Modification for web pages

Author: yuta-nakahara

README.md

@@ -23,7 +23,7 @@ For more details, see our [website](https://yuta-nakahara.github.io/BayesML/ "Ba
 
 Please use the following commands to install BayesML.
 
-``` shell
+``` bash
 pip install bayesml
 ```
 

README_jp.md

@@ -22,7 +22,7 @@ BayesMLは以下の特徴を持っています．
 
 以下のコマンドによりインストール可能です．
 
-``` shell
+``` bash
 pip install bayesml
 ```
 

bayesml/_test.py

@@ -1,18 +1,8 @@
-from bayesml import bernoulli
-
-gen_model = bernoulli.GenModel(theta=0.7)
-
-gen_model.visualize_model()
-
-x = gen_model.gen_sample(sample_size=20)
-
-learn_model = bernoulli.LearnModel()
-
-learn_model.visualize_posterior()
-
-learn_model.update_posterior(x)
-learn_model.visualize_posterior()
-
-print(learn_model.estimate_params(loss='squared'))
-print(learn_model.estimate_params(loss='abs'))
-print(learn_model.estimate_params(loss='0-1'))
+from bayesml import hiddenmarkovnormal
+import numpy as np
+model = hiddenmarkovnormal.GenModel(
+        c_num_classes=2,
+        c_degree=1,
+        mu_vecs=np.array([[5],[-5]]),
+        a_mat=np.array([[0.95,0.05],[0.1,0.9]]))
+model.visualize_model()

bayesml/contexttree/__init__.py

@@ -54,7 +54,7 @@
 where the updating rule of the hyperparameter is as follows:
 
 .. math::
-    \beta_n(k|s) = \beta_0(k|s) + \sum_{i=1}^n I \left\{ \text{:math:`s` is the ancestor of :math:`s_{T_\mathrm{max}}(x^{i-1})` and :math:`x_i=k` } \right\}.
+    \beta_n(k|s) = \beta_0(k|s) + \sum_{i=1}^n I \left\{ s \ \mathrm{is \ the \ ancestor \ of} \ s_{T_\mathrm{max}}(x^{i-1}) \ \mathrm{and} \ x_i=k \right\}.
 
 For :math:`T \in \mathcal{T}`,
 
@@ -66,18 +66,18 @@
 .. math::
     g_{n,s} =
     \begin{cases}
-        g_{0,s} & \text{if :math:`n=0`}, \\
+        g_{0,s}, & n=0, \\
         \frac{ g_{n-1,s} \tilde{q}_{s_{\mathrm{child}}} (x_n|x^{n-1}) }
-        { \tilde{q}_s(x_n|x^{n-1}) } & \text{otherwise},
+        { \tilde{q}_s(x_n|x^{n-1}) } & \mathrm{otherwise},
     \end{cases}
 
 where :math:`s_{\mathrm{child}}` is the child node of :math:`s` on the path from :math:`s_\lambda` to :math:`s_{T_\mathrm{max}}(x^n)` and
 
 .. math::
     \tilde{q}_s(x_n|x^{n-1}) =
     \begin{cases}
-        q_s(x_n|x^{n-1}) & \text{if :math:`s\in\mathcal{L}(T_\mathrm{max})`}, \\
-        (1-g_{n-1,s}) q_s(x_n|x^{n-1}) + g_{n-1,s} \tilde{q}_{s_{\mathrm{child}}}(x_n|x^{n-1}) & \text{otherwise}.
+        q_s(x_n|x^{n-1}) & s\in\mathcal{L}(T_\mathrm{max}), \\
+        (1-g_{n-1,s}) q_s(x_n|x^{n-1}) + g_{n-1,s} \tilde{q}_{s_{\mathrm{child}}}(x_n|x^{n-1}) & \mathrm{otherwise}.
     \end{cases}
 
 Here,

bayesml/contexttree/_contexttree.py

@@ -191,7 +191,7 @@ def _visualize_model_recursion(self,tree_graph,node,node_id,parent_id,sibling_nu
         tmp_p_v = p_v
 
         # add node information
-        label_string = f'h_g={node.h_g:.2f}\\lp_v={tmp_p_v:.2f}\\ltheta_vec='
+        label_string = f'h_g={node.h_g:.2f}\\lp_v={tmp_p_v:.2f}\\ltheta_vec\\l='
         if node.leaf:
             label_string += '['
             for i in range(self.c_k):
@@ -388,16 +388,16 @@ def visualize_model(self,filename=None,format=None,sample_length=10):
         Examples
         --------
         >>> from bayesml import contexttree
-        >>> model = contexttree.GenModel(c_k=2,c_d_max=3,h_g=0.75)
+        >>> gen_model = contexttree.GenModel(c_k=2,c_d_max=3,h_g=0.75)
         >>> gen_model.gen_params()
-        >>> model.visualize_model()
-        [1 1 1 1 1 1 0 0 0 1]
+        >>> gen_model.visualize_model()
+        [1 0 1 0 0 0 1 0 0 0]
 
         .. image:: ./images/contexttree_example.png
 
         See Also
         --------
-        graphbiz.Digraph
+        graphviz.Digraph
         """
         #例外処理
         _check.pos_int(sample_length,'sample_length',DataFormatError)
@@ -781,7 +781,7 @@ def estimate_params(self,loss="0-1",visualize=True,filename=None,format=None):
 
         See Also
         --------
-        graphbiz.Digraph
+        graphviz.Digraph
         """
 
         if loss == "0-1":
@@ -820,7 +820,7 @@ def _visualize_model_recursion(self,tree_graph,node:_LearnNode,node_id,parent_id
         tmp_p_v = p_v
 
         # add node information
-        label_string = f'hn_g={node.hn_g:.2f}\\lp_v={tmp_p_v:.2f}\\ltheta_vec='
+        label_string = f'hn_g={node.hn_g:.2f}\\lp_v={tmp_p_v:.2f}\\ltheta_vec\\l='
         label_string += '['
         for i in range(self.c_k):
             theta_vec_hat = node.hn_beta_vec / node.hn_beta_vec.sum()
@@ -869,7 +869,7 @@ def visualize_posterior(self,filename=None,format=None):
 
         See Also
         --------
-        graphbiz.Digraph
+        graphviz.Digraph
         """
         try:
             import graphviz

bayesml/gaussianmixture/_gaussianmixture.py

@@ -298,9 +298,7 @@ def visualize_model(self,sample_size=100):
          [ 2. ]]
         lambda_mats:
          [[[  6.25]]
-
          [[  6.25]]
-
          [[100.  ]]]
         
         .. image:: ./images/gaussianmixture_example.png
@@ -535,7 +533,7 @@ def set_h0_params(
             self.h0_w_mats[:] = h0_w_mats
             self.h0_w_mats_inv[:] = np.linalg.inv(self.h0_w_mats)
 
-        self.calc_prior_char()
+        self._calc_prior_char()
         self.reset_hn_params()
 
     def get_h0_params(self):
@@ -635,7 +633,7 @@ def get_hn_params(self):
                 "hn_nus":self.hn_nus,
                 "hn_w_mats":self.hn_w_mats}
 
-    def calc_prior_char(self):
+    def _calc_prior_char(self):
         self._ln_c_h0_alpha = gammaln(self.h0_alpha_vec.sum()) - gammaln(self.h0_alpha_vec).sum()
         self._ln_b_h0_w_nus = (
             - self.h0_nus*np.linalg.slogdet(self.h0_w_mats)[1]
@@ -645,7 +643,7 @@ def calc_prior_char(self):
                      axis=1) * 2.0
             ) / 2.0
 
-    def calc_vl(self):
+    def _calc_vl(self):
         # E[ln p(X|Z,mu,Lambda)]
         self._vl_p_x = np.sum(
             self.ns
@@ -797,10 +795,10 @@ def update_posterior(
         tolerance : float, optional
             convergence croterion of variational lower bound, by default 1.0E-8
         init_type : str, optional
-            type of initialization, by default 'subsampling'
-            * 'subsampling': for each latent class, extract a subsample whose size is int(np.sqrt(x.shape[0])).
-              and use its mean and covariance matrix as an initial values of hn_m_vecs and hn_lambda_mats.
-            * 'random_responsibility': randomly assign responsibility to r_vecs
+            * ``'subsampling'``: for each latent class, extract a subsample whose size is ``int(np.sqrt(x.shape[0]))``.
+              and use its mean and covariance matrix as an initial values of ``hn_m_vecs`` and ``hn_lambda_mats``.
+            * ``'random_responsibility'``: randomly assign responsibility to ``r_vecs``
+            Type of initialization, by default ``'subsampling'``
         """
         _check.float_vecs(x,'x',DataFormatError)
         if x.shape[-1] != self.c_degree:
@@ -833,14 +831,14 @@ def update_posterior(
                     f'init_type={init_type} is unsupported. '
                     + 'This function supports only '
                     + '"subsampling" and "random_responsibility"'))
-            self.calc_vl()
+            self._calc_vl()
             print(f'\r{i}. VL: {self.vl}',end='')
             for t in range(max_itr):
                 vl_before = self.vl
                 self._update_q_mu_lambda()
                 self._update_q_pi()
                 self._update_q_z(x)
-                self.calc_vl()
+                self._calc_vl()
                 print(f'\r{i}. VL: {self.vl} t={t} ',end='')
                 if np.abs((self.vl-vl_before)/vl_before) < tolerance:
                     convergence_flag = False
@@ -959,11 +957,9 @@ def visualize_posterior(self):
         [47.68878373 54.31121627]
         hn_w_mats:
         [[[0.02226992]]
-
         [[0.01575793]]]
         E[lambda_mats]=
         [[[1.06202546]]
-
         [[0.85583258]]]
 
         .. image:: ./images/gaussianmixture_posterior.png
@@ -1056,7 +1052,7 @@ def make_prediction(self,loss="squared"):
 
         Returns
         -------
-        Predicted_value : {float, numpy.ndarray}
+        predicted_value : {float, numpy.ndarray}
             The predicted value under the given loss function. 
         """
         if loss == "squared":
@@ -1105,14 +1101,14 @@ def pred_and_update(
         tolerance : float, optional
             convergence croterion of variational lower bound, by default 1.0E-8
         init_type : str, optional
-            type of initialization, by default 'random_responsibility'
-            * 'random_responsibility': randomly assign responsibility to r_vecs
-            * 'subsampling': for each latent class, extract a subsample whose size is int(np.sqrt(x.shape[0])).
-              and use its mean and covariance matrix as an initial values of hn_m_vecs and hn_lambda_mats.
+            * ``'random_responsibility'``: randomly assign responsibility to ``r_vecs``
+            * ``'subsampling'``: for each latent class, extract a subsample whose size is ``int(np.sqrt(x.shape[0]))``.
+              and use its mean and covariance matrix as an initial values of ``hn_m_vecs`` and ``hn_lambda_mats``.
+            Type of initialization, by default ``'random_responsibility'``
 
         Returns
         -------
-        Predicted_value : {float, numpy.ndarray}
+        predicted_value : {float, numpy.ndarray}
             The predicted value under the given loss function. 
         """
         _check.float_vec(x,'x',DataFormatError)
@@ -1143,7 +1139,7 @@ def estimate_latent_vars(self,x,loss="0-1"):
 
         Returns
         -------
-        Estimates : numpy.ndarray
+        estimates : numpy.ndarray
             The estimated values under the given loss function. 
             If the loss function is \"KL\", the posterior distribution will be returned 
             as a numpy.ndarray whose elements consist of occurence probabilities.
@@ -1196,14 +1192,14 @@ def estimate_latent_vars_and_update(
         tolerance : float, optional
             convergence croterion of variational lower bound, by default 1.0E-8
         init_type : str, optional
-            type of initialization, by default 'subsampling'
-            * 'subsampling': for each latent class, extract a subsample whose size is int(np.sqrt(x.shape[0])).
-              and use its mean and covariance matrix as an initial values of hn_m_vecs and hn_lambda_mats.
-            * 'random_responsibility': randomly assign responsibility to r_vecs
+            * ``'subsampling'``: for each latent class, extract a subsample whose size is ``int(np.sqrt(x.shape[0]))``.
+              and use its mean and covariance matrix as an initial values of ``hn_m_vecs`` and ``hn_lambda_mats``.
+            * ``'random_responsibility'``: randomly assign responsibility to ``r_vecs``
+            Type of initialization, by default ``'subsampling'``
 
         Returns
         -------
-        Predicted_value : numpy.ndarray
+        predicted_value : numpy.ndarray
             The estimated values under the given loss function. 
         """
         z_hat = self.estimate_latent_vars(x,loss=loss)

bayesml/hiddenmarkovnormal/_hiddenmarkovnormal.py

@@ -391,7 +391,6 @@ def visualize_model(self,sample_length=200):
         [-5.]]
         lambda_mats:
         [[[1.]]
-
         [[1.]]]
 
         .. image:: ./images/hiddenmarkovnormal_example.png
@@ -843,7 +842,7 @@ def _calc_q_lambda_char(self):
                      axis=1) * 2.0
             ) / 2.0
 
-    def calc_vl(self):
+    def _calc_vl(self):
         # E[ln p(X|Z,mu,Lambda)]
         self._vl_p_x = np.sum(
             self.ns
@@ -1025,10 +1024,10 @@ def update_posterior(
         tolerance : float, optional
             convergence criterion of variational lower bound, by default 1.0E-8
         init_type : str, optional
-            type of initialization, by default 'subsampling'
-            * 'subsampling': for each latent class, extract a subsample whose size is int(np.sqrt(x.shape[0])).
+            * ``'subsampling'``: for each latent class, extract a subsample whose size is ``int(np.sqrt(x.shape[0]))``, 
               and use its mean and covariance matrix as an initial values of hn_m_vecs and hn_lambda_mats.
-            * 'random_responsibility': randomly assign responsibility to gamma_vecs
+            * ``'random_responsibility'``: randomly assign responsibility to gamma_vecs
+            Type of initialization, by default 'subsampling'
         """
         _check.float_vecs(x,'x',DataFormatError)
         _check.shape_consistency(
@@ -1069,15 +1068,15 @@ def update_posterior(
                     f'init_type={init_type} is unsupported. '
                     + 'This function supports only '
                     + '"subsampling" and "random_responsibility"'))
-            self.calc_vl()
+            self._calc_vl()
             print(f'\r{i}. VL: {self.vl}',end='')
             for t in range(max_itr):
                 vl_before = self.vl
                 self._update_q_mu_lambda()
                 self._update_q_pi()
                 self._update_q_a()
                 self._update_q_z(x)
-                self.calc_vl()
+                self._calc_vl()
                 print(f'\r{i}. VL: {self.vl} t={t} ',end='')
                 if np.abs((self.vl-vl_before)/vl_before) < tolerance:
                     convergence_flag = False
@@ -1221,11 +1220,9 @@ def visualize_posterior(self):
         [154.15657765  47.84342235]
         hn_w_mats:
         [[[0.00525177]]
-
         [[0.02569298]]]
         E[lambda_mats]=
         [[[0.8095951 ]]
-
         [[1.22924015]]]
 
         .. image:: ./images/hiddenmarkovnormal_posterior.png
@@ -1375,10 +1372,10 @@ def pred_and_update(
         tolerance : float, optional
             convergence croterion of variational lower bound, by default 1.0E-8
         init_type : str, optional
-            type of initialization, by default 'random_responsibility'
-            * 'random_responsibility': randomly assign responsibility to r_vecs
-            * 'subsampling': for each latent class, extract a subsample whose size is int(np.sqrt(x.shape[0])).
+            * ``'random_responsibility'``: randomly assign responsibility to ``xi_mats`` and ``gamma_vecs``
+            * ``'subsampling'``: for each latent class, extract a subsample whose size is ``int(np.sqrt(x.shape[0]))``. 
               and use its mean and covariance matrix as an initial values of hn_m_vecs and hn_lambda_mats.
+            Type of initialization, by default 'random_responsibility'
 
         Returns
         -------
@@ -1508,10 +1505,10 @@ def estimate_latent_vars_and_update(
         tolerance : float, optional
             convergence croterion of variational lower bound, by default 1.0E-8
         init_type : str, optional
-            type of initialization, by default 'random_responsibility'
-            * 'random_responsibility': randomly assign responsibility to r_vecs
-            * 'subsampling': for each latent class, extract a subsample whose size is int(np.sqrt(x.shape[0])).
+            * ``'random_responsibility'``: randomly assign responsibility to ``xi_mats`` and ``gamma_vecs``
+            * ``'subsampling'``: for each latent class, extract a subsample whose size is ``int(np.sqrt(x.shape[0]))``.
               and use its mean and covariance matrix as an initial values of hn_m_vecs and hn_lambda_mats.
+            Type of initialization, by default ``'random_responsibility'``
 
         Returns
         -------

bayesml/metatree/_metatree_x_discrete.py

@@ -596,7 +596,7 @@ def visualize_model(self,filename=None,format=None,sample_size=10):
 
         See Also
         --------
-        graphbiz.Digraph
+        graphviz.Digraph
         """
         #例外処理
         _check.pos_int(sample_size,'sample_size',DataFormatError)
@@ -1208,7 +1208,7 @@ def estimate_params(self,loss="0-1",visualize=True,filename=None,format=None):
 
         See Also
         --------
-        graphbiz.Digraph
+        graphviz.Digraph
         """
 
         if loss == "0-1":
@@ -1293,7 +1293,7 @@ def visualize_posterior(self,filename=None,format=None):
 
         See Also
         --------
-        graphbiz.Digraph
+        graphviz.Digraph
         """
         MAP_index = np.argmax(self.hn_metatree_prob_vec)
         print(f'MAP probability of metatree:{self.hn_metatree_prob_vec[MAP_index]}')

doc/bayesml.autoregressive.rst

@@ -1,6 +1,8 @@
 bayesml.autoregressive package
 ==============================
 
+.. image:: ./images/autoregressive_example.png
+
 Module contents
 ---------------