LinearSoftmaxClassifier¶
- class mlpractice.linear_classifier.LinearSoftmaxClassifier¶
Linear softmax classifier class.
- Attributes
- Wnp.ndarray
Weights.
Methods
fit(X, y[, batch_size, learning_rate, ...])Trains linear classifier.
predict(X)Predicts classes for X.
- fit(X, y, batch_size=100, learning_rate=1e-07, reg_strength=1e-05, epochs=1)¶
Trains linear classifier.
- Parameters
- Xnp.ndarray, shape(n_samples, n_features)
Training data.
- ynp.ndarray, shape(n_samples)
Training data class labels.
- batch_sizeint, optional
The number of samples to use for each batch.
- learning_ratefloat, optional
Learning rate for gradient descent.
- reg_strengthfloat, optional
L2 regularization strength.
- epochsint, optional
The number of passes over the training data.
- Returns
- loss_historyarray_like
Holds a record of the loss values during training.
- predict(X)¶
Predicts classes for X.
- Parameters
- Xnp.ndarray, shape(n_samples, n_features)
Input samples.
- Returns
- y_prednp.ndarray, shape(n_samples)
Predicted classes.