Abstract: Estimation of distribution algorithms (EDAs) are a class of novel stochastic optimization algorithms, which have recently become a hot topic in field of evolutionary computation. EDAs acquire solutions by statistically learning and sampling the probability distribution of the best individuals of the population at each iteration of the algorithm. EDAs have introduced a new paradigm of evolutionary computation without using conventional evolutionary operatorssuch as crossover and mutation. In such a way, the relationships between the variables involved in the problem domain are explicitly and effectively exploited. According to the complexity of probability models for learning the interdependencies between the variables from the selected individuals, this paper gives a review of EDAs in the order of interactions: dependency-free, bivariate dependencies, and multivariate dependencies, aiming to bring the reader into this novel filed of optimization technology. In addition, the future research directions are discussed.