Systems, methods, apparatuses, and computer-readable media for computational model optimization. A plurality of sampled values for a hyperparameter of a computational model may be received, the plurality of sampled values comprising a subset of a plurality of possible values for the hyperparameter, each sampled value associated with a performance metric for the computational model with the sampled value assigned to the hyperparameter. A first candidate value from the plurality of possible values may be determined, the first candidate value having a distance to a first sampled value of the plurality of sampled values that exceeds a threshold distance, wherein the distance is in a space comprising the plurality of possible values. The first candidate value may be assigned to the hyperparameter of the computational model. A first performance metric may be determined for the computational model with the first candidate value assigned to the hyperparameter.