confusion_matrix

Name	Description
`ConfusionMatrix`	Accumulates a confusion matrix and returns row-normalized values.

Name

Description

ConfusionMatrix

Accumulates a confusion matrix and returns row-normalized values.

ConfusionMatrix(num_classes: int, name='confusion_matrix', **kwargs={})

Accumulate class counts and return a row-normalized confusion matrix.

Rows represent true classes and columns represent predicted classes. Rows with no observations return zeros. State accumulates across batches until reset_state() is called.

Parameters

Parameters of ConfusionMatrix
Name	Type	Default	Description
`num_classes`	`int`	Required	Number of classes, defining both matrix dimensions.
`name`	`str`	'confusion_matrix'	Keras metric name.
`**kwargs`	`Any`	{}	Additional Keras Metric arguments, including state dtype.

num_classes = int(num_classes)

conf_matrix = self.add_variable(name='conf_matrix', shape=(self.num_classes, self.num_classes), initializer='zeros', dtype=self._state_dtype)

update_state(y_true, y_pred, sample_weight=None)

Add a batch of labels and predictions to the accumulated counts.

Parameters

Parameters of update_state
Name	Type	Default	Description
`y_true`	`Tensor`	Required	Integer class labels, flattened before accumulation.
`y_pred`	`Tensor`	Required	Class scores with shape (batch, ..., num_classes), reduced with argmax over the last axis, or matching integer labels.
`sample_weight`	`Tensor \| None`	None	Optional weights, flattened to one value per label.

Raises

Errors raised by update_state
Type	Description
`ValueError`	A true or predicted label is outside the class range in eager execution. TensorFlow graph execution uses an assertion.

result()

Returns the row-normalized confusion matrix as float64.

reset_state()

get_config()

confusion_matrix

Confusion Matrix Metric API