Random Search is a hyper-parameter optimization technique which consists of selecting the best performing model from the set of N models corresponding to hyper-parameters sampled from the hyper-parameter space according to a given probability distribution. For example: in a multi-layer perceptron model with a single hidden layer, we can perform a random search for the optimal combination of the number of hidden units and the learning rate, out of N sample combinations where the former is from the discrete uniform distribution between two positive integers and the latter is from the uniform distribution between a thousandth and a tenth.