gfdl 0.1.0__py3-none-any.whl

gfdl/__init__.py

@@ -0,0 +1,36 @@
+import platform
+
+import numpy as np
+import packaging
+import scipy
+import sklearn
+from packaging.version import Version
+
+from .model import GFDL
+
+__all__ = ["GFDL"]
+
+packaging_version = Version(packaging.__version__)
+min_packaging_version = "24.0"
+if packaging_version < Version(min_packaging_version):
+    raise ImportError(f"{packaging_version=}, but {min_packaging_version=}")
+
+python_version = Version(platform.python_version())
+min_python_version = "3.12"
+if python_version < Version(min_python_version):
+    raise ImportError(f"{python_version=}, but {min_python_version=}")
+
+numpy_version = Version(np.__version__)
+min_numpy_version = "2.0.0"
+if numpy_version < Version(min_numpy_version):
+    raise ImportError(f"{numpy_version=}, but {min_numpy_version=}")
+
+sklearn_version = Version(sklearn.__version__)
+min_sklearn_version = "1.5.0"
+if sklearn_version < Version(min_sklearn_version):
+    raise ImportError(f"{sklearn_version=}, but {min_sklearn_version=}")
+
+scipy_version = Version(scipy.__version__)
+min_scipy_version = "1.13.0"
+if scipy_version < Version(min_scipy_version):
+    raise ImportError(f"{scipy_version=}, but {min_scipy_version=}")

gfdl/activations.py

@@ -0,0 +1,194 @@
+"""
+Activation functions for Gradient Free Deep Learning estimators.
+"""
+
+import numpy as np
+import scipy
+
+
+def relu(z):
+    """
+    The rectified linear unit activation function.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The output array with only positive values.
+    """
+    return np.maximum(0, z)
+
+
+def tanh(z):
+    """
+    The hyperbolic tangent activation function.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The output array with hyperbolic tangent values.
+
+    See Also
+    --------
+    numpy.tanh : The hyperbolic tangent function.
+    """
+    return np.tanh(z)
+
+
+def sigmoid(z):
+    """
+    The logistic sigmoid activation function.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The output array with the function values.
+
+    See Also
+    --------
+    scipy.special.expit : The logistic sigmoid function.
+    """
+    return scipy.special.expit(z)
+
+
+def identity(z):
+    """
+    The identity activation function.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The input array is returned unchanged.
+    """
+    return z
+
+
+def softmax(z):
+    """
+    The softmax activation function.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The output array with the function values.
+
+    See Also
+    --------
+    scipy.special.softmax : The softmax function.
+    """
+    return scipy.special.softmax(z, axis=-1)
+
+
+def softmin(z):
+    """
+    The softmin activation function.
+
+    It is the softmax function applied to negative of the input values.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The output array with the function values.
+
+    See Also
+    --------
+    scipy.special.softmax : The softmax function.
+    """
+    return softmax(-z)
+
+
+def log_sigmoid(z):
+    """
+    The logarithm of logistic sigmoid activation function.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The output array with the function values.
+
+    See Also
+    --------
+    scipy.special.log_expit : The logistic sigmoid function.
+    """
+    return scipy.special.log_expit(z)
+
+
+def log_softmax(z):
+    """
+    The log softmax activation function.
+
+    Parameters
+    ----------
+    z : array_like
+        Input array.
+
+    Returns
+    -------
+    numpy.ndarray
+        The output array with the function values.
+
+    See Also
+    --------
+    scipy.special.log_softmax : The logarithm of softmax function.
+    """
+    return scipy.special.log_softmax(z, axis=-1)
+
+
+ACTIVATIONS = {
+    "relu": relu,
+    "tanh": tanh,
+    "sigmoid": sigmoid,
+    "identity": identity,
+    "linear": identity,
+    "softmax": softmax,
+    "softmin": softmin,
+    "log_sigmoid": log_sigmoid,
+    "log_softmax": log_softmax,
+}
+
+
+def resolve_activation(activation):
+    # numpydoc ignore=GL08
+    name = activation.strip().lower()
+    try:
+        fn = ACTIVATIONS[name]
+    except KeyError as e:
+        allowed = sorted(ACTIVATIONS.keys())
+        raise ValueError(
+            f"activation='{activation}' is not supported; choose from {allowed}"
+        ) from e
+    return name, fn