add CNN support

README.md

@@ -46,10 +46,12 @@ for _ in range(1000):
 
 ## Demo
 * A naked and step-by-step [network](/simple_nn.py) without using my module.
-* [Train a regressor](/train_regressor.py)
-* [Train a classifier](/train_classifier.py)
+* [Train regressor](/train_regressor.py)
+* [Train classifier](/train_classifier.py)
+* [Train CNN](/train_cnn.py)
 * [Save and restore a trained net](/save_model.py)
 
+
 ## Download or fork
 Download [link](https://github.com/MorvanZhou/simple-neural-networks/archive/master.zip)
 

neuralnets/__init__.py

@@ -1,8 +1,10 @@
-from .nn import layers
-from .nn import initializers as init
-from .nn import activations as act
-from .nn.module import Module
-from .nn import losses
-from .nn import optimizers as optim
-from .nn.variable import Variable
-from .nn.saver import Saver
+from . import layers
+from . import initializers as init
+from . import activations as act
+from .module import Module
+from . import losses
+from . import optimizers as optim
+from .variable import Variable
+from .saver import Saver
+from .dataloader import DataLoader
+from . import metrics

neuralnets/dataloader.py

@@ -0,0 +1,28 @@
+import numpy as np
+
+
+class DataLoader:
+    def __init__(self, x, y, batch_size):
+        self.x = x
+        self.y = y
+        self.bs = batch_size
+        self.p = 0
+        self.bg = self.batch_generator()
+
+    def batch_generator(self):
+        while True:
+            p_ = self.p + self.bs
+            if p_ > len(self.x):
+                self.p = 0
+                continue
+            if self.p == 0:
+                indices = np.random.permutation(len(self.x))
+                self.x[:] = self.x[indices]
+                self.y[:] = self.y[indices]
+            bx = self.x[self.p:p_]
+            by = self.y[self.p:p_]
+            self.p = p_
+            yield bx, by
+
+    def next_batch(self):
+        return next(self.bg)