How many bits does it take to quantize your neural network?

Quantization converts neural networks into low-bit fixed-point computations which can be carried out by efficient integer-only hardware, and is standard practice for the deployment of neural networks on real-time embedded devices. However, like their real-numbered counterpart, quantized networks are...

詳細記述

書誌詳細
主要な著者: Giacobbe, M, Henzinger, TA, Lechner, M
フォーマット: Conference item
言語:English
出版事項: Springer 2020
_version_ 1826261221993611264
author Giacobbe, M
Henzinger, TA
Lechner, M
author_facet Giacobbe, M
Henzinger, TA
Lechner, M
author_sort Giacobbe, M
collection OXFORD
description Quantization converts neural networks into low-bit fixed-point computations which can be carried out by efficient integer-only hardware, and is standard practice for the deployment of neural networks on real-time embedded devices. However, like their real-numbered counterpart, quantized networks are not immune to malicious misclassification caused by adversarial attacks. We investigate how quantization affects a network’s robustness to adversarial attacks, which is a formal verification question. We show that neither robustness nor non-robustness are monotonic with changing the number of bits for the representation and, also, neither are preserved by quantization from a real-numbered network. For this reason, we introduce a verification method for quantized neural networks which, using SMT solving over bit-vectors, accounts for their exact, bit-precise semantics. We built a tool and analyzed the effect of quantization on a classifier for the MNIST dataset. We demonstrate that, compared to our method, existing methods for the analysis of real-numbered networks often derive false conclusions about their quantizations, both when determining robustness and when detecting attacks, and that existing methods for quantized networks often miss attacks. Furthermore, we applied our method beyond robustness, showing how the number of bits in quantization enlarges the gender bias of a predictor for students’ grades.
first_indexed 2024-03-06T19:18:07Z
format Conference item
id oxford-uuid:191fabd6-e8b9-41c1-8018-6842c9e9e111
institution University of Oxford
language English
last_indexed 2024-03-06T19:18:07Z
publishDate 2020
publisher Springer
record_format dspace
spelling oxford-uuid:191fabd6-e8b9-41c1-8018-6842c9e9e1112022-03-26T10:47:10ZHow many bits does it take to quantize your neural network?Conference itemhttp://purl.org/coar/resource_type/c_5794uuid:191fabd6-e8b9-41c1-8018-6842c9e9e111EnglishSymplectic ElementsSpringer2020Giacobbe, MHenzinger, TALechner, MQuantization converts neural networks into low-bit fixed-point computations which can be carried out by efficient integer-only hardware, and is standard practice for the deployment of neural networks on real-time embedded devices. However, like their real-numbered counterpart, quantized networks are not immune to malicious misclassification caused by adversarial attacks. We investigate how quantization affects a network’s robustness to adversarial attacks, which is a formal verification question. We show that neither robustness nor non-robustness are monotonic with changing the number of bits for the representation and, also, neither are preserved by quantization from a real-numbered network. For this reason, we introduce a verification method for quantized neural networks which, using SMT solving over bit-vectors, accounts for their exact, bit-precise semantics. We built a tool and analyzed the effect of quantization on a classifier for the MNIST dataset. We demonstrate that, compared to our method, existing methods for the analysis of real-numbered networks often derive false conclusions about their quantizations, both when determining robustness and when detecting attacks, and that existing methods for quantized networks often miss attacks. Furthermore, we applied our method beyond robustness, showing how the number of bits in quantization enlarges the gender bias of a predictor for students’ grades.
spellingShingle Giacobbe, M
Henzinger, TA
Lechner, M
How many bits does it take to quantize your neural network?
title How many bits does it take to quantize your neural network?
title_full How many bits does it take to quantize your neural network?
title_fullStr How many bits does it take to quantize your neural network?
title_full_unstemmed How many bits does it take to quantize your neural network?
title_short How many bits does it take to quantize your neural network?
title_sort how many bits does it take to quantize your neural network
work_keys_str_mv AT giacobbem howmanybitsdoesittaketoquantizeyourneuralnetwork
AT henzingerta howmanybitsdoesittaketoquantizeyourneuralnetwork
AT lechnerm howmanybitsdoesittaketoquantizeyourneuralnetwork