Distances

Functions:

calc_ecdf_dist(r, f[, rw, fw])	Calculate distance metrics between two samples over multiple dimensions using ecdf_distance and convert to a numpy array.
calc_hist_dist(r, f, bins[, rw, fw])	Calculate distance metrics between two samples over multiple dimensions using hist_distance and convert to a numpy array.
calc_sw1_dist(r, f[, rw, fw])	Calculate scaled Wasserstein-1 distance metrics between two samples over multiple dimensions.
ecdf(arr[, weights])	Calculate an approximation to the cumulative distribution function (CDF) using empirical distribution.
ecdf_distance(r, f[, rw, fw])	Calculate a distance metric between two 1D samples using the ecdf function.
energy_distance(r, f[, rw, fw])	Calculate the Energy Distance between two 1D samples.
hist_distance(r, f, bins[, rw, fw])	Calculate a distance metric between two 1D samples using the histcdf function.
histcdf(arr, bins[, weight])	Calculate an approximation to the cumulative distribution function (CDF) using histograms.
scaled_w1_distance(r, f[, rw, fw])	Calculate the scaled Wasserstein-1 Distance between two 1D samples.
wasserstein_distance(r, f[, rw, fw])	Calculate the Wasserstein Distance (Earth Mover's Distance) between two 1D samples.

caloutils.distances.calc_ecdf_dist(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.ndarray' id='140552768342912'>

Calculate distance metrics between two samples over multiple dimensions using ecdf_distance and convert to a numpy array.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

np.ndarray: Array of distance metrics calculated for each dimension.

caloutils.distances.calc_hist_dist(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, bins: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.ndarray' id='140552768342912'>

Calculate distance metrics between two samples over multiple dimensions using hist_distance and convert to a numpy array.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. bins (torch.Tensor): Bins for histogram calculation. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

np.ndarray: Array of distance metrics calculated for each dimension.

caloutils.distances.calc_sw1_dist(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.ndarray' id='140552768342912'>

Calculate scaled Wasserstein-1 distance metrics between two samples over multiple dimensions.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

np.ndarray: Array of scaled Wasserstein-1 distance metrics calculated for each dimension.

caloutils.distances.ecdf(arr: <Mock name='mock.Tensor' id='140552768372112'>, weights: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → tuple[<Mock name='mock.Tensor' id='140552768372112'>, typing.Optional[<Mock name='mock.Tensor' id='140552768372112'>]]

Calculate an approximation to the cumulative distribution function (CDF) using empirical distribution.

Args:

arr (torch.Tensor): Input tensor containing data points. weights (Optional[torch.Tensor]): Weights for each data point (optional).

Returns:

tuple[torch.Tensor, Optional[torch.Tensor]]: Tuple containing the calculated CDF values and weights (if provided).

caloutils.distances.ecdf_distance(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.Tensor' id='140552768372112'>

Calculate a distance metric between two 1D samples using the ecdf function.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

torch.Tensor: Calculated distance metric between the two samples.

caloutils.distances.energy_distance(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.Tensor' id='140552768372112'>

Calculate the Energy Distance between two 1D samples.

The Energy Distance is a statistical measure that quantifies the difference between the empirical distributions of two sets of data points.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

torch.Tensor: Calculated Energy Distance between the two samples.

caloutils.distances.hist_distance(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, bins: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.Tensor' id='140552768372112'>

Calculate a distance metric between two 1D samples using the histcdf function.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. bins (torch.Tensor): Bins for histogram calculation. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

torch.Tensor: Calculated distance metric between the two samples.

caloutils.distances.histcdf(arr: <Mock name='mock.Tensor' id='140552768372112'>, bins: <Mock name='mock.Tensor' id='140552768372112'>, weight: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.Tensor' id='140552768372112'>

Calculate an approximation to the cumulative distribution function (CDF) using histograms.

Args:

arr (torch.Tensor): Input tensor containing data points. bins (torch.Tensor): Bins for histogram calculation. weight (Optional[torch.Tensor]): Weights for each data point (optional).

Returns:

torch.Tensor: Approximated cumulative distribution function values.

caloutils.distances.scaled_w1_distance(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.Tensor' id='140552768372112'>

Calculate the scaled Wasserstein-1 Distance between two 1D samples.

The scaled Wasserstein-1 Distance measures the difference between the distributions of two samples while accounting for differences in scale and location.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

torch.Tensor: Calculated scaled Wasserstein-1 Distance between the two samples.

caloutils.distances.wasserstein_distance(r: <Mock name='mock.Tensor' id='140552768372112'>, f: <Mock name='mock.Tensor' id='140552768372112'>, rw: <Mock name='mock.Tensor' id='140552768372112'> | None = None, fw: <Mock name='mock.Tensor' id='140552768372112'> | None = None) → <Mock name='mock.Tensor' id='140552768372112'>

Calculate the Wasserstein Distance (Earth Mover’s Distance) between two 1D samples.

The Wasserstein Distance measures the minimum “cost” of transforming one distribution into another.

Args:

r (torch.Tensor): Reference sample. f (torch.Tensor): Comparison sample. rw (Optional[torch.Tensor]): Weights for the reference sample (optional). fw (Optional[torch.Tensor]): Weights for the comparison sample (optional).

Returns:

torch.Tensor: Calculated Wasserstein Distance between the two samples.