With the wide application of deep learning technology, the updated speed of neural architecture and demand for computing power is increasing daily. As a result, there is an urgent need for multi-level optimization for task performance and hardware overhead, from neural network architecture to hardware architecture.
