Global minimum search using DMC algorithm with continuous weights

In this study we presented an algorithm for an unconstrained optimization of a continuous objective function, inspired by the Diffusion Monte Carlo method using a weight-based implementation. In this algorithm a cloud of replicas explores the solution space. Replicas are moved and evaluated after each step. Each replica carries an additional parameter (weight) which reflects the quality of its local solution. This parameter is updated after each step. Most inefficient replicas, i.e. replicas with the lowest weights, are occasionally replaced with their highest weight counterparts. In our study we present the basic implementation of the algorithm and compare its performance with other approaches, including the previously used implementation of DMC algorithm with a fluctuating population.