An Efficient Algorithm for Mapping Deep Learning Applications on the NoC Architecture

Artificial Intelligence (AI) has a very important role in the modern world. Through the use of the AI, organizations are able to make better decisions, and can improve business processes because speed and accuracy both are increased in decision making. This study is related to mapping of Artificial Intelligence Algorithms, particularly Neural Networks (NN) on cores in a Network-on-chip (NoC) platform. In this work, neurons are the tasks of a NN in real life and our goal is to divide those tasks among processing cores of the NoC. To complete this process, multiple optimization algorithms are used to map neurons of NN onto the NoC to reduce its computation time. Furthermore, to evaluate the solutions, the hidden layer complexity of the NN is varied, and Octave/Google Colaboratory based simulations are used to get these results. The result indicates improvement in terms of energy consumption, on-chip communication, and application processing time.