depth first search

I’ve added some new code to my library of algorithms (check the sidebar for others). If you are interested in regression, you should probably check out Gaussian Processes. We have at our disposal one of those rare cases when we can work with a class of infinite objects using only finite structures.

Now, as usual, some pictures. We’ll start with some training points drawn from a Gaussian process prior (check out Bishop’s book for details [eqn. 6.63]):

After fitting, which in this case just involves constructing the covariance matrix, we can find the mean and covariance estimate for the value of our function at any point: