Optimization of many complex systems is often viewed as a black-box optimization problem. Such problems are often difficult to solve using conventional techniques, for a variety of reasons, such as the absence of derivatives, mixed data types, and so on. Techniques such as Genetic Algorithms, Estimation of Distribution Algorithms such as MIMIC and the CE method, and more recently, mathematically rigorous approaches such as Probability Collectives have been used for black-box optimization. It turns out that many of these techniques fall under the category of Monte Carlo Optimization. In this technique, we present a brief statistical analysis of Monte Carlo Optimization (MCO), and show that it is identical to Parametric Machine Learning (PL). Owing to this identity, we can use PL techniques to improve the performance of MCO. Then, we present a new version of the black-box optimization technique of Probability Collectives., and demonstrate the use of PL techniques to improve its optimization performance.