Fix bugs

Author: yuta-nakahara

bayesml/_test.py

@@ -0,0 +1,9 @@
+from bayesml import contexttree
+
+gen_model = contexttree.GenModel(c_k=2,c_d_max=3,h_g=0.75)
+gen_model.gen_params()
+gen_model.visualize_model()
+x = gen_model.gen_sample(500)
+learn_model = contexttree.LearnModel(c_k=2,c_d_max=3)
+learn_model.update_posterior(x)
+# learn_model.visualize_posterior()

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/linearregression/_linearregression.py

@@ -542,7 +542,7 @@ def update_posterior(self, x, y):
             float array.
         """
         _check.float_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}"))
         _check.floats(y,'y',DataFormatError)
         if type(y) is np.ndarray:

bayesml/metatree/_metatree_x_discrete.py

@@ -173,7 +173,8 @@ def __init__(
         )
 
         # params
-        self.root = _GenNode(0,list(range(self.c_k)),self.c_num_children,self.h_g,None,None)
+        self.root = _GenNode(0,list(range(self.c_k)),self.c_num_children,self.h_g,0,self.SubModel(**self.sub_h_params))
+        self.root.leaf = True
 
         self.set_params(root)
 
@@ -1282,6 +1283,7 @@ def visualize_posterior(self,filename=None,format=None):
         --------
         >>> from bayesml import metatree
         >>> gen_model = metatree.GenModel(c_k=3,h_g=0.75)
+        >>> gen_model.gen_params()
         >>> x,y = gen_model.gen_sample(500)
         >>> learn_model = metatree.LearnModel(c_k=3)
         >>> learn_model.update_posterior(x,y)

bayesml/multivariate_normal/_multivariatenormal.py

@@ -614,7 +614,7 @@ def update_posterior(self,x):
             All the elements must be real number.
         """
         _check.float_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)
 
@@ -721,7 +721,7 @@ def visualize_posterior(self):
         print(f"{self.hn_w_mat}")
         print("E[lambda_mat]=")
         print(f"{self.hn_nu * self.hn_w_mat}")
-        mu_vec_pdf, w_mat_pdf = self.estimate_params(loss="KL")
+        mu_vec_pdf, lambda_mat_pdf = self.estimate_params(loss="KL")
         if self.degree == 1:
             fig, axes = plt.subplots(1,2)
             # for mu_vec
@@ -732,12 +732,12 @@ def visualize_posterior(self):
             axes[0].set_xlabel("mu_vec")
             axes[0].set_ylabel("Density")
             # for lambda_mat
-            x = np.linspace(max(1.0e-8,self.hn_nu*self.hn_w_mat)-4.0*np.sqrt(self.hn_nu/2.0)*(2.0*self.hn_w_mat),
+            x = np.linspace(max(1.0e-8,self.hn_nu*self.hn_w_mat-4.0*np.sqrt(self.hn_nu/2.0)*(2.0*self.hn_w_mat)),
                             self.hn_nu*self.hn_w_mat+4.0*np.sqrt(self.hn_nu/2.0)*(2.0*self.hn_w_mat),
                             100)
             print(self.hn_w_mat)
-            axes[1].plot(x[:,0,0],w_mat_pdf.pdf(x[:,0,0]))
-            axes[1].set_xlabel("w_mat")
+            axes[1].plot(x[:,0,0],lambda_mat_pdf.pdf(x[:,0,0]))
+            axes[1].set_xlabel("lambda_mat")
             axes[1].set_ylabel("Density")
 
             fig.tight_layout()