Random search. You evaluate a function (say, a loss function for a statistical model given certain parameters) at a random point, and then randomly displace it

Random search. You evaluate a function (say, a loss function for a statistical model given certain parameters) at a random point, and then randomly displace it by a constant amount in a random direction. Random search is quite useful for hyperparameter optimization, and it can be a useful building block for something like simulated annealing or parallel tempering.