Memory Faults in Activation-sparse Quantized Deep Neural Networks: Analysis and Mitigation using Sharpness-aware Training Article Swipe
Akul Malhotra
,
Sumeet Kumar Gupta
·
YOU?
·
· 2024
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2406.10528
YOU?
·
· 2024
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2406.10528
Improving the hardware efficiency of deep neural network (DNN) accelerators with techniques such as quantization and sparsity enhancement have shown an immense promise. However, their inference accuracy in non-ideal real-world settings (such as in the presence of hardware faults) is yet to be systematically analyzed. In this work, we investigate the impact of memory faults on activation-sparse quantized DNNs (AS QDNNs). We show that a high level of activation sparsity comes at the cost of larger vulnerability to faults, with AS QDNNs exhibiting up to 11.13% lower accuracy than the standard QDNNs. We establish that the degraded accuracy correlates with a sharper minima in the loss landscape for AS QDNNs, which makes them more sensitive to perturbations in the weight values due to faults. Based on this observation, we employ sharpness-aware quantization (SAQ) training to mitigate the impact of memory faults. The AS and standard QDNNs trained with SAQ have up to 19.50% and 15.82% higher inference accuracy, respectively compared to their conventionally trained equivalents. Moreover, we show that SAQ-trained AS QDNNs show higher accuracy in faulty settings than standard QDNNs trained conventionally. Thus, sharpness-aware training can be instrumental in achieving sparsity-related latency benefits without compromising on fault tolerance.
Related Topics
Concepts
Metadata
- Type
- preprint
- Language
- en
- Landing Page
- http://arxiv.org/abs/2406.10528
- https://arxiv.org/pdf/2406.10528
- OA Status
- green
- Related Works
- 10
- OpenAlex ID
- https://openalex.org/W4399794287
All OpenAlex metadata
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4399794287Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2406.10528Digital Object Identifier
- Title
-
Memory Faults in Activation-sparse Quantized Deep Neural Networks: Analysis and Mitigation using Sharpness-aware TrainingWork title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2024Year of publication
- Publication date
-
2024-06-15Full publication date if available
- Authors
-
Akul Malhotra, Sumeet Kumar GuptaList of authors in order
- Landing page
-
https://arxiv.org/abs/2406.10528Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2406.10528Direct link to full text PDF
- Open access
-
YesWhether a free full text is available
- OA status
-
greenOpen access status per OpenAlex
- OA URL
-
https://arxiv.org/pdf/2406.10528Direct OA link when available
- Concepts
-
Training (meteorology), Deep neural networks, Computer science, Artificial neural network, Artificial intelligence, Pattern recognition (psychology), Geography, MeteorologyTop concepts (fields/topics) attached by OpenAlex
- Cited by
-
0Total citation count in OpenAlex
- Related works (count)
-
10Other works algorithmically related by OpenAlex
Full payload
| id | https://openalex.org/W4399794287 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2406.10528 |
| ids.doi | https://doi.org/10.48550/arxiv.2406.10528 |
| ids.openalex | https://openalex.org/W4399794287 |
| fwci | |
| type | preprint |
| title | Memory Faults in Activation-sparse Quantized Deep Neural Networks: Analysis and Mitigation using Sharpness-aware Training |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| topics[0].id | https://openalex.org/T12676 |
| topics[0].field.id | https://openalex.org/fields/17 |
| topics[0].field.display_name | Computer Science |
| topics[0].score | 0.9850999712944031 |
| topics[0].domain.id | https://openalex.org/domains/3 |
| topics[0].domain.display_name | Physical Sciences |
| topics[0].subfield.id | https://openalex.org/subfields/1702 |
| topics[0].subfield.display_name | Artificial Intelligence |
| topics[0].display_name | Machine Learning and ELM |
| topics[1].id | https://openalex.org/T10036 |
| topics[1].field.id | https://openalex.org/fields/17 |
| topics[1].field.display_name | Computer Science |
| topics[1].score | 0.9715999960899353 |
| topics[1].domain.id | https://openalex.org/domains/3 |
| topics[1].domain.display_name | Physical Sciences |
| topics[1].subfield.id | https://openalex.org/subfields/1707 |
| topics[1].subfield.display_name | Computer Vision and Pattern Recognition |
| topics[1].display_name | Advanced Neural Network Applications |
| topics[2].id | https://openalex.org/T10320 |
| topics[2].field.id | https://openalex.org/fields/17 |
| topics[2].field.display_name | Computer Science |
| topics[2].score | 0.9679999947547913 |
| topics[2].domain.id | https://openalex.org/domains/3 |
| topics[2].domain.display_name | Physical Sciences |
| topics[2].subfield.id | https://openalex.org/subfields/1702 |
| topics[2].subfield.display_name | Artificial Intelligence |
| topics[2].display_name | Neural Networks and Applications |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| concepts[0].id | https://openalex.org/C2777211547 |
| concepts[0].level | 2 |
| concepts[0].score | 0.7062739133834839 |
| concepts[0].wikidata | https://www.wikidata.org/wiki/Q17141490 |
| concepts[0].display_name | Training (meteorology) |
| concepts[1].id | https://openalex.org/C2984842247 |
| concepts[1].level | 3 |
| concepts[1].score | 0.6067180633544922 |
| concepts[1].wikidata | https://www.wikidata.org/wiki/Q197536 |
| concepts[1].display_name | Deep neural networks |
| concepts[2].id | https://openalex.org/C41008148 |
| concepts[2].level | 0 |
| concepts[2].score | 0.5912292003631592 |
| concepts[2].wikidata | https://www.wikidata.org/wiki/Q21198 |
| concepts[2].display_name | Computer science |
| concepts[3].id | https://openalex.org/C50644808 |
| concepts[3].level | 2 |
| concepts[3].score | 0.5770753026008606 |
| concepts[3].wikidata | https://www.wikidata.org/wiki/Q192776 |
| concepts[3].display_name | Artificial neural network |
| concepts[4].id | https://openalex.org/C154945302 |
| concepts[4].level | 1 |
| concepts[4].score | 0.4831685423851013 |
| concepts[4].wikidata | https://www.wikidata.org/wiki/Q11660 |
| concepts[4].display_name | Artificial intelligence |
| concepts[5].id | https://openalex.org/C153180895 |
| concepts[5].level | 2 |
| concepts[5].score | 0.3234529495239258 |
| concepts[5].wikidata | https://www.wikidata.org/wiki/Q7148389 |
| concepts[5].display_name | Pattern recognition (psychology) |
| concepts[6].id | https://openalex.org/C205649164 |
| concepts[6].level | 0 |
| concepts[6].score | 0.07160419225692749 |
| concepts[6].wikidata | https://www.wikidata.org/wiki/Q1071 |
| concepts[6].display_name | Geography |
| concepts[7].id | https://openalex.org/C153294291 |
| concepts[7].level | 1 |
| concepts[7].score | 0.0 |
| concepts[7].wikidata | https://www.wikidata.org/wiki/Q25261 |
| concepts[7].display_name | Meteorology |
| keywords[0].id | https://openalex.org/keywords/training |
| keywords[0].score | 0.7062739133834839 |
| keywords[0].display_name | Training (meteorology) |
| keywords[1].id | https://openalex.org/keywords/deep-neural-networks |
| keywords[1].score | 0.6067180633544922 |
| keywords[1].display_name | Deep neural networks |
| keywords[2].id | https://openalex.org/keywords/computer-science |
| keywords[2].score | 0.5912292003631592 |
| keywords[2].display_name | Computer science |
| keywords[3].id | https://openalex.org/keywords/artificial-neural-network |
| keywords[3].score | 0.5770753026008606 |
| keywords[3].display_name | Artificial neural network |
| keywords[4].id | https://openalex.org/keywords/artificial-intelligence |
| keywords[4].score | 0.4831685423851013 |
| keywords[4].display_name | Artificial intelligence |
| keywords[5].id | https://openalex.org/keywords/pattern-recognition |
| keywords[5].score | 0.3234529495239258 |
| keywords[5].display_name | Pattern recognition (psychology) |
| keywords[6].id | https://openalex.org/keywords/geography |
| keywords[6].score | 0.07160419225692749 |
| keywords[6].display_name | Geography |
| language | en |
| locations[0].id | pmh:oai:arXiv.org:2406.10528 |
| locations[0].is_oa | True |
| locations[0].source.id | https://openalex.org/S4306400194 |
| locations[0].source.issn | |
| locations[0].source.type | repository |
| locations[0].source.is_oa | True |
| locations[0].source.issn_l | |
| locations[0].source.is_core | False |
| locations[0].source.is_in_doaj | False |
| locations[0].source.display_name | arXiv (Cornell University) |
| locations[0].source.host_organization | https://openalex.org/I205783295 |
| locations[0].source.host_organization_name | Cornell University |
| locations[0].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[0].license | |
| locations[0].pdf_url | https://arxiv.org/pdf/2406.10528 |
| locations[0].version | submittedVersion |
| locations[0].raw_type | text |
| locations[0].license_id | |
| locations[0].is_accepted | False |
| locations[0].is_published | False |
| locations[0].raw_source_name | |
| locations[0].landing_page_url | http://arxiv.org/abs/2406.10528 |
| locations[1].id | doi:10.48550/arxiv.2406.10528 |
| locations[1].is_oa | True |
| locations[1].source.id | https://openalex.org/S4306400194 |
| locations[1].source.issn | |
| locations[1].source.type | repository |
| locations[1].source.is_oa | True |
| locations[1].source.issn_l | |
| locations[1].source.is_core | False |
| locations[1].source.is_in_doaj | False |
| locations[1].source.display_name | arXiv (Cornell University) |
| locations[1].source.host_organization | https://openalex.org/I205783295 |
| locations[1].source.host_organization_name | Cornell University |
| locations[1].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[1].license | |
| locations[1].pdf_url | |
| locations[1].version | |
| locations[1].raw_type | article |
| locations[1].license_id | |
| locations[1].is_accepted | False |
| locations[1].is_published | |
| locations[1].raw_source_name | |
| locations[1].landing_page_url | https://doi.org/10.48550/arxiv.2406.10528 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5035901172 |
| authorships[0].author.orcid | https://orcid.org/0000-0002-6152-2377 |
| authorships[0].author.display_name | Akul Malhotra |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | Malhotra, Akul |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5044276472 |
| authorships[1].author.orcid | https://orcid.org/0000-0001-5609-9722 |
| authorships[1].author.display_name | Sumeet Kumar Gupta |
| authorships[1].author_position | last |
| authorships[1].raw_author_name | Gupta, Sumeet Kumar |
| authorships[1].is_corresponding | False |
| has_content.pdf | False |
| has_content.grobid_xml | False |
| is_paratext | False |
| open_access.is_oa | True |
| open_access.oa_url | https://arxiv.org/pdf/2406.10528 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2024-06-19T00:00:00 |
| display_name | Memory Faults in Activation-sparse Quantized Deep Neural Networks: Analysis and Mitigation using Sharpness-aware Training |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-11-06T06:51:31.235846 |
| primary_topic.id | https://openalex.org/T12676 |
| primary_topic.field.id | https://openalex.org/fields/17 |
| primary_topic.field.display_name | Computer Science |
| primary_topic.score | 0.9850999712944031 |
| primary_topic.domain.id | https://openalex.org/domains/3 |
| primary_topic.domain.display_name | Physical Sciences |
| primary_topic.subfield.id | https://openalex.org/subfields/1702 |
| primary_topic.subfield.display_name | Artificial Intelligence |
| primary_topic.display_name | Machine Learning and ELM |
| related_works | https://openalex.org/W230091440, https://openalex.org/W2233261550, https://openalex.org/W2810751659, https://openalex.org/W258997015, https://openalex.org/W2997094352, https://openalex.org/W3216976533, https://openalex.org/W100620283, https://openalex.org/W2495260952, https://openalex.org/W2033914206, https://openalex.org/W2042327336 |
| cited_by_count | 0 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2406.10528 |
| best_oa_location.is_oa | True |
| best_oa_location.source.id | https://openalex.org/S4306400194 |
| best_oa_location.source.issn | |
| best_oa_location.source.type | repository |
| best_oa_location.source.is_oa | True |
| best_oa_location.source.issn_l | |
| best_oa_location.source.is_core | False |
| best_oa_location.source.is_in_doaj | False |
| best_oa_location.source.display_name | arXiv (Cornell University) |
| best_oa_location.source.host_organization | https://openalex.org/I205783295 |
| best_oa_location.source.host_organization_name | Cornell University |
| best_oa_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| best_oa_location.license | |
| best_oa_location.pdf_url | https://arxiv.org/pdf/2406.10528 |
| best_oa_location.version | submittedVersion |
| best_oa_location.raw_type | text |
| best_oa_location.license_id | |
| best_oa_location.is_accepted | False |
| best_oa_location.is_published | False |
| best_oa_location.raw_source_name | |
| best_oa_location.landing_page_url | http://arxiv.org/abs/2406.10528 |
| primary_location.id | pmh:oai:arXiv.org:2406.10528 |
| primary_location.is_oa | True |
| primary_location.source.id | https://openalex.org/S4306400194 |
| primary_location.source.issn | |
| primary_location.source.type | repository |
| primary_location.source.is_oa | True |
| primary_location.source.issn_l | |
| primary_location.source.is_core | False |
| primary_location.source.is_in_doaj | False |
| primary_location.source.display_name | arXiv (Cornell University) |
| primary_location.source.host_organization | https://openalex.org/I205783295 |
| primary_location.source.host_organization_name | Cornell University |
| primary_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| primary_location.license | |
| primary_location.pdf_url | https://arxiv.org/pdf/2406.10528 |
| primary_location.version | submittedVersion |
| primary_location.raw_type | text |
| primary_location.license_id | |
| primary_location.is_accepted | False |
| primary_location.is_published | False |
| primary_location.raw_source_name | |
| primary_location.landing_page_url | http://arxiv.org/abs/2406.10528 |
| publication_date | 2024-06-15 |
| publication_year | 2024 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 64, 100 |
| abstract_inverted_index.AS | 80, 108, 142, 170 |
| abstract_inverted_index.In | 45 |
| abstract_inverted_index.We | 61, 92 |
| abstract_inverted_index.an | 20 |
| abstract_inverted_index.as | 13, 32 |
| abstract_inverted_index.at | 71 |
| abstract_inverted_index.be | 42, 187 |
| abstract_inverted_index.in | 27, 33, 103, 117, 175, 189 |
| abstract_inverted_index.is | 39 |
| abstract_inverted_index.of | 4, 36, 52, 67, 74, 138 |
| abstract_inverted_index.on | 55, 125, 196 |
| abstract_inverted_index.to | 41, 77, 84, 115, 122, 134, 151, 160 |
| abstract_inverted_index.up | 83, 150 |
| abstract_inverted_index.we | 48, 128, 166 |
| abstract_inverted_index.(AS | 59 |
| abstract_inverted_index.SAQ | 148 |
| abstract_inverted_index.The | 141 |
| abstract_inverted_index.and | 15, 143, 153 |
| abstract_inverted_index.can | 186 |
| abstract_inverted_index.due | 121 |
| abstract_inverted_index.for | 107 |
| abstract_inverted_index.the | 1, 34, 50, 72, 89, 95, 104, 118, 136 |
| abstract_inverted_index.yet | 40 |
| abstract_inverted_index.DNNs | 58 |
| abstract_inverted_index.cost | 73 |
| abstract_inverted_index.deep | 5 |
| abstract_inverted_index.have | 18, 149 |
| abstract_inverted_index.high | 65 |
| abstract_inverted_index.loss | 105 |
| abstract_inverted_index.more | 113 |
| abstract_inverted_index.show | 62, 167, 172 |
| abstract_inverted_index.such | 12 |
| abstract_inverted_index.than | 88, 178 |
| abstract_inverted_index.that | 63, 94, 168 |
| abstract_inverted_index.them | 112 |
| abstract_inverted_index.this | 46, 126 |
| abstract_inverted_index.with | 10, 79, 99, 147 |
| abstract_inverted_index.(DNN) | 8 |
| abstract_inverted_index.(SAQ) | 132 |
| abstract_inverted_index.(such | 31 |
| abstract_inverted_index.Based | 124 |
| abstract_inverted_index.QDNNs | 81, 145, 171, 180 |
| abstract_inverted_index.Thus, | 183 |
| abstract_inverted_index.comes | 70 |
| abstract_inverted_index.fault | 197 |
| abstract_inverted_index.level | 66 |
| abstract_inverted_index.lower | 86 |
| abstract_inverted_index.makes | 111 |
| abstract_inverted_index.shown | 19 |
| abstract_inverted_index.their | 24, 161 |
| abstract_inverted_index.which | 110 |
| abstract_inverted_index.work, | 47 |
| abstract_inverted_index.11.13% | 85 |
| abstract_inverted_index.15.82% | 154 |
| abstract_inverted_index.19.50% | 152 |
| abstract_inverted_index.QDNNs, | 109 |
| abstract_inverted_index.QDNNs. | 91 |
| abstract_inverted_index.employ | 129 |
| abstract_inverted_index.faults | 54 |
| abstract_inverted_index.faulty | 176 |
| abstract_inverted_index.higher | 155, 173 |
| abstract_inverted_index.impact | 51, 137 |
| abstract_inverted_index.larger | 75 |
| abstract_inverted_index.memory | 53, 139 |
| abstract_inverted_index.minima | 102 |
| abstract_inverted_index.neural | 6 |
| abstract_inverted_index.values | 120 |
| abstract_inverted_index.weight | 119 |
| abstract_inverted_index.QDNNs). | 60 |
| abstract_inverted_index.faults) | 38 |
| abstract_inverted_index.faults, | 78 |
| abstract_inverted_index.faults. | 123, 140 |
| abstract_inverted_index.immense | 21 |
| abstract_inverted_index.latency | 192 |
| abstract_inverted_index.network | 7 |
| abstract_inverted_index.sharper | 101 |
| abstract_inverted_index.trained | 146, 163, 181 |
| abstract_inverted_index.without | 194 |
| abstract_inverted_index.However, | 23 |
| abstract_inverted_index.accuracy | 26, 87, 97, 174 |
| abstract_inverted_index.benefits | 193 |
| abstract_inverted_index.compared | 159 |
| abstract_inverted_index.degraded | 96 |
| abstract_inverted_index.hardware | 2, 37 |
| abstract_inverted_index.mitigate | 135 |
| abstract_inverted_index.presence | 35 |
| abstract_inverted_index.promise. | 22 |
| abstract_inverted_index.settings | 30, 177 |
| abstract_inverted_index.sparsity | 16, 69 |
| abstract_inverted_index.standard | 90, 144, 179 |
| abstract_inverted_index.training | 133, 185 |
| abstract_inverted_index.Improving | 0 |
| abstract_inverted_index.Moreover, | 165 |
| abstract_inverted_index.accuracy, | 157 |
| abstract_inverted_index.achieving | 190 |
| abstract_inverted_index.analyzed. | 44 |
| abstract_inverted_index.establish | 93 |
| abstract_inverted_index.inference | 25, 156 |
| abstract_inverted_index.landscape | 106 |
| abstract_inverted_index.non-ideal | 28 |
| abstract_inverted_index.quantized | 57 |
| abstract_inverted_index.sensitive | 114 |
| abstract_inverted_index.activation | 68 |
| abstract_inverted_index.correlates | 98 |
| abstract_inverted_index.efficiency | 3 |
| abstract_inverted_index.exhibiting | 82 |
| abstract_inverted_index.real-world | 29 |
| abstract_inverted_index.techniques | 11 |
| abstract_inverted_index.tolerance. | 198 |
| abstract_inverted_index.SAQ-trained | 169 |
| abstract_inverted_index.enhancement | 17 |
| abstract_inverted_index.investigate | 49 |
| abstract_inverted_index.accelerators | 9 |
| abstract_inverted_index.compromising | 195 |
| abstract_inverted_index.equivalents. | 164 |
| abstract_inverted_index.instrumental | 188 |
| abstract_inverted_index.observation, | 127 |
| abstract_inverted_index.quantization | 14, 131 |
| abstract_inverted_index.respectively | 158 |
| abstract_inverted_index.perturbations | 116 |
| abstract_inverted_index.vulnerability | 76 |
| abstract_inverted_index.conventionally | 162 |
| abstract_inverted_index.systematically | 43 |
| abstract_inverted_index.conventionally. | 182 |
| abstract_inverted_index.sharpness-aware | 130, 184 |
| abstract_inverted_index.sparsity-related | 191 |
| abstract_inverted_index.activation-sparse | 56 |
| cited_by_percentile_year | |
| countries_distinct_count | 0 |
| institutions_distinct_count | 2 |
| citation_normalized_percentile |