Home
Data Process
Genome
Transcriptome
Protein
Integration
Model
Train > supervised
Train > unsupervised
Train > semi-supervised
Use
Data Analyse
Help
``` ```
Train your semi-supervised model
Home
Model
Train
1. Select strategy and upload files
Upload train/validation/testing datasets
Upload a dataset to conduct train/validation/testing split
Use stratified K-Fold cross validation method
Upload training set (NOTE: The categorized labels of unlabeled samples are marked as Unknown)
Example
Upload validation set (NOTE: All samples must be labeled)
Upload testing set (NOTE: All samples must be labeled)
Upload a dataset (NOTE: The categorized labels of unlabeled samples are marked as Unknown)
Example
Input a ratio of training:validation:testing (NOTE: The separator is a colon ':')
Upload training set (NOTE: The categorized labels of unlabeled samples are marked as Unknown)
Example
Input K value (NOTE: If you want to use 3-Fold cross validation method, then K value is 3.)
Upload testing set (NOTE: All samples must be labeled)
2. Choose model
LadderNetwork > Utilizing unlabeled data through an encoder-decoder architecture
Pseudo-Labeling > Iteratively add labels to high-confidence unlabeled data
3. Upload training hyperparameters
(1) Epoch
(2) Learning Rate
(3) EarlyStopping Patience
(4) Batch Size
(5) Random Seed
(6) Loss Function
Semi-Supervised Loss
Focal Loss
Label Smoothing Loss
Contrastive Loss
Pseudo Labeling Loss
Focal Loss
Label Smoothing Loss
(7) Optimizer
Adam
SGD
AdamW
RMSprop
(8) Supervision Loss Weight
(9) Reconstruction Loss Weight
(10) Regularization Loss Weight
(9) Pseudo-label loss weight
(10) Pseudo-label confidence threshold
(11) Pseudo-label ratio
Your Email
Submit