Image Classification on MNIST (Error Rate)
0.42Error Rate4l-spec-cp
Evaluation results
| Method | Links | |
|---|---|---|
| 4l-spec-cpDomain=local k-NN graph, Kernel=spectral graph conv2018.03 | 0.42 | |
| Network in NetworkDomain=local img patch, Kernel=spatial conv2018.03 | 0.47 | |
| 3l-pointnet++Domain=local points, Kernel=spatial point-MLP2018.03 | 0.55 | |
| LeNet5Domain=local img patch, Kernel=spatial conv2018.03 | 0.8 | |
| MoNetDomain=local graph, Kernel=spatial graph conv2018.03 | 0.81 | |
| ChebNetDomain=full graph, Kernel=spectral graph conv2018.03 | 0.86 | |
| Relational Variational Autoencoder (RVAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 0.9 | |
| Relational Denoising Autoencoder (RDAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 1.1 | |
| Variational Autoencoder (VAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 1.2 | |
| Denoising Autoencoder (DAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 1.6 | |
| Multi-layer perceptronDomain=full image, Kernel=spatial MLP2018.03 | 1.6 | |
| Relational Sparse Autoencoder (RSAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 1.8 | |
| Sparse Autoencoder (SAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 2.2 | |
| Relational Autoencoder (RAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 3.8 | |
| Generative Autoencoder (GAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 5.7 | |
| Basic Autoencoder (BAE)Validation Strategy=10-fold cross validation, Classifier=Softmax regression2018.02 | 8.9 | |
| CAMNet3Params(M)=2.02019.07 | 22 | |
| Multi-column DNN2018.03 | 23 | |
| CAMNet2Params(M)=1.22019.07 | 25 | |
| SENetParams(M)=0.52019.07 | 25 | |
| CAMNet4Params(M)=3.02019.07 | 26 | |
| BaseCNNParams(M)=0.52019.07 | 30 | |
| MultiCNN3Params(M)=1.52019.07 | 30 | |
| tinyCAMNet3Params(M)=0.472019.07 | 32 | |
| Cr-Stitch2Params(M)=12019.07 | 33 | |
| SO-Net (2-layer)Representation=points, Input=512 x 2, Pre-training=false2018.03 | 44 | |
| SO-Net2018.03 | 44 | |
| Network in Network2018.03 | 47 | |
| MultiCNN3Params(M)=1.5, Data Augmentation=None2019.07 | 48 | |
| PointNet++Representation=points + normal, Input=512 x 22018.03 | 51 | |
| PointNet++2018.03 | 51 | |
| CAMNet3Params(M)=2.0, Data Augmentation=None2019.07 | 53 | |
| ECCRepresentation=points2018.03 | 63 | |
| ECC2018.03 | 63 | |
| PointNetRepresentation=points, Input=256 x 22018.03 | 78 | |
| PointNet2018.03 | 78 | |
| LeNet52018.03 | 80 | |
| Kd-NetRepresentation=points, Input=1024 x 22018.03 | 90 | |
| Kd-Net2018.03 | 90 | |
| Multi-layer perceptron2018.03 | 160 |