Table 15 Comparison of the accuracy and training times for the best found ELM and BP solutions for given image classification datasets
From: Extreme learning machine versus classical feedforward network
Dataset | Model | Accuracy (%) | Training time\(^2\) |
|---|---|---|---|
MNIST | LRF-ELM (own) | 98.85 | 1 min |
Branching and merging CNN with homogeneous filter capsules [4] | 99.79 | N/A | |
CIFAR-10 | LRF-ELM (own) | 58.63 | 1 min |
TResNet-XL [38] | 99.00 | Around 1 h | |
NORB | LRF-ELM (own) | 92.61 | 15 s |
VB-Routing CapsuleNet [37] | 98.40 | N/A | |
GTSRB | LRF-ELM (own) | 92.80 | 40 s |
Spatial transformer [1] | 99.71 | Around 1.5 h | |
ImageNet16 | LRF-ELM (own) | 4.91 | 1.5 min |
Wide ResNet-20, width 10 [6] | 40.06 | 2.7 days | |
ImageNet32 | LRF-ELM (own) | 3.07 | 4 min |
Wide ResNet-28, width 10 [6] | 59.04 | 13.8 days |