Boundary-Aware Quantization: Finite-Scale Decision Geometry of Neural Classifiers
Not on the map. Why?
Abstract (arXiv)
We measured quantization-induced decision-boundary changes using local logit-margin radii, first-order boundary displacement, normal variation, slice-boundary Jaccard distance, grid prediction changes, multiclass junction counts, and low-margin boundary-band flips. On the digits benchmark, 8-bit weight quantization preserved all test labels while producing boundary-mask Jaccard (0.428) on the PCA slice; at 4 bits, accuracy remained (0.9733), while boundary Jaccard rose to (0.970) and median local boundary shift reached (0.0290). Interpolation between adjacent quantization levels localized the visible reconfigurations at multiclass junctions, with 12, 34, and 17 triple-junction cells in the selected transitions. Calibration-to-test stopping reduced the digits held-out flip rate from (0.0094) to (0.0022) and boundary Jaccard from (0.825) to (0.524); the same stopping rule also reduced flips on MNIST and Fashion-MNIST. On official CIFAR-10 subsets, PTQ-W selected by accuracy gave 6-bit flip (0.0367) and boundary Jaccard (0.184), whereas boundary-aware stopping selected 8-bit flip (0.0083) and boundary Jaccard (0.048). On full CIFAR-10 with three seeds, 6-bit PTQ-W lost (0.0029) accuracy relative to float, changed (5.3%) of held-out decisions, and changed (24.5%) of low-margin boundary-band decisions. A fixed-bit boundary-gap rounding term changed the trade-off at 4 bits by reducing boundary Jaccard from (0.457) to (0.435) and boundary-band pair-order flip from (0.3600) to (0.3558), with an accuracy trade-off; the 3-bit stress test exposed the tuning limit of this surrogate. Calibration boundary Jaccard predicted held-out boundary Jaccard across PTQ-W and optimized rounding variants with (r=0.947)--(0.994).
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 10 of its 12 references that are in the map. First posted on arXiv 1 Jul 2026.
Builds on
10 of 12 references.
References · 10
- 2014Explaining and Harnessing Adversarial Examplescited by 7,287
- 2013Intriguing properties of neural networkscited by 5,114
- 2017Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inferencecited by 1,548
- 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,315
- 2020Up or Down? Adaptive Rounding for Post-Training Quantizationcited by 600
- 2018Quantizing deep convolutional networks for efficient inference: A whitepapercited by 507
- 2018Robustness via curvature regularization, and vice versacited by 174
- 2020Understanding the Decision Boundary of Deep Neural Networks: An Empirical Studycited by 19
- 2021On the Adversarial Robustness of Quantized Neural Networkscited by 11
- 2021A Layer-wise Adversarial-aware Quantization Optimization for Improving Robustnesscited by 3
Built on by
Citing papers per year: 2026: 2.
2 papers.