Models - Model Parameters Compatibility

This documentation shows you on which algorithms can be switched with others.

However, it assumes that you will ignore the difference in properties, hyperparameters and data distribution that the model trained on.

Full Compatibility

The algorithms under the same group can have its model parameters swapped without additional changes.

Closed-Form Linear Regression

  • BayesianLinearRegression
  • BayesianQuantileLinearRegression

Gradient-Based

  • LinearRegression
  • QuantileRegression
  • PoissonRegression
  • NegativeBinomialRegression
  • GammaRegression
  • SupportVectorRegressionGradientVariant
  • PassiveAggressiveRegressor
  • BinaryRegression
  • SupportVectorMachineGradientVariant
  • PassiveAggressiveClassifier
  • OneClassPassiveAggressiveClassifier

Naive Bayes

  • All except for CategoricalNaiveBayes

K-Nearest Neigbours

  • All

Mean-Based With Data Point Number

  • NearestCentroids
  • KMeans

Deep Reinforcement Learning (Single)

  • All

Deep Reinforcement Learning (Actor-Critic)

  • All

Tabular Reinforcement Learning

  • All

Generative Adversarial Network

  • All

Outlier Detection

  • All

Partial Compatibility

The algorithms under the same group can have its model parameters swapped with some changes to the model parameters. Refer to the models’ API reference for more information.

Statistical-Based Clustering

  • MeanShift
  • ExpectedMaximization
  • FuzzyCMeans

Kalman Filters

  • All