AFarePart: Accuracy-aware Fault-resilient Partitioner for DNN Edge Accelerators

This article introduces AFarePart, a new approach for partitioning Deep Neural Networks (DNNs) to improve their performance on edge accelerators. The focus is on accuracy and fault tolerance, which are crucial for reliable edge computing. The research likely explores how to divide DNN models effectively to minimize accuracy loss while also ensuring resilience against hardware failures. The use of 'accuracy-aware' suggests the system dynamically adjusts partitioning based on the model's sensitivity to errors. The 'fault-resilient' aspect implies mechanisms to handle potential hardware issues. The source being ArXiv indicates this is a preliminary research paper, likely undergoing peer review.

• •AFarePart is a new partitioning approach for DNNs on edge accelerators.
• •It focuses on accuracy and fault tolerance.
• •The system is likely accuracy-aware, dynamically adjusting partitioning.
• •It incorporates fault-resilient mechanisms to handle hardware issues.