Merge pull request #85 from bayesml/release-0.3.0

Release 0.3.0

Author: yuta-nakahara

README.md

@@ -3,6 +3,10 @@ Document Author
 Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
 Shota Saito <shota.s@gunma-u.ac.jp>
 -->
+<div align="center">
+<b>Our algorithm for the meta-tree model is accepted at AISTATS 2025! Click <a href="https://bayesml.github.io/BayesML/examples/metatree_prediction_interval.html">here</a>!</b>
+</div>
+
 <img src="./doc/logos/BayesML_logo.png" width="600">
 
 ## Purpose
@@ -17,7 +21,11 @@ BayesML has the following characteristics.
 * Many of our learning algorithms are much faster than general-purpose Bayesian learning algorithms such as MCMC methods because they effectively use the conjugate property of a probabilistic data generative model and a prior distribution. Moreover, they are suitable for online learning.
 * All packages have methods to visualize the probabilistic data generative model, generated data from that model, and the posterior distribution learned from the data in 2~3 dimensional space. Thus, you can effectively understand the characteristics of probabilistic data generative models and algorithms through the generation of synthetic data and learning from them.
 
-For more details, see our [website](https://yuta-nakahara.github.io/BayesML/ "BayesML's Documentation").
+For more details, see our [website](https://bayesml.github.io/BayesML/ "BayesML's Documentation").
+
+## News
+
+* Our algorithm for the meta-tree model is accepted at AISTATS 2025! A sample code is [here](https://bayesml.github.io/BayesML/examples/metatree_prediction_interval.html).
 
 ## Installation
 
@@ -110,18 +118,18 @@ Different settings of a loss function yield different optimal estimates.
 
 The following packages are currently available. In this library, a probabilistic data generative model, prior distribution, posterior distribution (or approximate posterior distribution), and predictive distribution (or approximate predictive distribution) are collectively called a model.
 
-* [Bernoulli model](https://yuta-nakahara.github.io/BayesML/bayesml.bernoulli.html "Bayesml Bernoulli Model")
-* [Categorical model](https://yuta-nakahara.github.io/BayesML/bayesml.categorical.html "BayesML Categorical Model")
-* [Poisson model](https://yuta-nakahara.github.io/BayesML/bayesml.poisson.html "BayesML Poisson Model")
-* [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")
+* [Bernoulli model](https://bayesml.github.io/BayesML/bayesml.bernoulli.html "Bayesml Bernoulli Model")
+* [Categorical model](https://bayesml.github.io/BayesML/bayesml.categorical.html "BayesML Categorical Model")
+* [Poisson model](https://bayesml.github.io/BayesML/bayesml.poisson.html "BayesML Poisson Model")
+* [Normal model](https://bayesml.github.io/BayesML/bayesml.normal.html "BayesML Normal Model")
+* [Multivariate normal model](https://bayesml.github.io/BayesML/bayesml.multivariate_normal.html "BayesML Multivariate Normal Model")
+* [Exponential model](https://bayesml.github.io/BayesML/bayesml.exponential.html "BayesML Exponential Model")
+* [Gaussian mixture model](https://bayesml.github.io/BayesML/bayesml.gaussianmixture.html "BayesML Gaussian Mixture Model")
+* [Linear regression model](https://bayesml.github.io/BayesML/bayesml.linearregression.html "BayesML Lenear Regression Model")
+* [Meta-tree model](https://bayesml.github.io/BayesML/bayesml.metatree.html "BayesML Meta-tree Model")
+* [Autoregressive model](https://bayesml.github.io/BayesML/bayesml.autoregressive.html "BayesML Autoregressive Model")
+* [Hidden Markov normal model](https://bayesml.github.io/BayesML/bayesml.hiddenmarkovnormal.html "BayesML Hidden Markov Normal Model")
+* [Context tree model](https://bayesml.github.io/BayesML/bayesml.contexttree.html "BayesML Context Tree Model")
 
 In the future, we will add packages to deal with more complicated hierarchical models.
 
@@ -137,8 +145,9 @@ Plain text
 
 ```
 Y. Nakahara, N. Ichijo, K. Shimada, Y. Iikubo, 
-S. Saito, K. Kazama, T. Matsushima, BayesML Developers, ``BayesML 0.2.5,'' 
-[Online] https://github.com/yuta-nakahara/BayesML
+S. Saito, K. Kazama, T. Matsushima, BayesML Developers, ``BayesML,'' 
+Python package version 0.3.0, 
+[Online] https://github.com/bayesml/BayesML
 ```
 
 BibTeX
@@ -148,8 +157,9 @@ BibTeX
   author = {Nakahara, Yuta and Ichijo, Naoki and Shimada, Koshi and
             Iikubo, Yuji and Saito, Shota and Kazama, Koki and
             Matsushima, Toshiyasu and {BayesML Developers}},
-  title = {BayesML 0.2.5},
-  howpublished = {\url{https://github.com/yuta-nakahara/BayesML}},
-  year = {2022}
+  title = {{BayesML}},
+  howpublished = {Python package version 0.3.0},
+  note = {\url{https://github.com/bayesml/BayesML}},
+  year = {2025}
 }
 ```

README_jp.md

@@ -2,6 +2,10 @@
 Document Author
 Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
 -->
+<div align="center">
+<b>メタツリーモデルに対するアルゴリズムがAISTATS 2025に採択！詳細は<a href="https://bayesml.github.io/BayesML/examples/metatree_prediction_interval.html">こちら</a>！</b>
+</div>
+
 <img src="./doc/logos/BayesML_logo.png" width="600">
 
 ## 目的
@@ -16,7 +20,11 @@ BayesMLは以下の特徴を持っています．
 * 学習アルゴリズムの多くはデータ生成確率モデルと事前分布の共役性を効果的に用いているため，MCMC法などの汎用的なベイズ学習アルゴリズムと比べて非常に高速で，オンライン学習にも適しています．
 * 2~3次元空間上で，データ生成確率モデル，生成された人工データ，学習された事後分布を可視化するメソッドが全てのパッケージに備わっています．そのため，人工データの生成と学習を通じて確率モデル，アルゴリズムの特性を効果的に把握することができます．
 
-詳細は[Webサイト](https://yuta-nakahara.github.io/BayesML/ "BayesML's Documentation")をご覧ください．
+詳細は[Webサイト](https://bayesml.github.io/BayesML/ "BayesML's Documentation")をご覧ください．
+
+## ニュース
+
+* メタツリーモデルに対するアルゴリズムがAISTATS 2025に採択されました．サンプルコードは[こちら](https://bayesml.github.io/BayesML/examples/metatree_prediction_interval.html)をご覧ください．
 
 ## インストール
 
@@ -107,18 +115,18 @@ print(learn_model.estimate_params(loss='0-1'))
 
 現在，以下のモデルに関するパッケージが利用可能です．本ライブラリでは，データ生成確率モデル，事前分布，事後分布（または近似事後分布），予測分布（または近似予測分布）を合わせてモデルと呼んでいます．
 
-* [ベルヌーイモデル](https://yuta-nakahara.github.io/BayesML/bayesml.bernoulli.html "Bayesml Bernoulli Model")
-* [カテゴリカルモデル](https://yuta-nakahara.github.io/BayesML/bayesml.categorical.html "BayesML Categorical Model")
-* [ポアソンモデル](https://yuta-nakahara.github.io/BayesML/bayesml.poisson.html "BayesML Poisson Model")
-* [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")
+* [ベルヌーイモデル](https://bayesml.github.io/BayesML/bayesml.bernoulli.html "Bayesml Bernoulli Model")
+* [カテゴリカルモデル](https://bayesml.github.io/BayesML/bayesml.categorical.html "BayesML Categorical Model")
+* [ポアソンモデル](https://bayesml.github.io/BayesML/bayesml.poisson.html "BayesML Poisson Model")
+* [正規モデル](https://bayesml.github.io/BayesML/bayesml.normal.html "BayesML Normal Model")
+* [多変量正規モデル](https://bayesml.github.io/BayesML/bayesml.multivariate_normal.html "BayesML Multivariate Normal Model")
+* [指数モデル](https://bayesml.github.io/BayesML/bayesml.exponential.html "BayesML Exponential Model")
+* [混合正規モデル](https://bayesml.github.io/BayesML/bayesml.gaussianmixture.html "BayesML Gaussian Mixture Model")
+* [線形回帰モデル](https://bayesml.github.io/BayesML/bayesml.linearregression.html "BayesML Lenear Regression Model")
+* [メタツリーモデル](https://bayesml.github.io/BayesML/bayesml.metatree.html "BayesML Meta-tree Model")
+* [自己回帰モデル](https://bayesml.github.io/BayesML/bayesml.autoregressive.html "BayesML Autoregressive Model")
+* [隠れマルコフモデル](https://bayesml.github.io/BayesML/bayesml.hiddenmarkovnormal.html "BayesML Hidden Markov Normal Model")
+* [文脈木モデル](https://bayesml.github.io/BayesML/bayesml.contexttree.html "BayesML Context Tree Model")
 
 また，今後はより複雑な階層的モデルを取り扱うパッケージを追加していく予定です．
 
@@ -134,8 +142,9 @@ 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
 
 ```
 Y. Nakahara, N. Ichijo, K. Shimada, Y. Iikubo, 
-S. Saito, K. Kazama, T. Matsushima, BayesML Developers, ``BayesML 0.2.5,'' 
-[Online] https://github.com/yuta-nakahara/BayesML
+S. Saito, K. Kazama, T. Matsushima, BayesML Developers, ``BayesML,'' 
+Python package version 0.3.0, 
+[Online] https://github.com/bayesml/BayesML
 ```
 
 BibTeX
@@ -145,8 +154,9 @@ BibTeX
   author = {Nakahara, Yuta and Ichijo, Naoki and Shimada, Koshi and
             Iikubo, Yuji and Saito, Shota and Kazama, Koki and
             Matsushima, Toshiyasu and {BayesML Developers}},
-  title = {BayesML 0.2.5},
-  howpublished = {\url{https://github.com/yuta-nakahara/BayesML}},
-  year = {2022}
+  title = {{BayesML}},
+  howpublished = {Python package version 0.3.0},
+  note = {\url{https://github.com/bayesml/BayesML}},
+  year = {2025}
 }
 ```

bayesml/_check.py

@@ -1,5 +1,5 @@
 # Code Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
 # Yuji Iikubo <yuji-iikubo.8@fuji.waseda.jp>
 # Yasushi Esaki <esakiful@gmail.com>
 # Jun Nishikawa <jun.b.nishikawa@gmail.com>

bayesml/_exceptions.py

@@ -1,5 +1,5 @@
 # Code Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
 class ParameterFormatError(Exception):
     def __init__(self, value):
         self.value = value

bayesml/autoregressive/__init__.py

@@ -1,6 +1,6 @@
 # Document Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
-# Koki Kazama <kokikazama@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
+# Koki Kazama <kazama@info.shonan-it.ac.jp>
 r"""
 The linear autoregressive model with the normal-gamma prior distribution.
 

bayesml/autoregressive/_autoregressive.py

@@ -1,7 +1,7 @@
 # Code Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
 # Document Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
 import warnings
 import numpy as np
 from scipy.stats import gamma as ss_gamma
@@ -475,31 +475,32 @@ def update_posterior(self,x,padding=None):
             If \"zeros\" is given, the zero vector is used as a initial value.
         """
         _check.float_vec(x,'x',DataFormatError)
-        if x.shape[0] <= self.c_degree:
+        n = x.shape[0]
+        if n <= self.c_degree:
             raise(DataFormatError("The length of x must greater than self.c_degree"))
-        x_mat = np.zeros((x.shape[0],self.c_degree+1))
+        x_mat = np.zeros((n,self.c_degree+1))
         x_mat[:,0] = 1.0
-        for n in range(1,self.c_degree+1):
-            x_mat[n,-n:] = x[:n]
-        for n in range(self.c_degree+1,x.shape[0]):
-            x_mat[n,1:] = x[n-self.c_degree:n]
+        for i in range(1,self.c_degree+1):
+            x_mat[i,-i:] = x[:i]
+        for i in range(self.c_degree+1,n):
+            x_mat[i,1:] = x[i-self.c_degree:i]
 
         mu_tmp = np.array(self.hn_mu_vec)
         lambda_tmp = np.array(self.hn_lambda_mat)
         if padding == "zeros":
             self.hn_lambda_mat += x_mat.T @ x_mat
             self.hn_mu_vec[:] = np.linalg.solve(self.hn_lambda_mat,
-                                                lambda_tmp @ mu_tmp[:,np.newaxis] + x_mat.T @ x[:,np.newaxis])[:,0]
-            self.hn_alpha += x.shape[0] / 2.0
-            self.hn_beta += (-self.hn_mu_vec[np.newaxis,:] @ self.hn_lambda_mat @ self.hn_mu_vec[:,np.newaxis]
-                            + x @ x + mu_tmp[np.newaxis,:] @ lambda_tmp @ mu_tmp[:,np.newaxis])[0,0] / 2.0
+                                                lambda_tmp @ mu_tmp + x_mat.T @ x)
+            self.hn_alpha += n / 2.0
+            self.hn_beta += (-self.hn_mu_vec @ self.hn_lambda_mat @ self.hn_mu_vec
+                            + x @ x + mu_tmp @ lambda_tmp @ mu_tmp) / 2.0
         else:
             self.hn_lambda_mat += x_mat[self.c_degree:].T @ x_mat[self.c_degree:]
             self.hn_mu_vec[:] = np.linalg.solve(self.hn_lambda_mat,
-                                                lambda_tmp @ mu_tmp[:,np.newaxis] + x_mat[self.c_degree:].T @ x[self.c_degree:,np.newaxis])[:,0]
-            self.hn_alpha += (x.shape[0]-self.c_degree) / 2.0
-            self.hn_beta += (-self.hn_mu_vec[np.newaxis,:] @ self.hn_lambda_mat @ self.hn_mu_vec[:,np.newaxis] 
-                            + x[self.c_degree:] @ x[self.c_degree:] + mu_tmp[np.newaxis,:] @ lambda_tmp @ mu_tmp[:,np.newaxis])[0,0] / 2.0
+                                                lambda_tmp @ mu_tmp + x_mat[self.c_degree:].T @ x[self.c_degree:])
+            self.hn_alpha += (n-self.c_degree) / 2.0
+            self.hn_beta += (-self.hn_mu_vec @ self.hn_lambda_mat @ self.hn_mu_vec 
+                            + x[self.c_degree:] @ x[self.c_degree:] + mu_tmp @ lambda_tmp @ mu_tmp) / 2.0
         return self
 
     def estimate_params(self,loss="squared"):

bayesml/base.py

@@ -1,5 +1,5 @@
 # Code Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
 
 from abc import ABCMeta, abstractmethod
 import pickle

bayesml/bernoulli/__init__.py

@@ -1,6 +1,6 @@
 # Document Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
-# Koki Kazama <kokikazama@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
+# Koki Kazama <kazama@info.shonan-it.ac.jp>
 r"""
 The Bernoulli distribution with the beta prior distribution.
 

bayesml/bernoulli/_bernoulli.py

@@ -1,7 +1,7 @@
 # Code Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
 # Document Author
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
 import warnings
 import numpy as np
 from scipy.stats import beta as ss_beta

bayesml/categorical/__init__.py

@@ -1,8 +1,8 @@
 # Document Author
 # Kohei Horinouchi <horinochi_18@toki.waseda.jp>
 # Naoki Ichijo <1jonao@fuji.waseda.jp>
-# Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
-# Koki Kazama <kokikazama@aoni.waseda.jp>
+# Yuta Nakahara <y.nakahara@waseda.jp>
+# Koki Kazama <kazama@info.shonan-it.ac.jp>
 r"""
 The categorical distribution with the dirichlet prior distribution