We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Computer Science > Machine Learning

Title: Quantized Convolutional Neural Networks Through the Lens of Partial Differential Equations

Abstract: Quantization of Convolutional Neural Networks (CNNs) is a common approach to ease the computational burden involved in the deployment of CNNs, especially on low-resource edge devices. However, fixed-point arithmetic is not natural to the type of computations involved in neural networks. In this work, we explore ways to improve quantized CNNs using PDE-based perspective and analysis. First, we harness the total variation (TV) approach to apply edge-aware smoothing to the feature maps throughout the network. This aims to reduce outliers in the distribution of values and promote piece-wise constant maps, which are more suitable for quantization. Secondly, we consider symmetric and stable variants of common CNNs for image classification, and Graph Convolutional Networks (GCNs) for graph node-classification. We demonstrate through several experiments that the property of forward stability preserves the action of a network under different quantization rates. As a result, stable quantized networks behave similarly to their non-quantized counterparts even though they rely on fewer parameters. We also find that at times, stability even aids in improving accuracy. These properties are of particular interest for sensitive, resource-constrained, low-power or real-time applications like autonomous driving.
Subjects: Machine Learning (cs.LG)
Cite as: arXiv:2109.00095 [cs.LG]
  (or arXiv:2109.00095v2 [cs.LG] for this version)

Submission history

From: Ido Ben-Yair [view email]
[v1] Tue, 31 Aug 2021 22:18:52 GMT (911kb)
[v2] Wed, 3 Aug 2022 13:22:54 GMT (466kb)

Link back to: arXiv, form interface, contact.