标题： 专题会议：在非易失性计算内存加速器上的深度神经网络可持续部署 显示英文标题

标题： Special Session: Sustainable Deployment of Deep Neural Networks on Non-Volatile Compute-in-Memory Accelerators

摘要： 基于非易失性存储器（NVM）的存内计算（CIM）加速器已成为一种可持续解决方案，由于其就地数据处理能力，可显著提高深度神经网络（DNN）推理的能效并降低延迟。 然而，由于NVM器件的随机性和固有变化，NVCIM加速器的性能会下降。 传统的写入验证操作，在部署期间通过迭代写入和验证来提高推理准确性，但在能耗和时间上成本较高。 受负反馈理论的启发，我们提出了一种新的负优化训练机制，以实现针对NVCIM的鲁棒DNN部署。 我们开发了一种定向变分前向（OVF）训练方法来实现该机制。 实验表明，OVF在推理准确性方面比现有最先进技术提高了高达46.71%，同时减少了认知不确定性。 该机制减少了对写入验证操作的依赖，从而有助于NVCIM加速器的可持续和实用部署，在解决性能下降问题的同时保持了NVCIM加速器的可持续计算优势。 显示英文摘要

摘要： Non-volatile memory (NVM) based compute-in-memory (CIM) accelerators have emerged as a sustainable solution to significantly boost energy efficiency and minimize latency for Deep Neural Networks (DNNs) inference due to their in-situ data processing capabilities. However, the performance of NVCIM accelerators degrades because of the stochastic nature and intrinsic variations of NVM devices. Conventional write-verify operations, which enhance inference accuracy through iterative writing and verification during deployment, are costly in terms of energy and time. Inspired by negative feedback theory, we present a novel negative optimization training mechanism to achieve robust DNN deployment for NVCIM. We develop an Oriented Variational Forward (OVF) training method to implement this mechanism. Experiments show that OVF outperforms existing state-of-the-art techniques with up to a 46.71% improvement in inference accuracy while reducing epistemic uncertainty. This mechanism reduces the reliance on write-verify operations and thus contributes to the sustainable and practical deployment of NVCIM accelerators, addressing performance degradation while maintaining the benefits of sustainable computing with NVCIM accelerators.