fixed

Author: ctgk

notebooks/ch01_Introduction.ipynb

@@ -1,8 +1,21 @@
 from prml import (
-    nn
+    clusterings,
+    features,
+    feature_extractions,
+    linear,
+    nn,
+    rv,
+    sampling
 )
 
 
 __all__ = [
-    "nn"
+    "clusterings",
+    "features",
+    "feature_extractions",
+    "kernel",
+    "linear",
+    "nn",
+    "rv",
+    "sampling"
 ]

notebooks/ch02_Probability_Distributions.ipynb

@@ -1,8 +1,10 @@
-from .pca import PCA
-from .bayesian_pca import BayesianPCA
+from prml.feature_extractions.autoencoder import Autoencoder
+from prml.feature_extractions.bayesian_pca import BayesianPCA
+from prml.feature_extractions.pca import PCA
 
 
 __all__ = [
-    "PCA",
-    "BayesianPCA"
+    "Autoencoder",
+    "BayesianPCA",
+    "PCA"
 ]

notebooks/ch03_Linear_Models_for_Regression.ipynb

@@ -0,0 +1,38 @@
+import numpy as np
+from prml import nn
+
+
+class Autoencoder(nn.Network):
+
+    def __init__(self, *args):
+        self.n_unit = len(args)
+        super().__init__()
+        for i in range(self.n_unit - 1):
+            self.parameter[f"w_encode{i}"] = nn.Parameter(np.random.randn(args[i], args[i + 1]))
+            self.parameter[f"b_encode{i}"] = nn.Parameter(np.zeros(args[i + 1]))
+            self.parameter[f"w_decode{i}"] = nn.Parameter(np.random.randn(args[i + 1], args[i]))
+            self.parameter[f"b_decode{i}"] = nn.Parameter(np.zeros(args[i]))
+
+    def transform(self, x):
+        h = x
+        for i in range(self.n_unit - 1):
+            h = nn.tanh(h @ self.parameter[f"w_encode{i}"] + self.parameter[f"b_encode{i}"])
+        return h.value
+
+    def forward(self, x):
+        h = x
+        for i in range(self.n_unit - 1):
+            h = nn.tanh(h @ self.parameter[f"w_encode{i}"] + self.parameter[f"b_encode{i}"])
+        for i in range(self.n_unit - 2, 0, -1):
+            h = nn.tanh(h @ self.parameter[f"w_decode{i}"] + self.parameter[f"b_decode{i}"])
+        x_ = h @ self.parameter["w_decode0"] + self.parameter["b_decode0"]
+        self.px = nn.random.Gaussian(x_, 1., data=x)
+
+    def fit(self, x, n_iter=100, learning_rate=1e-3):
+        optimizer = nn.optimizer.Adam(self.parameter, learning_rate)
+        for _ in range(n_iter):
+            self.clear()
+            self.forward(x)
+            log_likelihood = self.log_pdf()
+            log_likelihood.backward()
+            optimizer.update()

notebooks/ch04_Linear_Models_for_Classfication.ipynb

@@ -1,5 +1,5 @@
 import numpy as np
-from .pca import PCA
+from prml.feature_extractions.pca import PCA
 
 
 class BayesianPCA(PCA):