Abstract: The distributed inference paradigm enables the computation workload to be distributed across multiple devices, facilitating the implementation of deep learning based intelligent services on ...