Merge 93bb349 into f97db2a

chainer/__init__.py

@@ -27,6 +27,7 @@
 from chainer.configuration import config  # NOQA
 from chainer.configuration import global_config  # NOQA
 from chainer.configuration import using_config  # NOQA
+from chainer.distribution import Distribution  # NOQA
 from chainer.function import force_backprop_mode  # NOQA
 from chainer.function import Function  # NOQA
 from chainer.function import FunctionAdapter  # NOQA

chainer/distribution.py

@@ -0,0 +1,270 @@
+import copy
+
+
+class Distribution(object):
+
+    """Interface of Distribution
+
+    `Distribution` is a bass class to treat probability distributions.
+    When initialization, it takes parameter as input.
+    """
+
+    def _copy_to(self, target):
+        target.__dict__ = copy.copy(self.__dict__)
+        return target
+
+    @property
+    def batch_shape(self):
+        """Returns the shape of a sample.
+
+        Returns:
+            ~chainer.Variable: Output variable representing the shape of a
+            sample.
+
+        """
+        raise NotImplementedError
+
+    def cdf(self, x):
+        """Returns Cumulative Distribution Function for a input variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing Cumulative
+            Distribution Function.
+
+        """
+        raise NotImplementedError
+
+    @property
+    def covariance(self):
+        """Returns covariance.
+
+        Returns:
+            ~chainer.Variable: Output variable representing covariance.
+        """
+        raise NotImplementedError
+
+    @property
+    def entropy(self):
+        """Returns entropy.
+
+        Returns:
+            ~chainer.Variable: Output variable representing entropy.
+
+        """
+        raise NotImplementedError
+
+    @property
+    def enumerate_support(self):
+        """Returns support values of discrete distribution.
+
+        Returns:
+            ~chainer.Variable: Output variable containing candidates.
+
+        """
+        raise NotImplementedError
+
+    @property
+    def event_shape(self):
+        """Returns the shape of an event.
+
+        Returns:
+            ~chainer.Variable: Output variable representing the shape of an
+            event.
+
+        """
+        raise NotImplementedError
+
+    def icdf(self, x):
+        """Returns Inverse Cumulative Distribution Function for a input Variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing Inverse Cumulative
+            Distribution Function.
+
+        """
+        raise NotImplementedError
+
+    def log_cdf(self, x):
+        """Returns logarithm of Cumulative Distribution Function for a input Variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing logarithm of
+            Cumulative Distribution Function.
+
+        """
+        raise NotImplementedError
+
+    def log_prob(self, x):
+        """Returns logarithm of probability for a input variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing logarithm of
+            probability.
+
+        """
+        raise NotImplementedError
+
+    def log_survival_function(self, x):
+        """Returns logarithm of survival function for a input Variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing logarithm of
+            survival function for a input variable.
+
+        """
+        raise NotImplementedError
+
+    @property
+    def mean(self):
+        """Returns mean value.
+
+        Returns:
+            ~chainer.Variable: Output variable representing mean value.
+
+        """
+        raise NotImplementedError
+
+    @property
+    def mode(self):
+        """Returns mode.
+
+        Returns:
+            ~chainer.Variable: Output variable representing mode.
+
+        """
+        raise NotImplementedError
+
+    def perplexity(self, x):
+        """Returns perplexity function for a input variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing perplexity function
+            for a input variable.
+
+        """
+        raise NotImplementedError
+
+    def prob(self, x):
+        """Returns probability for a input variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing probability.
+
+        """
+        raise NotImplementedError
+
+    def sample(self, shape=()):
+        """Samples from this distribution.
+
+        Args:
+            shape(:class:`tuple` of :class:`int`): Sampling shape.
+
+        Returns:
+            ~chainer.Variable: Output variable representing sampled random
+            variable.
+        """
+        final_shape = self.batch_shape + self.event_shape
+        if shape == ():
+            n = 1
+        elif isinstance(shape, int):
+            n = shape
+            final_shape = (n,) + final_shape
+        else:
+            n = 1
+            for shape_ in shape:
+                n *= shape_
+            final_shape = shape + final_shape
+        samples = self._sample_n(n)
+        return samples.reshape(final_shape)
+
+    def _sample_n(self, n):
+        """Samples from this distribution.
+
+        Args:
+            n(`int`): Sampling size.
+
+        Returns:
+            ~chainer.Variable: Output variable representing sampled random
+            variable.
+        """
+        raise NotImplementedError
+
+    @property
+    def stddev(self):
+        """Returns standard deviation.
+
+        Returns:
+            ~chainer.Variable: Output variable representing standard deviation.
+
+        """
+        raise NotImplementedError
+
+    @property
+    def support(self):
+        """Returns support.
+
+        Returns:
+            string: Output string that means support of this distribution.
+
+        """
+        raise NotImplementedError
+
+    def survival_function(self, x):
+        """Returns survival function for a input variable.
+
+        Args:
+            x(:class:`~chainer.Variable` or :class:`numpy.ndarray` or \
+            :class:`cupy.ndarray`): Input variable representing a random
+            variable.
+
+        Returns:
+            ~chainer.Variable: Output variable representing survival function
+            for a input variable.
+
+        """
+        raise NotImplementedError
+
+    @property
+    def variance(self):
+        """Returns variance.
+
+        Returns:
+            ~chainer.Variable: Output variable representing variance.
+
+        """
+        raise NotImplementedError