Let's say I have samples of a bounded (but potentially noisy) function at some fixed interval of points. How would I go about determining the likelihood that my observed data was sampled from an *increasing* function with heteroscedastic noise? #machinelearning #statistics #GaussianProcesses ??