The increase in the number of wireless devices also generated an increase in the density of the frequency band used, thus increasing the probability of interference between them and by conventional noise. Due to the need to minimize or reduce these effects, learning algorithms for performing statistical inference are studied, performing the function of monitoring and predicting complex systems. This project is focused on the study of adaptive learning algorithms capable of performing inference in order to avoid the effect of interference between devices and minimize possible errors, in particular using algorithms based on the stochastic gradient and the close gradient in a learning scenario distributed. This work was divided into three stages: study and simulation of distributed networks; test in real scenario: temperature measurement; and elaboration of the next algorithm and study of performance in the federative scenario.