AutoNet-Generated Deep Layer-Wise Convex Networks for ECG Classification
The design of neural networks typically involves trial-and-error, a time-consuming process for obtaining an optimal architecture, even for experienced researchers. Additionally, it is widely accepted that loss functions of deep neural networks are generally non-convex with respect to the parameters to be optimised. We propose the Layer-wise Convex Theorem to ensure that the loss is convex with respect to the parameters of a given layer, achieved by constraining each layer to be an overdetermined system of non-linear equations. Based on this theorem, we developed an end-to-end algorithm (the AutoNet) to automatically generate layer-wise convex networks (LCNs) for any given training set. We then demonstrate the performance of the AutoNet-generated LCNs (AutoNet-LCNs) compared to state-of-the-art models on three electrocardiogram (ECG) classification benchmark datasets, with further validation on two non-ECG benchmark datasets for more general tasks. The AutoNet-LCN was able to find networks customised for each dataset without manual fine-tuning under 2 GPU-hours, and the resulting networks outperformed the state-of-the-art models with fewer than 5% parameters on all the above five benchmark datasets. The efficiency and robustness of the AutoNet-LCN markedly reduce model discovery costs and enable efficient training of deep learning models in resource-constrained settings.
Medienart: |
E-Artikel |
---|
Erscheinungsjahr: |
2024 |
---|---|
Erschienen: |
2024 |
Enthalten in: |
Zur Gesamtaufnahme - volume:PP |
---|---|
Enthalten in: |
IEEE transactions on pattern analysis and machine intelligence - PP(2024) vom: 21. März |
Sprache: |
Englisch |
---|
Beteiligte Personen: |
Shen, Yanting [VerfasserIn] |
---|
Links: |
---|
Themen: |
---|
Anmerkungen: |
Date Revised 21.03.2024 published: Print-Electronic Citation Status Publisher |
---|
doi: |
10.1109/TPAMI.2024.3378843 |
---|
funding: |
|
---|---|
Förderinstitution / Projekttitel: |
|
PPN (Katalog-ID): |
NLM370023080 |
---|
LEADER | 01000naa a22002652 4500 | ||
---|---|---|---|
001 | NLM370023080 | ||
003 | DE-627 | ||
005 | 20240323001738.0 | ||
007 | cr uuu---uuuuu | ||
008 | 240323s2024 xx |||||o 00| ||eng c | ||
024 | 7 | |a 10.1109/TPAMI.2024.3378843 |2 doi | |
028 | 5 | 2 | |a pubmed24n1341.xml |
035 | |a (DE-627)NLM370023080 | ||
035 | |a (NLM)38512733 | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
100 | 1 | |a Shen, Yanting |e verfasserin |4 aut | |
245 | 1 | 0 | |a AutoNet-Generated Deep Layer-Wise Convex Networks for ECG Classification |
264 | 1 | |c 2024 | |
336 | |a Text |b txt |2 rdacontent | ||
337 | |a ƒaComputermedien |b c |2 rdamedia | ||
338 | |a ƒa Online-Ressource |b cr |2 rdacarrier | ||
500 | |a Date Revised 21.03.2024 | ||
500 | |a published: Print-Electronic | ||
500 | |a Citation Status Publisher | ||
520 | |a The design of neural networks typically involves trial-and-error, a time-consuming process for obtaining an optimal architecture, even for experienced researchers. Additionally, it is widely accepted that loss functions of deep neural networks are generally non-convex with respect to the parameters to be optimised. We propose the Layer-wise Convex Theorem to ensure that the loss is convex with respect to the parameters of a given layer, achieved by constraining each layer to be an overdetermined system of non-linear equations. Based on this theorem, we developed an end-to-end algorithm (the AutoNet) to automatically generate layer-wise convex networks (LCNs) for any given training set. We then demonstrate the performance of the AutoNet-generated LCNs (AutoNet-LCNs) compared to state-of-the-art models on three electrocardiogram (ECG) classification benchmark datasets, with further validation on two non-ECG benchmark datasets for more general tasks. The AutoNet-LCN was able to find networks customised for each dataset without manual fine-tuning under 2 GPU-hours, and the resulting networks outperformed the state-of-the-art models with fewer than 5% parameters on all the above five benchmark datasets. The efficiency and robustness of the AutoNet-LCN markedly reduce model discovery costs and enable efficient training of deep learning models in resource-constrained settings | ||
650 | 4 | |a Journal Article | |
700 | 1 | |a Lu, Lei |e verfasserin |4 aut | |
700 | 1 | |a Zhu, Tingting |e verfasserin |4 aut | |
700 | 1 | |a Wang, Xinshao |e verfasserin |4 aut | |
700 | 1 | |a Clifton, Lei |e verfasserin |4 aut | |
700 | 1 | |a Chen, Zhengming |e verfasserin |4 aut | |
700 | 1 | |a Clarke, Robert |e verfasserin |4 aut | |
700 | 1 | |a Clifton, David A |e verfasserin |4 aut | |
773 | 0 | 8 | |i Enthalten in |t IEEE transactions on pattern analysis and machine intelligence |d 1979 |g PP(2024) vom: 21. März |w (DE-627)NLM098212257 |x 1939-3539 |7 nnns |
773 | 1 | 8 | |g volume:PP |g year:2024 |g day:21 |g month:03 |
856 | 4 | 0 | |u http://dx.doi.org/10.1109/TPAMI.2024.3378843 |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a GBV_NLM | ||
951 | |a AR | ||
952 | |d PP |j 2024 |b 21 |c 03 |