Exploring Edge TPU for deep feed-forward neural networks

This paper explores the performance of Google's Edge TPU on feed-forward neural networks. We consider Edge TPU as a hardware platform and explore different architectures of deep neural network classifiers, which traditionally have been a challenge to run on resource-constrained edge devices. Based on the use of a joint-time-frequency data representation, also known as a spectrogram, we explore the trade-off between classification performance and the energy consumed for inference. The energy efficiency of Edge TPU is compared with that of the widely-used embedded CPU ARM Cortex-A53. Our results quantify the impact of neural network architectural specifications on the Edge TPU's performance, guiding decisions on the TPU's optimal operating point, where it can provide high classification accuracy with minimal energy consumption. Also, our evaluations highlight the crossover in performance between Edge TPU and Cortex-A53, depending on the neural network specifications. Based on our analysis, we provide a decision chart to guide decisions on platform selection based on the model parameters and context.

Automated summary

This paper examines the performance of Google's Edge TPU on feed-forward neural networks. It considers the Edge TPU as a hardware platform and explores different architectures of deep neural network classifiers, which have traditionally been challenging to run on resource-constrained edge devices. By utilizing a spectrogram data representation, the study examines the trade-off between classification performance and energy consumption for inference. The energy efficiency of the Edge TPU is compared to that of the widely-used embedded CPU ARM Cortex-A53. The results provide insights into the impact of neural network architecture on the performance of the Edge TPU and offer guidance for selecting the optimal operating point based on classification accuracy and energy consumption. Additionally, the evaluations highlight the performance crossover between the Edge TPU and Cortex-A53, depending on the neural network specifications. The analysis also provides a decision chart to assist in platform selection based on model parameters and context.