Runtime Deep Model Multiplexing for Reduced Latency and Energy Consumption Inference | IEEE Conference Publication | IEEE Xplore