cs.AREmpirical
CRIMP: Compact & Reliable DNN Inference on In-Memory Processing via Crossbar-Aligned Compression and Non-ideality Adaptation
Jul 9, 2026
This paper addresses the obstacles of using Crossbar-based In-Memory Processing (IMP) accelerators for deep neural networks (DNNs) by reusing bit-shift units for multiplication, applying pruning methods, and adapting to non-ideality.
The paper proposes a comprehensive learning framework that integrates three optimizations to reduce training overhead and improve accuracy in Crossbar-based In-Memory Processing for deep learning.
Before reading this…
Applications
To understand this paper, make sure you know these concepts first:
Crossbar-based In-Memory Processing (IMP) accelerators achieve high-speed, low-power computing for deep neural networks (DNNs), but face three obstacles. First, floating-point (FP) arithmetic is incompatible with crossbars, and existing quantization schemes still require FP processors for scaling factors, incurring hardware overhead. Second, redundant DNN parameters occupy too many crossbars, and current IMP-aware pruning methods require data aligning across crossbars, introducing significant memory and computing overhead. Third, non-ideal crossbar behaviors such as write variations degrade the accuracy of deployed models, and existing compensation methods add substantial overhead. In this paper, we address all three problems within a single training process. We reuse bit-shift units in crossbars to approximately multiply scaling factors, avoiding FP processors. We apply kernel-group pruning and crossbar pruning to remove the hardware units needed for data aligning. We adopt runtime-aware non-ideality adaptation to relieve the impact of device non-ideality from the training stage by exploiting crossbar features. Integrating these three optimizations into one comprehensive learning framework reduces training overhead and improves accuracy. Experiments show that our quantization incurs a negligible accuracy drop, and our pruning achieves higher sparsity and accuracy than state-of-the-art methods. Our framework produces integer-only, pruned, and reliable VGG-16 and ResNet-56 models for CIFAR-10 on IMP accelerators, with accuracy drops of only 2.19% and 1.26%, respectively, without hardware overhead.