forward than (1-t)*A+t*B. 1 Share this post Link to post Share on other sites
ryt 294 Report post Posted December 23, 2013 I don't understand that question. If you are asking about situations where it is not OK to take linear combinations like that, think of two points on the surface of a doughnut, two 3D rotations or even two points on a circle. There might be ways to make something that behaves similarly to LERP in those cases, but it will be much less straight forward than (1-t)*A+t*B. I was asking that to see if there were cases where I could not do this. Thank you, it is much clearer to me now. 0 Share this post Link to post Share on other sites
Tom KQT 1704 Report post Posted January 6, 2014 You'll maybe just have to be careful about possible discontinuities in f(x) or g(x). If one of them gives you infinity for a particular x, LERPing the values won't be straightforward. 1 Share this post Link to post Share on other sites