NeFT: Negative Feedback Training to Improve Robustness of Compute-In-Memory DNN Accelerators

TL;DR

NeFT introduces a control theory-inspired training method that enhances the robustness of compute-in-memory DNN accelerators against device variations, significantly improving inference accuracy and confidence.

Contribution

The paper proposes Negative Feedback Training (NeFT), a novel approach that better captures noisy information during training, outperforming existing methods in robustness against device variations.

Findings

Up to 45.08% accuracy improvement

Reduced epistemic uncertainty and increased confidence

Enhanced convergence probability in DNN inference

Abstract

Compute-in-memory accelerators built upon non-volatile memory devices excel in energy efficiency and latency when performing deep neural network (DNN) inference, thanks to their in-situ data processing capability. However, the stochastic nature and intrinsic variations of non-volatile memory devices often result in performance degradation during DNN inference. Introducing these non-ideal device behaviors in DNN training enhances robustness, but drawbacks include limited accuracy improvement, reduced prediction confidence, and convergence issues. This arises from a mismatch between the deterministic training and non-deterministic device variations, as such training, though considering variations, relies solely on the model's final output. In this work, inspired by control theory, we propose Negative Feedback Training (NeFT), a novel concept supported by theoretical analysis, to more effectively capture the multi-scale noisy information throughout the network. We instantiate this concept with two specific instances, oriented variational forward (OVF) and intermediate representation snapshot (IRS). Based on device variation models extracted from measured data, extensive experiments show that our NeFT outperforms existing state-of-the-art methods with up to a 45.08% improvement in inference accuracy while reducing epistemic uncertainty, boosting output confidence, and improving convergence probability. These results underline the generality and practicality of our NeFT framework for increasing the robustness of DNNs against device variations. The source code for these two instances is available at https://github.com/YifanQin-ND/NeFT_CIM