Python tensorflow.python.ops.metrics_impl._streaming_confusion_matrix() Examples

The following are 30 code examples of tensorflow.python.ops.metrics_impl._streaming_confusion_matrix(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module tensorflow.python.ops.metrics_impl , or try the search function .
Example #1
Source File: tf_metrics.py    From FoolNLTK with Apache License 2.0 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #2
Source File: tf_metrics.py    From pynlp with MIT License 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #3
Source File: tf_metrics.py    From pynlp with MIT License 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #4
Source File: tf_metrics.py    From pynlp with MIT License 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #5
Source File: calculate_model_score.py    From pynlp with MIT License 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #6
Source File: calculate_model_score.py    From pynlp with MIT License 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #7
Source File: tf_metrics.py    From pynlp with MIT License 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #8
Source File: tf_metrics.py    From pynlp with MIT License 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #9
Source File: tf_metrics.py    From pynlp with MIT License 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #10
Source File: tf_metrics.py    From FoolNLTK with Apache License 2.0 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #11
Source File: tf_metrics.py    From bert-chinese-ner with MIT License 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #12
Source File: tf_metrics.py    From FoolNLTK with Apache License 2.0 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #13
Source File: tf_metrics.py    From albert-chinese-ner with MIT License 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #14
Source File: tf_metrics.py    From albert-chinese-ner with MIT License 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #15
Source File: tf_metrics.py    From albert-chinese-ner with MIT License 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #16
Source File: tf_metrics.py    From BERT-for-Sequence-Labeling-and-Text-Classification with Apache License 2.0 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #17
Source File: tf_metrics.py    From BERT-for-Sequence-Labeling-and-Text-Classification with Apache License 2.0 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #18
Source File: tf_metrics.py    From BERT-for-Sequence-Labeling-and-Text-Classification with Apache License 2.0 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #19
Source File: calculate_model_score.py    From BERT-for-Sequence-Labeling-and-Text-Classification with Apache License 2.0 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #20
Source File: calculate_model_score.py    From BERT-for-Sequence-Labeling-and-Text-Classification with Apache License 2.0 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #21
Source File: metrics.py    From curriculum with GNU General Public License v3.0 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #22
Source File: tf_metrics.py    From tudouNLP with MIT License 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #23
Source File: tf_metrics.py    From tudouNLP with MIT License 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #24
Source File: __init__.py    From tf_metrics with Apache License 2.0 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow

    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.

    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #25
Source File: __init__.py    From tf_metrics with Apache License 2.0 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow

    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.

    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #26
Source File: __init__.py    From tf_metrics with Apache License 2.0 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow

    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean

    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #27
Source File: tf_metrics.py    From curriculum with GNU General Public License v3.0 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    """Multi-class precision metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)
Example #28
Source File: tf_metrics.py    From curriculum with GNU General Public License v3.0 5 votes vote down vote up 
def recall(labels, predictions, num_classes, pos_indices=None, weights=None,
           average='micro'):
    """Multi-class recall metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, re, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    _, op, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (re, op)
Example #29
Source File: tf_metrics.py    From curriculum with GNU General Public License v3.0 5 votes vote down vote up 
def fbeta(labels, predictions, num_classes, pos_indices=None, weights=None,
          average='micro', beta=1):
    """Multi-class fbeta metric for Tensorflow
    Parameters
    ----------
    labels : Tensor of tf.int32 or tf.int64
        The true labels
    predictions : Tensor of tf.int32 or tf.int64
        The predictions, same shape as labels
    num_classes : int
        The number of classes
    pos_indices : list of int, optional
        The indices of the positive classes, default is all
    weights : Tensor of tf.int32, optional
        Mask, must be of compatible shape with labels
    average : str, optional
        'micro': counts the total number of true positives, false
            positives, and false negatives for the classes in
            `pos_indices` and infer the metric from it.
        'macro': will compute the metric separately for each class in
            `pos_indices` and average. Will not account for class
            imbalance.
        'weighted': will compute the metric separately for each class in
            `pos_indices` and perform a weighted average by the total
            number of true labels for each class.
    beta : int, optional
        Weight of precision in harmonic mean
    Returns
    -------
    tuple of (scalar float Tensor, update_op)
    """
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    _, _, fbeta = metrics_from_confusion_matrix(
        cm, pos_indices, average=average, beta=beta)
    _, _, op = metrics_from_confusion_matrix(
        op, pos_indices, average=average, beta=beta)
    return (fbeta, op)
Example #30
Source File: metrics.py    From curriculum with GNU General Public License v3.0 5 votes vote down vote up 
def precision(labels, predictions, num_classes, pos_indices=None,
              weights=None, average='micro'):
    cm, op = _streaming_confusion_matrix(
        labels, predictions, num_classes, weights)
    pr, _, _ = metrics_from_confusion_matrix(
        cm, pos_indices, average=average)
    op, _, _ = metrics_from_confusion_matrix(
        op, pos_indices, average=average)
    return (pr, op)