Efficient Neural Network Training via Subset Pretraining
Journal
Proceedings of the 16th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management
Type
conference paper
Date Issued
2024-11-22
Author(s)
Abstract
In training neural networks, it is common practice to use partial gradients computed over batches, mostly very small subsets of the training set. This approach is motivated by the argument that such a partial gradient is close to the true one, with precision growing only with the square root of the batch size. A theoretical justification is with the help of stochastic approximation theory. However, the conditions for the validity of this theory are not satisfied in the usual learning rate schedules. Batch processing is also difficult to combine with efficient second-order optimization methods. This proposal is based on another hypothesis: the loss minimum of the training set can be expected to be well-approximated by the minima of its subsets. Such subset minima can be computed in a fraction of the time necessary for optimizing over the whole training set. This hypothesis has been tested with the help of the MNIST, CIFAR-10, and CIFAR-100 image classification benchmarks, optionally extended by training data augmentation. The experiments have confirmed that results equivalent to conventional training can be reached. In summary, even small subsets are representative if the overdetermination ratio for the given model parameter set sufficiently exceeds unity. The computing expense can be reduced to a tenth or less.
Language
English
Keywords
Deep Neural Network
Convolutional Network
Computer Vision
Efficient Training
Resource Optimization
Training Strategies
Overdetermination Ratio
Stochastic Approximation Theory
HSG Classification
contribution to scientific community
Publisher
SCITEPRESS - Science and Technology Publications
Start page
242
End page
249
Pages
8
Event Location
Porto, Portugal
Event Date
November 2024
Contact Email Address
bernhard.bermeitinger@unisg.ch
File(s)![Thumbnail Image]()
![Thumbnail Image]()
restricted
Name
KDIR2024 - Efficient Neural Network Training via Subset Pretraining.pdf
Size
1 MB
Format
Adobe PDF
Checksum (MD5)
249f96220ed6fd658c1c7679f65ed1b3
open.access
Name
KDIR2024 - Efficient Neural Network Training via Subset Pretraining - Slides.pdf
Size
1.05 MB
Format
Adobe PDF
Checksum (MD5)
4c75e24ab65efb1af550a8899a0b759a