Multi objective non dominated sorting whale optimization genetic algorithm for convolutional neural network-based on-chip networks

In recent years, chip technology has experienced rapid development, achieving the integration of multiple processing units and driving the development of modern multi-core systems. Network on Chip (NoC) has been widely adopted as an interconnect communication architecture superior to traditional buses. In the realm of artificial intelligence, the widespread application of convolutional neural networks has brought enormous computational demands, while traditional computers have relatively slow processing speeds. In order to address this challenge, people have begun to develop convolutional neural network accelerators, namely artificial intelligence chips, which use on-chip networks to achieve efficient data transmission and communication. In order to effectively and efficiently map convolutional neural networks onto on-chip networks and achieve goals such as low power consumption and low latency, a multi-objective mapping method for non dominated sorting whale optimization genetic algorithm is proposed by utilizing the strong global search ability of whale optimization algorithm and combining it with non dominated sorting algorithm, aiming to solve the mapping problem of on-chip networks. The experimental results demonstrate that the proposed method ∊an achieve reduced power consumption and latency.