API Reference¶
This section provides detailed technical documentation for all public APIs in the Tabular SSL library.
Models¶
TabularSSL¶
The main class for self-supervised learning on tabular data.
from tabular_ssl import TabularSSL
Parameters¶
input_dim(int): Dimension of input featureshidden_dim(int, optional): Dimension of hidden layers. Defaults to 256.num_layers(int, optional): Number of transformer layers. Defaults to 4.num_heads(int, optional): Number of attention heads. Defaults to 4.dropout(float, optional): Dropout rate. Defaults to 0.1.mask_ratio(float, optional): Ratio of features to mask during training. Defaults to 0.15.
Methods¶
train(data, batch_size=32, epochs=100, learning_rate=1e-4)¶
Train the model using self-supervised learning.
Parameters:
- data (pd.DataFrame): Input data
- batch_size (int): Batch size for training
- epochs (int): Number of training epochs
- learning_rate (float): Learning rate
Returns:
- dict: Training history
predict(data)¶
Make predictions on new data.
Parameters:
- data (pd.DataFrame): Input data
Returns:
- np.ndarray: Model predictions
Data Utilities¶
DataLoader¶
Utility class for loading and preprocessing tabular data.
from tabular_ssl.data import DataLoader
Methods¶
load_data(file_path, target_col=None)¶
Load data from a file.
Parameters:
- file_path (str): Path to the data file
- target_col (str, optional): Name of the target column
Returns:
- pd.DataFrame: Loaded data
preprocess(data, categorical_cols=None)¶
Preprocess the data.
Parameters:
- data (pd.DataFrame): Input data
- categorical_cols (list, optional): List of categorical column names
Returns:
- pd.DataFrame: Preprocessed data
Utility Functions¶
Evaluation¶
from tabular_ssl.utils import evaluate_model
evaluate_model(model, test_data, metrics=['accuracy', 'f1'])¶
Evaluate model performance.
Parameters:
- model: Trained model
- test_data (pd.DataFrame): Test data
- metrics (list): List of metrics to compute
Returns:
- dict: Dictionary of metric scores
Visualization¶
from tabular_ssl.utils import plot_training_history
plot_training_history(history)¶
Plot training history.
Parameters:
- history (dict): Training history dictionary
Returns:
- matplotlib.figure.Figure: Plot figure