Merge pull request #2 from yuta-nakahara/develop-categorical

Develop categorical model

Author: yuta-nakahara

bayesml/__init__.py

@@ -1,4 +1,5 @@
 from . import bernoulli
+from . import categorical
 from . import autoregressive
 from . import exponential
 from . import linearregression
@@ -7,6 +8,7 @@
 from . import poisson
 
 __all__ = ['bernoulli',
+           'categorical',
            'autoregressive',
            'exponential',
            'linearregression',

bayesml/_check.py

@@ -3,35 +3,34 @@
 # Yuji Iikubo <yuji-iikubo.8@fuji.waseda.jp>
 import numpy as np
 
-FLOATS = list({'float128','float64','float32','float16'} & set(dir(np)) | {float})
-INTS = list({'int64','int32','int16','int8'} & set(dir(np)) | {int})
+_EPSILON = np.sqrt(np.finfo(np.float64).eps)
 
 def float_in_closed01(val,val_name,exception_class):
-    if type(val) in FLOATS:
+    if np.issubdtype(type(val),np.floating):
         if val >= 0.0 and val <= 1.0:
             return val
-    if type(val) in INTS:
+    if np.issubdtype(type(val),np.integer):
         if val >= 0.0 and val <= 1.0:
             return float(val)
     raise(exception_class(val_name + " must be in [0,1]."))
 
 def pos_float(val,val_name,exception_class):
-    if type(val) in FLOATS:
+    if np.issubdtype(type(val),np.floating):
         if val > 0.0:
             return val
-    if type(val) in INTS:
+    if np.issubdtype(type(val),np.integer):
         if val > 0.0:
             return float(val)
     raise(exception_class(val_name + " must be positive (not including 0.0)."))
 
 def pos_int(val,val_name,exception_class):
-    if type(val) in INTS:
+    if np.issubdtype(type(val),np.integer):
         if val > 0:
             return val
     raise(exception_class(val_name + " must be int. Its value must be positive (not including 0)."))
 
 def nonneg_int(val,val_name,exception_class):
-    if type(val) in INTS:
+    if np.issubdtype(type(val),np.integer):
         if val >= 0:
             return val
     raise(exception_class(val_name + " must be int. Its value must be non-negative (including 0)."))
@@ -42,18 +41,18 @@ def nonneg_ints(val,val_name,exception_class):
     except:
         pass
     if type(val) is np.ndarray:
-        if val.dtype in INTS and np.all(val>=0):
+        if np.issubdtype(val.dtype,np.integer) and np.all(val>=0):
             return val
     raise(exception_class(val_name + " must be int or a numpy.ndarray whose dtype is int. Its values must be non-negative (including 0)."))
 
 def nonneg_int_vec(val,val_name,exception_class):
     if type(val) is np.ndarray:
-        if val.dtype in INTS and val.ndim == 1 and np.all(val>=0):
+        if np.issubdtype(val.dtype,np.integer) and val.ndim == 1 and np.all(val>=0):
             return val
     raise(exception_class(val_name + " must be a 1-dimensional numpy.ndarray whose dtype is int. Its values must be non-negative (including 0)."))
 
 def int_of_01(val,val_name,exception_class):
-    if type(val) in INTS:
+    if np.issubdtype(type(val),np.integer):
         if val == 0 or val ==1:
             return val
     raise(exception_class(val_name + " must be int. Its value must be 0 or 1."))
@@ -64,23 +63,23 @@ def ints_of_01(val,val_name,exception_class):
     except:
         pass
     if type(val) is np.ndarray:
-        if val.dtype in INTS and np.all(val >= 0) and np.all(val <= 1):
+        if np.issubdtype(val.dtype,np.integer) and np.all(val >= 0) and np.all(val <= 1):
             return val
     raise(exception_class(val_name + " must be int or a numpy.ndarray whose dtype is int. Its values must be 0 or 1."))
 
 def int_vec_of_01(val,val_name,exception_class):
     if type(val) is np.ndarray:
-        if val.dtype in INTS and val.ndim == 1 and np.all(val >= 0) and np.all(val <= 1):
+        if np.issubdtype(val.dtype,np.integer) and val.ndim == 1 and np.all(val >= 0) and np.all(val <= 1):
             return val
     raise(exception_class(val_name + " must be a 1-dimensional numpy.ndarray whose dtype is int. Its values must be 0 or 1."))
 
 def scalar(val,val_name,exception_class):
-    if type(val) in INTS or type(val) in FLOATS:
+    if np.issubdtype(type(val),np.integer) or np.issubdtype(type(val),np.floating):
         return val
     raise(exception_class(val_name + " must be a scalar."))
 
 def pos_scalar(val,val_name,exception_class):
-    if type(val) in INTS or type(val) in FLOATS:
+    if np.issubdtype(type(val),np.integer) or np.issubdtype(type(val),np.floating):
         if val > 0.0:
             return val
     raise(exception_class(val_name + " must be a positive scalar."))
@@ -102,9 +101,9 @@ def pos_def_sym_mat(val,val_name,exception_class):
     raise(exception_class(val_name + " must be a positive definite symmetric 2-dimensional numpy.ndarray."))
 
 def float_(val,val_name,exception_class):
-    if type(val) in FLOATS:
+    if np.issubdtype(type(val),np.floating):
         return val
-    if type(val) in INTS:
+    if np.issubdtype(type(val),np.integer):
         return float(val)
     raise(exception_class(val_name + " must be a scalar."))
 
@@ -114,9 +113,9 @@ def floats(val,val_name,exception_class):
     except:
         pass
     if type(val) is np.ndarray:
-        if val.dtype in INTS:
+        if np.issubdtype(val.dtype,np.integer):
             return val.astype(float)
-        if val.dtype in FLOATS:
+        if np.issubdtype(val.dtype,np.floating):
             return val
     raise(exception_class(val_name + " must be float or a numpy.ndarray."))
 
@@ -126,33 +125,67 @@ def pos_floats(val,val_name,exception_class):
     except:
         pass
     if type(val) is np.ndarray:
-        if val.dtype in INTS and np.all(val>0):
+        if np.issubdtype(val.dtype,np.integer) and np.all(val>0):
             return val.astype(float)
-        if val.dtype in FLOATS and np.all(val>0.0):
+        if np.issubdtype(val.dtype,np.floating) and np.all(val>0.0):
             return val
     raise(exception_class(val_name + " must be float or a numpy.ndarray. Its values must be positive (not including 0)"))
 
 def float_vec(val,val_name,exception_class):
     if type(val) is np.ndarray:
-        if val.dtype in INTS and val.ndim == 1:
+        if np.issubdtype(val.dtype,np.integer) and val.ndim == 1:
             return val.astype(float)
-        if val.dtype in FLOATS and val.ndim == 1:
+        if np.issubdtype(val.dtype,np.floating) and val.ndim == 1:
             return val
     raise(exception_class(val_name + " must be a 1-dimensional numpy.ndarray."))
 
 def pos_float_vec(val,val_name,exception_class):
     if type(val) is np.ndarray:
-        if val.dtype in INTS and val.ndim == 1 and np.all(val>0):
+        if np.issubdtype(val.dtype,np.integer) and val.ndim == 1 and np.all(val>0):
             return val.astype(float)
-        if val.dtype in FLOATS and val.ndim == 1 and np.all(val>0.0):
+        if np.issubdtype(val.dtype,np.floating) and val.ndim == 1 and np.all(val>0.0):
             return val
     raise(exception_class(val_name + " must be a 1-dimensional numpy.ndarray. Its values must be positive (not including 0)"))
 
 def float_vecs(val,val_name,exception_class):
     if type(val) is np.ndarray:
-        if val.dtype in INTS and val.ndim >= 1:
+        if np.issubdtype(val.dtype,np.integer) and val.ndim >= 1:
             return val.astype(float)
-        if val.dtype in FLOATS and val.ndim >= 1:
+        if np.issubdtype(val.dtype,np.floating) and val.ndim >= 1:
             return val
     raise(exception_class(val_name + " must be a numpy.ndarray whose ndim >= 1."))
 
+def float_vec_sum_1(val,val_name,exception_class):
+    if type(val) is np.ndarray:
+        if np.issubdtype(val.dtype,np.integer) and val.ndim == 1 and abs(val.sum() - 1.) <= _EPSILON:
+            return val.astype(float)
+        if np.issubdtype(val.dtype,np.floating) and val.ndim == 1 and abs(val.sum() - 1.) <= _EPSILON:
+            return val
+    raise(exception_class(val_name + " must be a 1-dimensional numpy.ndarray, and the sum of its elements must equal to 1."))
+
+def int_(val,val_name,exception_class):   
+    if np.issubdtype(type(val),np.integer):
+        return val
+    raise(exception_class(val_name + " must be an integer."))
+
+def ints(val,val_name,exception_class):
+    try:
+        return int_(val,val_name,exception_class)
+    except:
+        pass
+    if type(val) is np.ndarray:
+        if np.issubdtype(val.dtype,np.integer):
+            return val
+    raise(exception_class(val_name + " must be int or a numpy.ndarray whose dtype is int."))
+
+def onehot_vec(val,val_name,exception_class):
+    if type(val) is np.ndarray:
+        if np.issubdtype(val.dtype,np.integer) and val.ndim == 1 and np.all(val >= 0) and val.sum()==1:
+            return val
+    raise(exception_class(val_name + " must be a one-hot vector (1-dimensional ndarray) whose dtype must be int."))
+
+def onehot_vecs(val,val_name,exception_class):
+    if type(val) is np.ndarray:
+        if np.issubdtype(val.dtype,np.integer) and val.ndim >= 1 and np.all(val >= 0) and np.all(val.sum(axis=-1)==1):
+            return val
+    raise(exception_class(val_name + " must be a numpy.ndarray whose dtype is int and whose last axis constitutes one-hot vectors."))

bayesml/bernoulli/_bernoulli.py

@@ -4,8 +4,6 @@
 # Yuta Nakahara <yuta.nakahara@aoni.waseda.jp>
 import warnings
 import numpy as np
-import os
-import sys
 from scipy.stats import beta as ss_beta
 # from scipy.stats import betabino as ss_betabinom
 import matplotlib.pyplot as plt
@@ -155,29 +153,33 @@ def visualize_model(self,sample_size=20,sample_num=5):
         >>> model = bernoulli.GenModel()
         >>> model.visualize_model()
         p:0.5
-        x0:[0 0 1 1 1 1 1 1 1 0 1 0 1 1 1 0 1 0 1 0]
-        x1:[1 0 0 1 1 0 1 0 0 0 0 0 0 0 0 1 0 1 0 1]
-        x2:[1 1 1 1 0 1 0 1 0 0 0 1 1 0 0 1 1 1 0 1]
-        x3:[0 0 1 1 1 0 0 1 1 0 0 1 0 0 1 0 0 1 0 1]
-        x4:[0 1 0 1 1 0 1 0 1 1 1 1 1 0 1 0 0 1 1 0]
-
+        x0:[1 1 0 0 0 1 0 1 0 0 0 1 0 1 0 1 0 1 0 0]
+        x1:[1 1 0 0 0 0 0 1 1 0 0 0 1 0 1 0 0 0 0 0]
+        x2:[0 1 0 1 0 0 1 0 0 0 1 0 1 1 1 0 1 0 1 1]
+        x3:[0 0 0 1 1 0 1 0 1 0 0 0 1 0 1 0 1 0 1 1]
+        x4:[1 0 1 1 1 1 0 1 0 0 1 1 0 0 0 0 0 0 1 1]
+        
         .. image:: ./images/bernoulli_example.png
         """
         _check.pos_int(sample_size,'sample_size',DataFormatError)
         _check.pos_int(sample_num,'sample_num',DataFormatError)
         print(f"p:{self.p}")
-        fig, ax = plt.subplots(figsize=(5,sample_num))
+        fig, ax = plt.subplots(2,1,figsize=(5, sample_num+1),gridspec_kw={'height_ratios': [1,sample_num]})
+        ax[0].set_title("True distribution")
+        ax[0].barh(0,self.p,label=1,color="C0")
+        ax[0].barh(0,1.0-self.p,left=self.p,label=0,color="C1")
+        ax[1].set_title("Generated sample")
         for i in range(sample_num):
             x = self.gen_sample(sample_size)
             print(f"x{i}:{x}")
             if i == 0:
-                ax.barh(i,x.sum(),label=1,color="C0")
-                ax.barh(i,sample_size-x.sum(),left=x.sum(),label=0,color="C1")
+                ax[1].barh(i,x.sum(),label=1,color="C0")
+                ax[1].barh(i,sample_size-x.sum(),left=x.sum(),label=0,color="C1")
             else:
-                ax.barh(i,x.sum(),color="C0")
-                ax.barh(i,sample_size-x.sum(),left=x.sum(),color="C1")
-        ax.legend()
-        ax.set_xlabel("Number of occurrences")
+                ax[1].barh(i,x.sum(),color="C0")
+                ax[1].barh(i,sample_size-x.sum(),left=x.sum(),color="C1")
+        ax[1].legend()
+        ax[1].set_xlabel("Number of occurrences")
         plt.show()
 
 class LearnModel(base.Posterior,base.PredictiveMixin):