linear_softmax¶
- mlpractice.linear_classifier.linear_softmax(X, W, target_index)¶
Performs linear classification and returns loss and gradient with respect to W.
- Parameters
- Xnp.ndarray, shape(batch_size, n_features)
Batch of images.
- Wnp.ndarray, shape(n_features, n_classes)
Weights.
- target_indexnp.ndarray, shape(batch_size)
Indices of the true classes for given samples.
- Returns
- lossfloat
Computed cross-entropy loss value.
- gradientnp.ndarray, shape(n_features, n_classes)
Gradient of loss with respect to weights.