Error corrected

Author: Kyubyong

Neural_Network_Part1.ipynb

@@ -12,7 +12,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 1,
    "metadata": {
     "collapsed": true,
     "deletable": true,
@@ -298,14 +298,20 @@
   },
   {
    "cell_type": "markdown",
-   "metadata": {},
+   "metadata": {
+    "deletable": true,
+    "editable": true
+   },
    "source": [
     "# Fully Connected"
    ]
   },
   {
    "cell_type": "markdown",
-   "metadata": {},
+   "metadata": {
+    "deletable": true,
+    "editable": true
+   },
    "source": [
     "Q5. Apply a fully connected layer to x with 2 outputs and then an sigmoid function."
    ]
@@ -314,7 +320,9 @@
    "cell_type": "code",
    "execution_count": 3,
    "metadata": {
-    "collapsed": false
+    "collapsed": false,
+    "deletable": true,
+    "editable": true
    },
    "outputs": [
     {
@@ -685,7 +693,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": 2,
    "metadata": {
     "collapsed": false,
     "deletable": true,
@@ -696,13 +704,13 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "(4, 9, 9, 5)\n"
+      "(4, 11, 11, 5)\n"
      ]
     }
    ],
    "source": [
     "x = tf.random_uniform(shape=(4, 5, 5, 4), dtype=tf.float32)\n",
-    "filter = tf.get_variable(\"filter\", shape=..., dtype=tf.float32, \n",
+    "filter = tf.get_variable(\"filter\", shape=(3, 3, 5, 4), dtype=tf.float32, \n",
     "                    initializer=tf.random_uniform_initializer())\n",
     "shp = x.get_shape().as_list()\n",
     "output_shape = ...\n",

Neural_Network_Part1_Solutions.ipynb

@@ -12,7 +12,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 1,
    "metadata": {
     "collapsed": true,
     "deletable": true,
@@ -298,14 +298,20 @@
   },
   {
    "cell_type": "markdown",
-   "metadata": {},
+   "metadata": {
+    "deletable": true,
+    "editable": true
+   },
    "source": [
     "# Fully Connected"
    ]
   },
   {
    "cell_type": "markdown",
-   "metadata": {},
+   "metadata": {
+    "deletable": true,
+    "editable": true
+   },
    "source": [
     "Q5. Apply a fully connected layer to x with 2 outputs and then an sigmoid function."
    ]
@@ -314,7 +320,9 @@
    "cell_type": "code",
    "execution_count": 24,
    "metadata": {
-    "collapsed": false
+    "collapsed": false,
+    "deletable": true,
+    "editable": true
    },
    "outputs": [
     {
@@ -453,7 +461,10 @@
     "with tf.Session() as sess:\n",
     "    sess.run(init)\n",
     "    _out = sess.run(out)\n",
-    "    print(_out.shape)"
+    "    print(_out.shape)\n",
+    "# Do we really have to distinguish between these two functions? \n",
+    "# Unless you want to use stride of 2 or more,\n",
+    "# You can just use tf.nn.atrous_conv2d. For normal convolution, set rate 1."
    ]
   },
   {
@@ -678,7 +689,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 19,
+   "execution_count": 11,
    "metadata": {
     "collapsed": true,
     "deletable": true,
@@ -691,7 +702,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": 12,
    "metadata": {
     "collapsed": false,
     "deletable": true,
@@ -702,7 +713,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "(4, 9, 9, 5)\n"
+      "(4, 11, 11, 5)\n"
      ]
     }
    ],
@@ -711,8 +722,8 @@
     "filter = tf.get_variable(\"filter\", shape=(3, 3, 5, 4), dtype=tf.float32, \n",
     "                    initializer=tf.random_uniform_initializer())\n",
     "shp = x.get_shape().as_list()\n",
-    "output_shape = [shp[0], (shp[1]-1)*2+1, (shp[2]-1)*2+1, 5]\n",
-    "out = tf.nn.conv2d_transpose(x, filter, strides=[1, 2, 2, 1], output_shape=output_shape, padding=\"SAME\")\n",
+    "output_shape = [shp[0], (shp[1]-1)*2+3, (shp[2]-1)*2+3, 5]\n",
+    "out = tf.nn.conv2d_transpose(x, filter, strides=[1, 2, 2, 1], output_shape=output_shape, padding=\"VALID\")\n",
     "init = tf.global_variables_initializer()\n",
     "with tf.Session() as sess:\n",
     "    sess.run(init)\n",