Online Retraining
Select a previously trained model.
Train a new predictive model for Sleep Staging or Sleep-Disordered Breathing. While training is in progress, you may navigate away from the page or close your browser. Trained models may be saved for later use in the Analysis section of the platform.
Select an existing model to test on out-of-sample data, or configure and train a new model.
Sleep Staging
Configuration
Select between training a binary classifier to predict whether the AHI of a child is ≥5 or to classify recording as Sleep or Wake, or a regressor to produce a point estimate of the AHI.
Select a Linear, Gaussian, or Polynomial kernel. The Linear kernel allows for better model interpretability, while the Gaussian kernel is more flexible and may capture non-linearities. Polynomial kernels offer a balance in complexity and the degree of non-linearity is configurable.Cost-sensitive Support Vector Classification may be used for imbalanced datasets. When different costs are applied, the resulting box constraints are rescaled so that the average matches the C value chosen during hyperparameter tuning.
Model Information
Feature Selection
| Feature | Description | |
|---|---|---|
| Epoch Fraction | Fraction of recording elapsed. | |
| SpO₂ Mean Difference | Difference between the mean SpO₂ of the current epoch compared to the complete recording. | |
| SpO₂ Variance Difference | Difference between the second central moment of the SpO₂ in the current epoch compared to the complete recording. | |
| SpO₂ SD1/SD2 Difference | Difference between the ratio of short-term and long-term variability in the SpO₂ of the current epoch compared to the complete recording. | |
| SpO₂ Mean | Mean of the SpO₂ in the current epoch. | |
| SpO₂ Variance | Second central moment of the SpO₂ in the current epoch. | |
| SpO₂ Spread | Spread (maximum - minimum) of the SpO₂. | |
| SpO₂ SD1 | Standard deviation of the first difference in the SpO₂. A measure of short-term SpO₂ variability. | |
| SpO₂ SD2 | Long-axis spread of the SpO₂ Poincaré plot, reflecting longer-term SpO₂ variability. | |
| SpO₂ SD1/SD2 | Ratio of short-term and long-term variability in the SpO₂. | |
| Pulse Rate Mean Difference | Difference between the mean Pulse Rate of the current epoch compared to the complete recording. | |
| Pulse Rate Variance Difference | Difference between the second central moment of the Pulse Rate in the current epoch compared to the complete recording. | |
| Pulse Rate SD1/SD2 Difference | Difference between the ratio of short-term and long-term variability in the Pulse Rate of the current epoch compared to the complete recording. | |
| Pulse Rate Mean | Mean of the Pulse Rate in the current epoch. | |
| Pulse Rate Variance | Second central moment of the Pulse Rate in the current epoch. | |
| Pulse Rate Spread | Spread (maximum - minimum) of the Pulse Rate. | |
| Pulse Rate SD1 | Standard deviation of the first difference in the Pulse Rate. A measure of short-term Pulse Rate variability. | |
| Pulse Rate SD2 | Long-axis spread of the Pulse Rate Poincaré plot, reflecting longer-term Pulse Rate variability. | |
| Pulse Rate SD1/SD2 | Ratio of short-term and long-term variability in the Pulse Rate. |
Select a precomputed dataset to use for model training. Only datasets specifically configured for training are available. Cross-validation results are available during model testing.
Bayesian Optimization
Number of objective function evaluations.
Number of candidates evaluated per iteration.
Configure the Bayesian Optimizer and search space for each hyperparameter. The number of iterations determines how many points are evaluated, and the number of candidates controls the resolution of the acquisition search.
Training Progress
Finished
Best Parameters: C=0.28
Selected Parameters: C=0.28
Bayesian Optimization is used to search the hyperparameter ranges provided. This process attempts to balance exploration and exploitation to select the optimal values for the trained model. Gaussian Process Regression (GPR) and the Expected Improvement acquisition function are used.
Model Validation data are collected from out-of-fold testing during cross-validation. These data are saved with the model and do not change.
Models are calibrated using cross-validation data when training is complete so that predictions are accompanied by estimates of uncertainty. A trained Ridge Logistic Regression model is embedded within Support Vector Classifiers to produce a probability point-estimate during prediction, and a Laplace distribution is computed and embedded within Linear and Support Vector Regression to produce confidence intervals that accompany the AHI point-estimate during prediction.
Trained models may be saved and are immediately accessible through the Analysis and Batch Analysis areas of the platform.
Select a precomputed dataset to use for out-of-sample Model Testing. Out-of-sample data are recommended.