Skip to content

Metrics

This module provides utility functions for evaluating model performance and activation functions. It includes functions to compute the accuracy, top-k accuracy of model predictions, and the sigmoid function.

accuracy(out, yb)

Computes the accuracy of model predictions.

Parameters: out (Tensor): The output tensor from the model, containing predicted class scores. yb (Tensor): The ground truth labels tensor.

Returns: Tensor: The mean accuracy of the predictions, computed as a float tensor.

Source code in temp_dir/pytorch/openml_pytorch/metrics.py 
 9
10
11
12
13
14
15
16
17
18
19
20
21
def accuracy(out, yb):
    """

    Computes the accuracy of model predictions.

    Parameters:
    out (Tensor): The output tensor from the model, containing predicted class scores.
    yb (Tensor): The ground truth labels tensor.

    Returns:
    Tensor: The mean accuracy of the predictions, computed as a float tensor.
    """
    return (torch.argmax(out, dim=1) == yb.long()).float().mean()

accuracy_topk(out, yb, k=5)

Computes the top-k accuracy of the given model outputs.

Args: out (torch.Tensor): The output predictions of the model, of shape (batch_size, num_classes). yb (torch.Tensor): The ground truth labels, of shape (batch_size,). k (int, optional): The number of top predictions to consider. Default is 5.

Returns: float: The top-k accuracy as a float value.

The function calculates how often the true label is among the top-k predicted labels.

Source code in temp_dir/pytorch/openml_pytorch/metrics.py 
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
def accuracy_topk(out, yb, k=5):
    """

    Computes the top-k accuracy of the given model outputs.

    Args:
        out (torch.Tensor): The output predictions of the model, of shape (batch_size, num_classes).
        yb (torch.Tensor): The ground truth labels, of shape (batch_size,).
        k (int, optional): The number of top predictions to consider. Default is 5.

    Returns:
        float: The top-k accuracy as a float value.

    The function calculates how often the true label is among the top-k predicted labels.
    """
    return (torch.topk(out, k, dim=1)[1] == yb.long().unsqueeze(1)).float().mean()

sigmoid(x)

Computes the sigmoid function

The sigmoid function is defined as 1 / (1 + exp(-x)). This function is used to map any real-valued number into the range (0, 1). It is widely used in machine learning, especially in logistic regression and neural networks.

Args: x (numpy.ndarray or float): The input value or array over which the sigmoid function should be applied.

Returns: numpy.ndarray or float: The sigmoid of the input value or array.

Source code in temp_dir/pytorch/openml_pytorch/metrics.py 
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
def sigmoid(x):
    """
    Computes the sigmoid function

    The sigmoid function is defined as 1 / (1 + exp(-x)). This function is used
    to map any real-valued number into the range (0, 1). It is widely used in
    machine learning, especially in logistic regression and neural networks.

    Args:
        x (numpy.ndarray or float): The input value or array over which the
        sigmoid function should be applied.

    Returns:
        numpy.ndarray or float: The sigmoid of the input value or array.
    """
    return 1 / (1 + np.exp(-x))