Release 0.2.0

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
 ```
 
@@ -33,6 +33,7 @@ The following are required.
 * NumPy (>= 1.20)
 * SciPy (>= 1.7)
 * MatplotLib (>= 3.5)
+* Scikit-learn (>= 1.1)
 
 ## Example
 
@@ -53,11 +54,11 @@ gen_model.visualize_model()
 ```
 
 >theta:0.7  
->x0:[1 1 1 1 1 0 1 0 0 1 1 1 1 0 1 1 0 1 1 1]  
->x1:[1 0 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 1 1]  
->x2:[0 0 1 1 0 1 0 1 1 1 1 1 1 0 1 0 1 1 1 1]  
->x3:[1 0 1 1 1 1 1 0 0 0 1 0 0 1 0 1 1 0 1 0]  
->x4:[1 1 0 1 0 1 1 1 0 1 1 1 0 0 1 1 1 1 1 1]  
+>x0:[1 1 1 0 1 1 1 0 1 1 1 1 1 1 0 1 1 1 0 1]  
+>x1:[1 1 0 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 0]  
+>x2:[1 0 1 1 0 1 1 1 0 1 1 1 1 1 0 0 1 1 1 1]  
+>x3:[1 1 1 0 1 1 0 1 0 0 0 0 1 0 1 1 1 1 1 1]  
+>x4:[0 0 1 0 0 0 1 1 1 1 1 1 1 1 0 0 1 1 1 1]  
 >![bernoulli_example1](./doc/images/README_ex_img1.png)
 
 After confirming that the frequency of occurrence of 1 is around `theta=0.7`, we generate a sample and store it to variable `x`.
@@ -99,9 +100,9 @@ print(learn_model.estimate_params(loss='abs'))
 print(learn_model.estimate_params(loss='0-1'))
 ```
 
->0.6428571428571429  
->0.6474720009710451  
->0.6578947368421053  
+>0.7380952380952381  
+>0.7457656349087012  
+>0.7631578947368421  
 
 Different settings of a loss function yield different optimal estimates.
 
@@ -115,8 +116,12 @@ The following packages are currently available. In this library, a probabilistic
 * [Normal model](https://yuta-nakahara.github.io/BayesML/bayesml.normal.html "BayesML Normal Model")
 * [Multivariate normal model](https://yuta-nakahara.github.io/BayesML/bayesml.multivariate_normal.html "BayesML Multivariate Normal Model")
 * [Exponential model](https://yuta-nakahara.github.io/BayesML/bayesml.exponential.html "BayesML Exponential Model")
+* [Gaussian mixture model](https://yuta-nakahara.github.io/BayesML/bayesml.gaussianmixture.html "BayesML Gaussian Mixture Model")
 * [Linear regression model](https://yuta-nakahara.github.io/BayesML/bayesml.linearregression.html "BayesML Lenear Regression Model")
+* [Meta-tree model](https://yuta-nakahara.github.io/BayesML/bayesml.metatree.html "BayesML Meta-tree Model")
 * [Autoregressive model](https://yuta-nakahara.github.io/BayesML/bayesml.autoregressive.html "BayesML Autoregressive Model")
+* [Hidden Markov normal model](https://yuta-nakahara.github.io/BayesML/bayesml.hiddenmarkovnormal.html "BayesML Hidden Markov Normal Model")
+* [Context tree model](https://yuta-nakahara.github.io/BayesML/bayesml.contexttree.html "BayesML Context Tree Model")
 
 In the future, we will add packages to deal with a mixture normal model and a hidden Markov model, which are difficult to perform exact Bayesian inference, by using variational Bayes methods.
 
@@ -131,11 +136,8 @@ When you use BayesML for your academic work, please provide the following biblio
 Plain text
 
 ```
-Y. Nakahara, N. Ichijo, K. Shimada, 
-K. Tajima, K. Horinouchi, L. Ruan, 
-N. Namegaya, R. Maniwa, T. Ishiwatari, 
-W. Yu, Y. Iikubo, S. Saito, 
-K. Kazama, T. Matsushima, ``BayesML,'' 
+Y. Nakahara, N. Ichijo, K. Shimada, Y. Iikubo, 
+S. Saito, K. Kazama, T. Matsushima, ``BayesML 0.2.0,'' 
 [Online] https://github.com/yuta-nakahara/BayesML
 ```
 
@@ -144,11 +146,9 @@ BibTeX
 ``` bibtex
 @misc{bayesml,
   author = {Nakahara Yuta and Ichijo Naoki and Shimada Koshi and
-            Tajima Keito and Horinouchi Kohei and Ruan Luyu and
-            Namegaya Noboru and Maniwa Ryota and Ishiwatari Taisuke and
-            Yu Wenbin and Iikubo Yuji and Saito Shota and Kazama Koki and
+            Iikubo Yuji and Saito Shota and Kazama Koki and
             Matsushima Toshiyasu}
-  title = {BayesML},
+  title = {BayesML 0.2.0},
   howpublished = {\url{https://github.com/yuta-nakahara/BayesML}},
   year = {2022}
 }

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
 ```
 
@@ -32,6 +32,7 @@ BayesMLの実行には以下が必要です．
 * NumPy (>= 1.20)
 * SciPy (>= 1.7)
 * MatplotLib (>= 3.5)
+* Scikit-learn (>= 1.1)
 
 ## 実行例
 
@@ -52,11 +53,11 @@ gen_model.visualize_model()
 ```
 
 >theta:0.7  
->x0:[1 1 1 1 1 0 1 0 0 1 1 1 1 0 1 1 0 1 1 1]  
->x1:[1 0 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 1 1]  
->x2:[0 0 1 1 0 1 0 1 1 1 1 1 1 0 1 0 1 1 1 1]  
->x3:[1 0 1 1 1 1 1 0 0 0 1 0 0 1 0 1 1 0 1 0]  
->x4:[1 1 0 1 0 1 1 1 0 1 1 1 0 0 1 1 1 1 1 1]  
+>x0:[1 1 1 0 1 1 1 0 1 1 1 1 1 1 0 1 1 1 0 1]  
+>x1:[1 1 0 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 0]  
+>x2:[1 0 1 1 0 1 1 1 0 1 1 1 1 1 0 0 1 1 1 1]  
+>x3:[1 1 1 0 1 1 0 1 0 0 0 0 1 0 1 1 1 1 1 1]  
+>x4:[0 0 1 0 0 0 1 1 1 1 1 1 1 1 0 0 1 1 1 1]  
 >![bernoulli_example1](./doc/images/README_ex_img1.png)
 
 1の出現頻度が`theta=0.7`程度であることを確認したら，サンプルを生成し変数`x`に保存します．
@@ -96,9 +97,9 @@ print(learn_model.estimate_params(loss='abs'))
 print(learn_model.estimate_params(loss='0-1'))
 ```
 
->0.6428571428571429  
->0.6474720009710451  
->0.6578947368421053  
+>0.7380952380952381  
+>0.7457656349087012  
+>0.7631578947368421  
 
 損失関数の設定が異なると，そのもとでの最適な推定値も異なることがわかります．
 
@@ -112,8 +113,12 @@ print(learn_model.estimate_params(loss='0-1'))
 * [正規モデル](https://yuta-nakahara.github.io/BayesML/bayesml.normal.html "BayesML Normal Model")
 * [多変量正規モデル](https://yuta-nakahara.github.io/BayesML/bayesml.multivariate_normal.html "BayesML Multivariate Normal Model")
 * [指数モデル](https://yuta-nakahara.github.io/BayesML/bayesml.exponential.html "BayesML Exponential Model")
+* [混合正規モデル](https://yuta-nakahara.github.io/BayesML/bayesml.gaussianmixture.html "BayesML Gaussian Mixture Model")
 * [線形回帰モデル](https://yuta-nakahara.github.io/BayesML/bayesml.linearregression.html "BayesML Lenear Regression Model")
+* [メタツリーモデル](https://yuta-nakahara.github.io/BayesML/bayesml.metatree.html "BayesML Meta-tree Model")
 * [自己回帰モデル](https://yuta-nakahara.github.io/BayesML/bayesml.autoregressive.html "BayesML Autoregressive Model")
+* [隠れマルコフモデル](https://yuta-nakahara.github.io/BayesML/bayesml.hiddenmarkovnormal.html "BayesML Hidden Markov Normal Model")
+* [文脈木モデル](https://yuta-nakahara.github.io/BayesML/bayesml.contexttree.html "BayesML Context Tree Model")
 
 また，今後は混合正規モデルや隠れマルコフモデルなどの厳密なベイズ推論が困難なモデルを変分ベイズ法で学習するパッケージが追加される予定です．
 
@@ -128,11 +133,8 @@ BayesMLへのコントリビューションを考えてくださってありが
 プレーンテキスト
 
 ```
-Y. Nakahara, N. Ichijo, K. Shimada, 
-K. Tajima, K. Horinouchi, L. Ruan, 
-N. Namegaya, R. Maniwa, T. Ishiwatari, 
-W. Yu, Y. Iikubo, S. Saito, 
-K. Kazama, T. Matsushima, ``BayesML,'' 
+Y. Nakahara, N. Ichijo, K. Shimada, Y. Iikubo, 
+S. Saito, K. Kazama, T. Matsushima, ``BayesML,'' 
 [Online] https://github.com/yuta-nakahara/BayesML
 ```
 
@@ -141,9 +143,7 @@ BibTeX
 ``` bibtex
 @misc{bayesml,
   author = {Nakahara Yuta and Ichijo Naoki and Shimada Koshi and
-            Tajima Keito and Horinouchi Kohei and Ruan Luyu and
-            Namegaya Noboru and Maniwa Ryota and Ishiwatari Taisuke and
-            Yu Wenbin and Iikubo Yuji and Saito Shota and Kazama Koki and
+            Iikubo Yuji and Saito Shota and Kazama Koki and
             Matsushima Toshiyasu}
   title = {BayesML},
   howpublished = {\url{https://github.com/yuta-nakahara/BayesML}},

bayesml/__init__.py

@@ -7,6 +7,9 @@
 from . import normal
 from . import poisson
 from . import metatree
+from . import contexttree
+from . import gaussianmixture
+from . import hiddenmarkovnormal
 
 __all__ = ['bernoulli',
            'categorical',
@@ -16,5 +19,8 @@
            'multivariate_normal',
            'normal',
            'poisson',
-           'metatree'
+           'metatree',
+           'contexttree',
+           'gaussianmixture',
+           'hiddenmarkovnormal',
            ]

bayesml/bernoulli/_bernoulli.py

@@ -368,7 +368,7 @@ def visualize_posterior(self):
         p_range = np.linspace(0,1,100,endpoint=False)
         fig, ax = plt.subplots()
         ax.plot(p_range,self.estimate_params(loss="KL").pdf(p_range))
-        ax.set_xlabel("p_theta")
+        ax.set_xlabel("theta")
         ax.set_ylabel("posterior")
         plt.show()
 

bayesml/categorical/_categorical.py

@@ -349,7 +349,7 @@ def update_posterior(self, x):
             2-dimensional array whose shape is ``(sample_size,degree)`` whose rows are one-hot vectors.
         """
         _check.onehot_vecs(x,'x',DataFormatError)
-        if self.degree > 1 and x.shape[-1] != self.degree:
+        if x.shape[-1] != self.degree:
             raise(DataFormatError(f"x.shape[-1] must be degree:{self.degree}"))
         x = x.reshape(-1,self.degree)
 

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