Training deep neural networks for power control in Multiple input multiple output systems

In wireless communication systems reliability can be achieved through spatial diversity and the data rate can be improved with spatial multiplexing. Multiple input multiple output systems otherwise known as MIMO systems achieves both. To improve the efficiency of MIMO systems, the available resources like power and spectrum should be optimally allocated among the users. Water filling algorithm is a classic approach for the same. This paper presents a version of the existing improved water filling technique approximated using deep learning technique in multi user systems inorder to obtain increased capacity and improved accuracy. Comparisons of already existing techniques, approximated with and without deep learning technique with proposed method is performed in single user and multi user based MIMO systems to obtain improved capacity.