Advanced Control Techniques

Distributed gradient algorithm over ordinary least squares problem

The project deals with the design and implementation of a distributed optimization algorithm in order to solve a Supervised Learning (SL) problem. In particular, the project activity includes to design a software based on Message Passing Interface (MPI) that implements a distributed gradient tracking algorithm. This software will be used to solve a Ordinary Least Squares problem and than the results obtained will be validated through numerical tests over a given dataset. All results will be desplayed graphically to give an immediate idea of the behavior of the agents during the estimate. Finally, the graph of the difference between the solution provided by the sequential algorithm and distributed one will highlight the computing differences between them.