Search papers, labs, and topics across Lattice.
This paper introduces the Non-ideality Optimized eNVM Accelerator (NOVA) architecture, which addresses the energy efficiency challenges of deep learning by leveraging emerging non-volatile memory (eNVM) for on-chip training. The authors develop a Non-ideality Avoidance Training (NAT) algorithm that optimizes weight convergence to the most stable conductance regions of eNVM devices, significantly mitigating the impact of intrinsic non-idealities. Experimental results show that NAT enhances accuracy by an average of 15.1% across various benchmark tasks while NOVA achieves a remarkable 33.58脳 improvement in energy efficiency compared to traditional GPUs.
Achieving a 15.1% accuracy boost in neural network training while enhancing energy efficiency by over 33 times could redefine on-chip training paradigms.
The rapid advancement of deep learning has presented significant energy efficiency challenges to the conventional von Neumann architecture. In-memory computing (IMC) architectures based on emerging non-volatile memory (eNVM) are widely regarded as a promising solution for accelerating neural network training due to their high parallelism and low power consumption. However, the intrinsic non-idealities of eNVM devices can cause conductance updates to deviate from target values, thereby limiting the performance of on-chip training. To address this challenge, this paper presents a Non-ideality Optimized eNVM Accelerator (NOVA) architecture for on-chip training. Specifically, we first fabricate a two-dimensional (2D) ferroelectric field-effect transistor (FeFET) and develop a conductance modulation behavioral model calibrated with experimental data. Building upon this device model, we propose, for the first time, a Non-ideality Avoidance Training (NAT) algorithm tailored for eNVM devices, which mitigates accuracy degradation by guiding weight convergence toward the most stable conductance regions of eNVM devices. Experimental results demonstrate that, even under severe device asymmetry, NAT improves the accuracy by an average of 15.1\% over the baseline methods across multiple benchmark tasks. Meanwhile, the NOVA achieves an average energy efficiency gain of approximately 33.58$\times$ compared with the peak energy efficiency of graphics processing units (GPUs).
