Efficient shallow learning as an alternative to deep learning
Abstract The realization of complex classification tasks requires training of deep learning (DL) architectures consisting of tens or even hundreds of convolutional and fully connected hidden layers, which is far from the reality of the human brain. According to the DL rationale, the first convolutio...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-04-01
|
| Series: | Scientific Reports |
| Online Access: | https://doi.org/10.1038/s41598-023-32559-8 |
| _version_ | 1797840991213846528 |
|---|---|
| author | Yuval Meir Ofek Tevet Yarden Tzach Shiri Hodassman Ronit D. Gross Ido Kanter |
| author_facet | Yuval Meir Ofek Tevet Yarden Tzach Shiri Hodassman Ronit D. Gross Ido Kanter |
| author_sort | Yuval Meir |
| collection | DOAJ |
| description | Abstract The realization of complex classification tasks requires training of deep learning (DL) architectures consisting of tens or even hundreds of convolutional and fully connected hidden layers, which is far from the reality of the human brain. According to the DL rationale, the first convolutional layer reveals localized patterns in the input and large-scale patterns in the following layers, until it reliably characterizes a class of inputs. Here, we demonstrate that with a fixed ratio between the depths of the first and second convolutional layers, the error rates of the generalized shallow LeNet architecture, consisting of only five layers, decay as a power law with the number of filters in the first convolutional layer. The extrapolation of this power law indicates that the generalized LeNet can achieve small error rates that were previously obtained for the CIFAR-10 database using DL architectures. A power law with a similar exponent also characterizes the generalized VGG-16 architecture. However, this results in a significantly increased number of operations required to achieve a given error rate with respect to LeNet. This power law phenomenon governs various generalized LeNet and VGG-16 architectures, hinting at its universal behavior and suggesting a quantitative hierarchical time–space complexity among machine learning architectures. Additionally, the conservation law along the convolutional layers, which is the square-root of their size times their depth, is found to asymptotically minimize error rates. The efficient shallow learning that is demonstrated in this study calls for further quantitative examination using various databases and architectures and its accelerated implementation using future dedicated hardware developments. |
| first_indexed | 2024-04-09T16:23:40Z |
| format | Article |
| id | doaj.art-300be2b85aec4c93b5d5055b0b71e259 |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-04-09T16:23:40Z |
| publishDate | 2023-04-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-300be2b85aec4c93b5d5055b0b71e2592023-04-23T11:18:09ZengNature PortfolioScientific Reports2045-23222023-04-011311710.1038/s41598-023-32559-8Efficient shallow learning as an alternative to deep learningYuval Meir0Ofek Tevet1Yarden Tzach2Shiri Hodassman3Ronit D. Gross4Ido Kanter5Department of Physics, Bar-Ilan UniversityDepartment of Physics, Bar-Ilan UniversityDepartment of Physics, Bar-Ilan UniversityDepartment of Physics, Bar-Ilan UniversityDepartment of Physics, Bar-Ilan UniversityDepartment of Physics, Bar-Ilan UniversityAbstract The realization of complex classification tasks requires training of deep learning (DL) architectures consisting of tens or even hundreds of convolutional and fully connected hidden layers, which is far from the reality of the human brain. According to the DL rationale, the first convolutional layer reveals localized patterns in the input and large-scale patterns in the following layers, until it reliably characterizes a class of inputs. Here, we demonstrate that with a fixed ratio between the depths of the first and second convolutional layers, the error rates of the generalized shallow LeNet architecture, consisting of only five layers, decay as a power law with the number of filters in the first convolutional layer. The extrapolation of this power law indicates that the generalized LeNet can achieve small error rates that were previously obtained for the CIFAR-10 database using DL architectures. A power law with a similar exponent also characterizes the generalized VGG-16 architecture. However, this results in a significantly increased number of operations required to achieve a given error rate with respect to LeNet. This power law phenomenon governs various generalized LeNet and VGG-16 architectures, hinting at its universal behavior and suggesting a quantitative hierarchical time–space complexity among machine learning architectures. Additionally, the conservation law along the convolutional layers, which is the square-root of their size times their depth, is found to asymptotically minimize error rates. The efficient shallow learning that is demonstrated in this study calls for further quantitative examination using various databases and architectures and its accelerated implementation using future dedicated hardware developments.https://doi.org/10.1038/s41598-023-32559-8 |
| spellingShingle | Yuval Meir Ofek Tevet Yarden Tzach Shiri Hodassman Ronit D. Gross Ido Kanter Efficient shallow learning as an alternative to deep learning Scientific Reports |
| title | Efficient shallow learning as an alternative to deep learning |
| title_full | Efficient shallow learning as an alternative to deep learning |
| title_fullStr | Efficient shallow learning as an alternative to deep learning |
| title_full_unstemmed | Efficient shallow learning as an alternative to deep learning |
| title_short | Efficient shallow learning as an alternative to deep learning |
| title_sort | efficient shallow learning as an alternative to deep learning |
| url | https://doi.org/10.1038/s41598-023-32559-8 |
| work_keys_str_mv | AT yuvalmeir efficientshallowlearningasanalternativetodeeplearning AT ofektevet efficientshallowlearningasanalternativetodeeplearning AT yardentzach efficientshallowlearningasanalternativetodeeplearning AT shirihodassman efficientshallowlearningasanalternativetodeeplearning AT ronitdgross efficientshallowlearningasanalternativetodeeplearning AT idokanter efficientshallowlearningasanalternativetodeeplearning |