initial commit for packaging

Author: ctgk

ch01_Introduction.ipynb

@@ -0,0 +1,10 @@
+from . import distributions
+from . import features
+from . import linear_models
+
+
+__all__ = [
+    "distributions",
+    "features",
+    "linear_models"
+]

ch02_Probability_Distributions.ipynb

@@ -0,0 +1,12 @@
+from .beta import BetaDistribution
+from .gaussian import GaussianDistribution
+from .gaussian_mixture import GaussianMixtureDistribution
+from .students_t import StudentsTDistribution
+
+
+__all__ = [
+    "BetaDistribution",
+    "GaussianDistribution",
+    "GaussianMixtureDistribution",
+    "StudentsTDistribution"
+]

ch03_Linear_Models_for_Regression.ipynb

@@ -0,0 +1,89 @@
+import numpy as np
+from scipy.special import gamma
+
+
+np.seterr(all="ignore")
+
+
+class BetaDistribution(object):
+
+    def __init__(self, pseudo_ones=1, pseudo_zeros=1):
+        """
+        construct beta distribution
+
+        Parameters
+        ----------
+        pseudo_ones : float
+            pseudo count of one
+        pseudo_zeros : float
+            pseudo count of zero
+        """
+        self.pseudo_ones = pseudo_ones
+        self.pseudo_zeros = pseudo_zeros
+        self.n_ones = pseudo_ones
+        self.n_zeros = pseudo_zeros
+
+    def fit(self, n_ones=None, n_zeros=None):
+        """
+        estimating parameter of posterior distribution of Beta(mu|a,b)
+
+        Parameters
+        ----------
+        n_ones : float
+            number of observed one
+        n_zeros : float
+            number of observed zero
+        """
+        self.n_ones = 0 if n_ones is None else n_ones
+        self.n_zeros = 0 if n_zeros is None else n_zeros
+        self.n_ones += self.pseudo_ones
+        self.n_zeros += self.pseudo_zeros
+
+    def fit_online(self, n_ones=None, n_zeros=None):
+        """
+        online estimation of posterior distribution Beta(mu|a,b)
+
+        Parameters
+        ----------
+        n_ones : float
+            number of observed one
+        n_zeros : float
+            number of observed zero
+        """
+        self.n_ones += 0 if n_ones is None else n_ones
+        self.n_zeros += 0 if n_zeros is None else n_zeros
+
+    def predict(self):
+        """
+        returns one or zero according to the posterior distribution
+
+        Returns
+        -------
+        output : int
+            prediction
+        """
+        return int(
+            self.n_ones / (self.n_ones + self.n_zeros)
+            > np.random.uniform())
+
+    def probability(self, x):
+        """
+        probability denstiy function
+        calculate posterior distribution beta(x|D)
+
+        Parameters
+        ----------
+        x : ndarray
+            input
+
+        Returns
+        -------
+        output : float
+            value of posterior distribution
+        """
+        return (
+            gamma(self.n_ones + self.n_zeros)
+            * np.power(x, self.n_ones - 1)
+            * np.power(1 - x, self.n_zeros - 1)
+            / gamma(self.n_ones)
+            / gamma(self.n_zeros))