Sensitivity-Aware Finetuning for Accuracy Recovery on Deep Learning Hardware

2023-06-05 · Lakshmi Nair, Darius Bunandar

Existing methods to recover model accuracy on analog-digital hardware in the presence of quantization and analog noise include noise-injection training. However, it can be slow in practice, incurring high computational costs, even when starting from pretrained models. We introduce the Sensitivity-Aware Finetuning (SAFT) approach that identifies noise sensitive layers in a model, and uses the information to freeze specific layers for noise-injection training. Our results show that SAFT achieves comparable accuracy to noise-injection training and is 2x to 8x faster.

📄 PDF Abstract BibTeX arXiv:2306.03076

Code (0)

등록된 구현이 없습니다.

Tasks

Deep LearningQuantizationSensitivity

ReRAM-aware Model Finetuning addressing I-V Non-linearity and Retention Errors

2026-06-16 · Ching-Yi Lin, Shamik Kundu, Arnab Raha, Sahil Shah arxiv

Traditional CPU, GPU, and NPU architectures are increasingly limited by the von Neumann bottleneck. While In-Memory Computing (IMC) using ReRAM crossbar arrays offers a high-density, energy-efficient alternative, its pra…

Image Classification

Adaptive Block Floating-Point for Analog Deep Learning Hardware

2022-05-12 · Ayon Basumallik, Darius Bunandar, Nicholas Dronen, Nicholas Harris 외

Analog mixed-signal (AMS) devices promise faster, more energy-efficient deep neural network (DNN) inference than their digital counterparts. However, recent studies show that DNNs on AMS devices with fixed-point numbers …

Deep Learning Quantization

MDQEC-QAS: Meta-Decoding for Quantum Error Correction with Hardware-Aware VQC Search and Confidence-Gated Recovery

2026-07-12 · Prashant Kumar Choudhary, Nouhaila Innan, Muhammad Shafique, Rajeev Singh arxiv

We propose a unified meta-decoding framework for quantum error correction that learns syndrome-to-recovery mappings across multiple stabilizer codes and noise settings, without requiring separate decoders for each config…

HCInfer: An Efficient Inference System via Error Compensation for Resource-Constrained Devices

2026-05-07 · Shen Xu, Xiangwen Zhuge, Zhe Xu, Yingkun Hu 외 arxiv

LLMs often struggle with memory-constrained deployment on consumer-grade hardware due to their massive parameter sizes. While existing solutions such as model compression and offloading improve deployment feasibility, th…

Model Compression

Provable Pruning for Efficient 3D Gaussian Splatting via Coresets

2026-07-02 · Waseem Mousa, Alaa Maalouf arxiv

3D Gaussian Splatting (3DGS) enables high-quality real-time novel-view synthesis, but practical scenes often contain millions of Gaussians, making compression essential for deployment on limited hardware. Existing reduct…