tfm.uplift.metrics.sliced_metric.SlicedMetric

A metric sliced by integer, boolean, or string features.

tfm.uplift.metrics.sliced_metric.SlicedMetric(
    metric: tf_keras.metrics.Metric,
    slicing_spec: (dict[str, str] | dict[str, int]),
    slicing_feature_dtype: (tf.DType | None) = None,
    name: (str | None) = None
)

A metric wrapper that computes a metric for different slices of an arbitrary feature. The slicing is specified via a slicing spec, which is a dictionary from a slice name to the unique value to be sliced on. For each pair of slice_name, slicing_value passed, the suffix /slice_name will be appended to the name of the result of the corresponding slice. An overall result is also computed without any slicing applied.

In order for this to work correctly, the given metric must support passing sample_weight to its update_state() method. Additionally, the slicing feature must also be passed to update_state() method of this class and must be of a broadcastable shape to the metric inputs. This wrapper creates a deep copy of the metric passed to it for each slice. At every call to update_state(), the wrapper will call the update_state() method of the metric for each slice with the sample_weights set to zero where the slicing feature is not equal to the corresponding slicing value.

If the given metric returns a tensor, the result of this metric will be a dictionary mapping from the sliced metric's name to the result for that slice. If the given metric returns a dictionary of tensors, the result of this metric will be a flattened dictionary consisting of each of the sliced metrics' results for every slice.

Example usage:

sliced_metric = SlicedMetric(
    tf_keras.metrics.Accuracy('accuracy'),
    slicing_spec={"control": False, "treatment": True},
)
sliced_metric.update_state(
    y_true=tf.constant([[0], [1], [0], [1]]),
    y_pred=tf.constant([[1], [0], [1], [1]]),
    slicing_feature=tf.constant([[True], [False], [True], [False]]),
)
sliced_metric.result()
{
    "accuracy": 0.25,
    "accuracy/control": 0.5,
    "accuracy/treatment": 0
}

metric A tf_keras.metrics.Metric instance.
slicing_spec A dictionary that maps from string slice names, to one of integer, boolean, or string slicing values.
slicing_feature_dtype The expected dtype of the slicing feature. The values in the slicing spec are casted to this type if passed. If None, the dtype of the slicing feature is inferred based on the values in the slicing spec.
name The name of the wrapper metric. Defaults to sliced_{metric.name}.

A ValueError if slicing_spec is empty, contains duplicate slicing values, or has slicing values of different types.

Methods

call

call(
    inputs, *args, **kwargs
)

This is where the layer's logic lives.

The call() method may not create state (except in its first invocation, wrapping the creation of variables or other resources in tf.init_scope()). It is recommended to create state, including tf.Variable instances and nested Layer instances, in __init__(), or in the build() method that is called automatically before call() executes for the first time.

Args
inputs Input tensor, or dict/list/tuple of input tensors. The first positional inputs argument is subject to special rules:

  • inputs must be explicitly passed. A layer cannot have zero arguments, and inputs cannot be provided via the default value of a keyword argument.
  • NumPy array or Python scalar values in inputs get cast as tensors.
  • TF-Keras mask metadata is only collected from inputs.
  • Layers are built (build(input_shape) method) using shape info from inputs only.
  • input_spec compatibility is only checked against inputs.
  • Mixed precision input casting is only applied to inputs. If a layer has tensor arguments in *args or **kwargs, their casting behavior in mixed precision should be handled manually.
  • The SavedModel input specification is generated using inputs only.
  • Integration with various ecosystem packages like TFMOT, TFLite, TF.js, etc is only supported for inputs and not for tensors in positional and keyword arguments.
*args Additional positional arguments. May contain tensors, although this is not recommended, for the reasons above.
**kwargs Additional keyword arguments. May contain tensors, although this is not recommended, for the reasons above. The following optional keyword arguments are reserved:
  • training: Boolean scalar tensor of Python boolean indicating whether the call is meant for training or inference.
  • mask: Boolean input mask. If the layer's call() method takes a mask argument, its default value will be set to the mask generated for inputs by the previous layer (if input did come from a layer that generated a corresponding mask, i.e. if it came from a TF-Keras layer with masking support).

    • Returns
    A tensor or list/tuple of tensors.

    merge_state

    merge_state(
    metrics
)

    Merges the state from one or more metrics.

    This method can be used by distributed systems to merge the state computed by different metric instances. Typically the state will be stored in the form of the metric's weights. For example, a tf.keras.metrics.Mean metric contains a list of two weight values: a total and a count. If there were two instances of a tf.keras.metrics.Accuracy that each independently aggregated partial state for an overall accuracy calculation, these two metric's states could be combined as follows:

    m1 = tf.keras.metrics.Accuracy()
_ = m1.update_state([[1], [2]], [[0], [2]])
     
    m2 = tf.keras.metrics.Accuracy()
_ = m2.update_state([[3], [4]], [[3], [4]])
     
    m2.merge_state([m1])
m2.result().numpy()
0.75

    Args
    metrics an iterable of metrics. The metrics must have compatible state.

    Raises
    ValueError If the provided iterable does not contain metrics matching the metric's required specifications.

    reset_state

    View source

    reset_state()

    Resets all of the metric state variables.

    This function is called between epochs/steps, when a metric is evaluated during training.

    result

    View source

    result() -> dict[str, tf.Tensor]

    Aggregates all the metrics' results into a flattened dictionary.

    update_state

    View source

    update_state(
    *args,
    sample_weight: (tf.Tensor | None) = None,
    slicing_feature: tf.Tensor,
    **kwargs
)

    Updates the state of the metrics for each slice.

    Args
    *args A variable amount of tf.Tensor instances that will be passed to the update_state method of each metric.
    sample_weight An optional tf.Tensor used to weight the sample. Its dimensions must be broadcastable to the shape(s) of *args.
    slicing_feature A tf.Tensor consisting of the feature to be sliced on. Its dimensions must be broadcastable to the shape(s) of *args.
    **kwargs Keyword arguments that will be passed to the update_state method of each metric.