Python package for TabSRA: An Attention based Self-Explainable Model for Tabular Learning
pip install skorch-tabsra==0.0.1
from tabsra.skorch_tabsra import InputShapeSetterTabSRA,TabSRALinearRegressor
TabSRA = TabSRALinearRegressor(**params,callbacks=callbacks)
_ = TabSRA.fit(X_train_,Y_train_)Parameters
module__dim_head : int, default=8. The attention head dimension , d_k in the paper. Typical values are {4,8,12}.
module__n_hidden_encoder : int, default=2. The number of hidden layers in in the Key/Query encoder. Typical values are {1,2}.
module__n_head : int, default=2. Number of SRA head/ensemble. Bigger value gives capacity to the model to produce less stable/robust explanations. Typical values are 1 or 2.
module__dropout_rate : float,default=0.0. The neuron dropout rate used in the Key/Query encorder during the training.
module__encoder_bias : bool. Whether to use bias term in the Key/Query encoder.
module__classifier_bias : bool, default=True. Whether to use bias term in the downstream linear classifier (regressor).
module__attention_scaler : str, default='sqrt'. The scaling function for the attention weights.
optimizer : default=torch.optim.Adam.
lr : float, default=0.05. Learning rate used for the training.
batch_size : int, default=256.
max_epochs :int, default=100. Maximal number of training iterations.
iterator_train__shuffle : bool, default=True. Whether to shuffle the training batch iterator during the training
verbose: int, default=1. Set to 0 to disable.
random_state : int. default=42
callbacks : The default is [InputShapeSetterTabSRA(regression=True)] which helps to infer the shape of the input data
Methods
fit(X,y,**fit_params): Fit the module
predict(X): The prediction
predict_inference(X): Alternative inference method to predict. Suitable for very large datasets
get_feature_attribution(X): Local feature attribution
get_attention(X): Produce attention weights
get_weights(): Returns the regression coefficients of the downstream linear part of the model
For remaining list of methods, please refer to https://skorch.readthedocs.io/en/stable/regressor.html