![[2602.16336] HAWX: A Hardware-Aware FrameWork for Fast and Scalable ApproXimation of DNNs](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.16336] HAWX: A Hardware-Aware FrameWork for Fast and Scalable ApproXimation of DNNs
Summary
HAWX introduces a hardware-aware framework for efficiently approximating deep neural networks (DNNs), achieving significant speedups while maintaining accuracy across various architectures.
Why It Matters
As deep learning models grow in complexity, optimizing their performance on hardware becomes crucial. HAWX addresses this need by providing a scalable method to enhance DNN efficiency, which is vital for real-time applications and resource-constrained environments.
Key Takeaways
- HAWX employs multi-level sensitivity scoring to optimize DNN configurations.
- The framework achieves over 23x speedup in layer-level searches and significant improvements in filter-level searches.
- HAWX supports both spatial and temporal accelerator architectures, enhancing flexibility in hardware deployment.
Computer Science > Machine Learning arXiv:2602.16336 (cs) [Submitted on 18 Feb 2026] Title:HAWX: A Hardware-Aware FrameWork for Fast and Scalable ApproXimation of DNNs Authors:Samira Nazari, Mohammad Saeed Almasi, Mahdi Taheri, Ali Azarpeyvand, Ali Mokhtari, Ali Mahani, Christian Herglotz View a PDF of the paper titled HAWX: A Hardware-Aware FrameWork for Fast and Scalable ApproXimation of DNNs, by Samira Nazari and 6 other authors View PDF HTML (experimental) Abstract:This work presents HAWX, a hardware-aware scalable exploration framework that employs multi-level sensitivity scoring at different DNN abstraction levels (operator, filter, layer, and model) to guide selective integration of heterogeneous AxC blocks. Supported by predictive models for accuracy, power, and area, HAWX accelerates the evaluation of candidate configurations, achieving over 23* speedup in a layer-level search with two candidate approximate blocks and more than (3*106)* speedup at the filter-level search only for LeNet-5, while maintaining accuracy comparable to exhaustive search. Experiments across state-of-the-art DNN benchmarks such as VGG-11, ResNet-18, and EfficientNetLite demonstrate that the efficiency benefits of HAWX scale exponentially with network size. The HAWX hardware-aware search algorithm supports both spatial and temporal accelerator architectures, leveraging either off-the-shelf approximate components or customized designs. Subjects: Machine Learning (cs.LG); Artificial Intellige...
