diff --git a/README.md b/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..dc7862137a278f71b3f5ac5fdedb70bf7044a00a
--- /dev/null
+++ b/README.md
@@ -0,0 +1,14 @@
+# Loss function custom
+
+## environment
+- miniforge
+- tensorflow 2.8.0
+- matplotlib
+
+## dataset
+
+- mnist dataset
+
+
+### Loss function
+- categorical cross entropy : 범주형 데이터에 대한 Loss function
\ No newline at end of file
diff --git a/loss_function copy.ipynb b/loss_function copy.ipynb
index d637d6814059942a3c7170734583898bb6c058c5..916400ba7553b176dd73ba1d93a6c22f0121f5eb 100644
--- a/loss_function copy.ipynb
+++ b/loss_function copy.ipynb
@@ -2,7 +2,7 @@
"cells": [
{
"cell_type": "code",
- "execution_count": 3,
+ "execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
@@ -12,31 +12,22 @@
"from keras.models import Sequential\n",
"from keras.layers import Dense, Dropout, Flatten\n",
"from keras.layers.convolutional import Conv2D, MaxPooling2D\n",
+ "from keras.utils import np_utils\n",
"import numpy as np"
]
},
{
"cell_type": "code",
- "execution_count": 4,
+ "execution_count": 3,
"metadata": {},
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
- "11493376/11490434 [==============================] - 2s 0us/step\n",
- "11501568/11490434 [==============================] - 2s 0us/step\n"
- ]
- }
- ],
+ "outputs": [],
"source": [
"(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()"
]
},
{
"cell_type": "code",
- "execution_count": 15,
+ "execution_count": 4,
"metadata": {},
"outputs": [
{
@@ -61,247 +52,132 @@
"\n",
"print('x_train shape:', x_train.shape)\n",
"print(x_train.shape[0], 'train samples')\n",
- "print(x_test.shape[0], 'test samples')\n",
- "\n"
+ "print(x_test.shape[0], 'test samples')\n"
]
},
{
"cell_type": "code",
- "execution_count": null,
+ "execution_count": 5,
"metadata": {},
"outputs": [],
- "source": [
- "def ccee(predict, label):\n",
- " delta = 1e-7\n",
- " log_pred = np.log(predict + delta)\n",
- "\n",
- " return -(np.sum(np.sum(label * log_pred, axis = 1)))/label.shape[0]"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 16,
- "metadata": {},
- "outputs": [
- {
- "ename": "AttributeError",
- "evalue": "module 'keras.utils' has no attribute 'to_categorical'",
- "output_type": "error",
- "traceback": [
- "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
- "\u001b[0;31mAttributeError\u001b[0m Traceback (most recent call last)",
- "\u001b[1;32m/Users/yongheewon/Documents/opensource/loss_function.ipynb Cell 4'\u001b[0m in \u001b[0;36m<cell line: 2>\u001b[0;34m()\u001b[0m\n\u001b[1;32m <a href='vscode-notebook-cell:/Users/yongheewon/Documents/opensource/loss_function.ipynb#ch0000006?line=0'>1</a>\u001b[0m num_classes \u001b[39m=\u001b[39m \u001b[39m10\u001b[39m\n\u001b[0;32m----> <a href='vscode-notebook-cell:/Users/yongheewon/Documents/opensource/loss_function.ipynb#ch0000006?line=1'>2</a>\u001b[0m y_train \u001b[39m=\u001b[39m keras\u001b[39m.\u001b[39;49mutils\u001b[39m.\u001b[39;49mto_categorical(y_train, num_classes)\n\u001b[1;32m <a href='vscode-notebook-cell:/Users/yongheewon/Documents/opensource/loss_function.ipynb#ch0000006?line=2'>3</a>\u001b[0m y_test \u001b[39m=\u001b[39m keras\u001b[39m.\u001b[39mutils\u001b[39m.\u001b[39mto_categorical(y_test, num_classes)\n",
- "\u001b[0;31mAttributeError\u001b[0m: module 'keras.utils' has no attribute 'to_categorical'"
- ]
- }
- ],
"source": [
"num_classes = 10\n",
- "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
- "y_test = keras.utils.to_categorical(y_test, num_classes)"
+ "y_train = keras.utils.np_utils.to_categorical(y_train, num_classes)\n",
+ "y_test = keras.utils.np_utils.to_categorical(y_test, num_classes)"
]
},
{
"cell_type": "code",
- "execution_count": 6,
+ "execution_count": 7,
"metadata": {},
"outputs": [
{
- "ename": "ModuleNotFoundError",
- "evalue": "No module named 'matplotlib'",
- "output_type": "error",
- "traceback": [
- "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
- "\u001b[0;31mModuleNotFoundError\u001b[0m Traceback (most recent call last)",
- "\u001b[1;32m/Users/yongheewon/Documents/opensource/loss_function.ipynb Cell 4'\u001b[0m in \u001b[0;36m<cell line: 2>\u001b[0;34m()\u001b[0m\n\u001b[1;32m <a href='vscode-notebook-cell:/Users/yongheewon/Documents/opensource/loss_function.ipynb#ch0000003?line=0'>1</a>\u001b[0m \u001b[39mimport\u001b[39;00m \u001b[39mrandom\u001b[39;00m\n\u001b[0;32m----> <a href='vscode-notebook-cell:/Users/yongheewon/Documents/opensource/loss_function.ipynb#ch0000003?line=1'>2</a>\u001b[0m \u001b[39mimport\u001b[39;00m \u001b[39mmatplotlib\u001b[39;00m\u001b[39m.\u001b[39;00m\u001b[39mpyplot\u001b[39;00m \u001b[39mas\u001b[39;00m \u001b[39mplt\u001b[39;00m\n\u001b[1;32m <a href='vscode-notebook-cell:/Users/yongheewon/Documents/opensource/loss_function.ipynb#ch0000003?line=3'>4</a>\u001b[0m predicted_result \u001b[39m=\u001b[39m model\u001b[39m.\u001b[39mpredict(x_test)\n\u001b[1;32m <a href='vscode-notebook-cell:/Users/yongheewon/Documents/opensource/loss_function.ipynb#ch0000003?line=4'>5</a>\u001b[0m predicted_labels \u001b[39m=\u001b[39m np\u001b[39m.\u001b[39margmax(predicted_result, axis\u001b[39m=\u001b[39m\u001b[39m1\u001b[39m)\n",
- "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'matplotlib'"
- ]
+ "data": {
+ "text/plain": [
+ "array([0., 0., 0., 0., 0., 1., 0., 0., 0., 0.], dtype=float32)"
+ ]
+ },
+ "execution_count": 7,
+ "metadata": {},
+ "output_type": "execute_result"
}
],
"source": [
- "import random\n",
- "import matplotlib.pyplot as plt\n",
- "\n",
- "predicted_result = model.predict(x_test)\n",
- "predicted_labels = np.argmax(predicted_result, axis=1)\n",
- "\n",
- "test_labels = np.argmax(y_test, axis=1)\n",
- "\n",
- "count = 0\n",
- "\n",
- "plt.figure(figsize=(12,8))\n",
- "for n in range(16):\n",
- " count += 1\n",
- " plt.subplot(4, 4, count)\n",
- " plt.imshow(x_test[n].reshape(28, 28), cmap='Greys', interpolation='nearest')\n",
- " tmp = \"Label:\" + str(test_labels[n]) + \", Prediction:\" + str(predicted_labels[n])\n",
- " plt.title(tmp)\n",
- "\n",
- "plt.tight_layout()\n",
- "plt.show()"
+ "y_train[0]"
]
},
{
"cell_type": "code",
- "execution_count": 13,
+ "execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
- "Model: \"as3s6g-confscore-center\"\n",
- "_________________________________________________________________\n",
- " Layer (type) Output Shape Param # \n",
- "=================================================================\n",
- " input_1 (InputLayer) [(None, 75, 3)] 0 \n",
- " \n",
- " c1 (Conv1D) (None, 75, 64) 1024 \n",
- " \n",
- " c2 (Conv1D) (None, 75, 64) 20544 \n",
- " \n",
- " conv1d (Conv1D) (None, 75, 64) 20544 \n",
- " \n",
- " c3 (Conv1D) (None, 75, 128) 41088 \n",
- " \n",
- " c4 (Conv1D) (None, 75, 128) 82048 \n",
- " \n",
- " c6 (Conv1D) (None, 75, 256) 164096 \n",
- " \n",
- " c7 (Conv1D) (None, 75, 256) 327936 \n",
- " \n",
- " c9 (Conv1D) (None, 75, 512) 655872 \n",
- " \n",
- " c10 (Conv1D) (None, 75, 512) 1311232 \n",
- " \n",
- " conv1d_1 (Conv1D) (None, 75, 256) 655616 \n",
- " \n",
- " conv1d_2 (Conv1D) (None, 75, 256) 65792 \n",
- " \n",
- " conv1d_3 (Conv1D) (None, 75, 128) 32896 \n",
- " \n",
- " conv1d_4 (Conv1D) (None, 75, 128) 16512 \n",
- " \n",
- " flatten (Flatten) (None, 9600) 0 \n",
- " \n",
- " fc3 (Dense) (None, 128) 1228928 \n",
- " \n",
- " dropout (Dropout) (None, 128) 0 \n",
- " \n",
- " fc4 (Dense) (None, 64) 8256 \n",
- " \n",
- " dropout_1 (Dropout) (None, 64) 0 \n",
- " \n",
- " dense (Dense) (None, 32) 2080 \n",
- " \n",
- " dropout_2 (Dropout) (None, 32) 0 \n",
- " \n",
- " dense_1 (Dense) (None, 16) 528 \n",
- " \n",
- " dropout_3 (Dropout) (None, 16) 0 \n",
- " \n",
- " fc5 (Dense) (None, 12) 204 \n",
- " \n",
- " reshape (Reshape) (None, 6, 2) 0 \n",
- " \n",
- "=================================================================\n",
- "Total params: 4,635,196\n",
- "Trainable params: 4,635,196\n",
- "Non-trainable params: 0\n",
- "_________________________________________________________________\n",
- "Model: \"as3s6g-confscore-center\"\n",
+ "28 28\n",
+ "Model: \"image_class\"\n",
"_________________________________________________________________\n",
" Layer (type) Output Shape Param # \n",
"=================================================================\n",
- " input_1 (InputLayer) [(None, 75, 3)] 0 \n",
- " \n",
- " c1 (Conv1D) (None, 75, 64) 1024 \n",
- " \n",
- " c2 (Conv1D) (None, 75, 64) 20544 \n",
- " \n",
- " conv1d (Conv1D) (None, 75, 64) 20544 \n",
- " \n",
- " c3 (Conv1D) (None, 75, 128) 41088 \n",
- " \n",
- " c4 (Conv1D) (None, 75, 128) 82048 \n",
+ " input_1 (InputLayer) [(None, 28, 28, 1)] 0 \n",
" \n",
- " c6 (Conv1D) (None, 75, 256) 164096 \n",
+ " conv2d (Conv2D) (None, 28, 28, 32) 832 \n",
" \n",
- " c7 (Conv1D) (None, 75, 256) 327936 \n",
+ " max_pooling2d (MaxPooling2D (None, 14, 14, 32) 0 \n",
+ " ) \n",
" \n",
- " c9 (Conv1D) (None, 75, 512) 655872 \n",
+ " conv2d_1 (Conv2D) (None, 14, 14, 64) 8256 \n",
" \n",
- " c10 (Conv1D) (None, 75, 512) 1311232 \n",
+ " max_pooling2d_1 (MaxPooling (None, 7, 7, 64) 0 \n",
+ " 2D) \n",
" \n",
- " conv1d_1 (Conv1D) (None, 75, 256) 655616 \n",
+ " dropout (Dropout) (None, 7, 7, 64) 0 \n",
" \n",
- " conv1d_2 (Conv1D) (None, 75, 256) 65792 \n",
+ " flatten (Flatten) (None, 3136) 0 \n",
" \n",
- " conv1d_3 (Conv1D) (None, 75, 128) 32896 \n",
+ " dense (Dense) (None, 1000) 3137000 \n",
" \n",
- " conv1d_4 (Conv1D) (None, 75, 128) 16512 \n",
+ " dropout_1 (Dropout) (None, 1000) 0 \n",
" \n",
- " flatten (Flatten) (None, 9600) 0 \n",
- " \n",
- " fc3 (Dense) (None, 128) 1228928 \n",
- " \n",
- " dropout (Dropout) (None, 128) 0 \n",
- " \n",
- " fc4 (Dense) (None, 64) 8256 \n",
- " \n",
- " dropout_1 (Dropout) (None, 64) 0 \n",
- " \n",
- " dense (Dense) (None, 32) 2080 \n",
- " \n",
- " dropout_2 (Dropout) (None, 32) 0 \n",
- " \n",
- " dense_1 (Dense) (None, 16) 528 \n",
- " \n",
- " dropout_3 (Dropout) (None, 16) 0 \n",
- " \n",
- " fc5 (Dense) (None, 12) 204 \n",
- " \n",
- " reshape (Reshape) (None, 6, 2) 0 \n",
+ " dense_1 (Dense) (None, 10) 10010 \n",
" \n",
"=================================================================\n",
- "Total params: 4,635,196\n",
- "Trainable params: 4,635,196\n",
+ "Total params: 3,156,098\n",
+ "Trainable params: 3,156,098\n",
"Non-trainable params: 0\n",
"_________________________________________________________________\n"
]
}
],
"source": [
- "def create_model( img_rows,img_cols, num_classes):\n",
- " inputs = tf.keras.Input(shape=(img_rows, img_cols, 1))\n",
- " x = tf.keras.layers.Conv2D(32, kernel_size = (5,5), name='c1', padding='same',\n",
+ "def create_model_( img_rows,img_cols, num_classes):\n",
+ " inputs = tf.keras.Input(shape=(28, 28, 1)) \n",
+ " x = tf.keras.layers.Conv2D(32, kernel_size = (5,5), padding='same',\n",
" activation='relu')(inputs)\n",
" x = tf.keras.layers.MaxPool2D(pool_size = (2,2),strides = (2,2))(x)\n",
" \n",
- " x = tf.keras.layers.Conv2D(64, kernel_size = (2,2), name='c1', padding='same',\n",
+ " x = tf.keras.layers.Conv2D(64, kernel_size = (2,2),padding='same',\n",
" activation='relu')(x)\n",
" x = tf.keras.layers.MaxPool2D(pool_size = (2,2))(x)\n",
" \n",
" x = tf.keras.layers.Dropout(0.25)(x)\n",
- " x = tf.keras.layers.Flatten(name='flatten')(x) \n",
- " x = tf.keras.layers.Dense(1000, name='fc3', activation='leaky_relu')(x)\n",
+ " x = tf.keras.layers.Flatten()(x)\n",
+ " x = tf.keras.layers.Dense(1000, activation='leaky_relu')(x)\n",
" x = tf.keras.layers.Dropout(0.5)(x)\n",
- " x = tf.keras.layers.Dense(num_classes, activation='softmax')(x)\n",
- " model.summary()\n",
+ " outputs = tf.keras.layers.Dense(num_classes, activation='softmax')(x)\n",
+ " \n",
+ " model = keras.Model(inputs, outputs, name='image_class')\n",
" return model\n",
"\n",
"\n",
- "\n",
- "model = create_model(img_cols, img_rows, num_classes)\n",
+ "print(img_cols, img_rows)\n",
+ "model = create_model_(img_rows,img_cols, num_classes)\n",
"model.summary()"
]
},
+ {
+ "cell_type": "code",
+ "execution_count": 9,
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def ccee(y_true, y_pred):\n",
+ " delta = 1e-7\n",
+ " log_pred = tf.math.log(y_pred + delta)\n",
+ "\n",
+ " return -(tf.reduce_sum(tf.reduce_sum(y_true * log_pred, axis = 1)))/y_true.shape[0]"
+ ]
+ },
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
- "source": []
+ "source": [
+ "model.compile( loss = ccee,optimizer='adam', metrics= ['accuracy'] )\n",
+ "hist = model.fit(x_train, y_train, batch_size=10, epochs= 10, verbose=1, validation_data=(x_test, y_test))"
+ ]
}
],
"metadata": {
diff --git a/loss_function.ipynb b/loss_function.ipynb
index 58d9b54f75fbc11b498e0f0c4ccbc562d8db7d7a..18456daadf40d06585ac88b414ddd0062ceca3fb 100644
--- a/loss_function.ipynb
+++ b/loss_function.ipynb
@@ -18,10 +18,10 @@
},
{
"cell_type": "markdown",
- "source": [],
"metadata": {
"collapsed": false
- }
+ },
+ "source": []
},
{
"cell_type": "code",
@@ -159,14 +159,14 @@
{
"cell_type": "code",
"execution_count": null,
- "outputs": [],
- "source": [],
"metadata": {
"collapsed": false,
"pycharm": {
"name": "#%%\n"
}
- }
+ },
+ "outputs": [],
+ "source": []
},
{
"cell_type": "code",
@@ -177,40 +177,39 @@
"name": "stdout",
"output_type": "stream",
"text": [
- "Epoch 1/10\n",
- " 1/6000 [..............................] - ETA: 17:42 - loss: 2.2814 - accuracy: 0.2000"
+ "Epoch 1/10\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
- "2022-06-26 11:32:02.208949: W tensorflow/core/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz\n"
+ "2022-06-26 12:24:22.787706: W tensorflow/core/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
- "6000/6000 [==============================] - 96s 16ms/step - loss: 0.1236 - accuracy: 0.9616 - val_loss: 0.0493 - val_accuracy: 0.9853\n",
+ "6000/6000 [==============================] - 68s 11ms/step - loss: 0.1259 - accuracy: 0.9608 - val_loss: 0.0342 - val_accuracy: 0.9888\n",
"Epoch 2/10\n",
- "6000/6000 [==============================] - 150s 25ms/step - loss: 0.0680 - accuracy: 0.9796 - val_loss: 0.0501 - val_accuracy: 0.9845\n",
+ "6000/6000 [==============================] - 77s 13ms/step - loss: 0.0657 - accuracy: 0.9800 - val_loss: 0.0442 - val_accuracy: 0.9865\n",
"Epoch 3/10\n",
- "6000/6000 [==============================] - 186s 31ms/step - loss: 0.0557 - accuracy: 0.9835 - val_loss: 0.0439 - val_accuracy: 0.9856\n",
+ "6000/6000 [==============================] - 81s 13ms/step - loss: 0.0554 - accuracy: 0.9827 - val_loss: 0.0312 - val_accuracy: 0.9902\n",
"Epoch 4/10\n",
- "6000/6000 [==============================] - 206s 34ms/step - loss: 0.0517 - accuracy: 0.9848 - val_loss: 0.0337 - val_accuracy: 0.9890\n",
+ "6000/6000 [==============================] - 91s 15ms/step - loss: 0.0503 - accuracy: 0.9850 - val_loss: 0.0385 - val_accuracy: 0.9885\n",
"Epoch 5/10\n",
- "6000/6000 [==============================] - 180s 30ms/step - loss: 0.0480 - accuracy: 0.9861 - val_loss: 0.0295 - val_accuracy: 0.9912\n",
+ "6000/6000 [==============================] - 110s 18ms/step - loss: 0.0481 - accuracy: 0.9862 - val_loss: 0.0359 - val_accuracy: 0.9887\n",
"Epoch 6/10\n",
- "6000/6000 [==============================] - 159s 26ms/step - loss: 0.0436 - accuracy: 0.9869 - val_loss: 0.0415 - val_accuracy: 0.9880\n",
+ "6000/6000 [==============================] - 110s 18ms/step - loss: 0.0481 - accuracy: 0.9860 - val_loss: 0.0637 - val_accuracy: 0.9829\n",
"Epoch 7/10\n",
- "6000/6000 [==============================] - 162s 27ms/step - loss: 0.0492 - accuracy: 0.9863 - val_loss: 0.0433 - val_accuracy: 0.9876\n",
+ "6000/6000 [==============================] - 108s 18ms/step - loss: 0.0444 - accuracy: 0.9873 - val_loss: 0.0417 - val_accuracy: 0.9903\n",
"Epoch 8/10\n",
- "6000/6000 [==============================] - 139s 23ms/step - loss: 0.0454 - accuracy: 0.9875 - val_loss: 0.0372 - val_accuracy: 0.9901\n",
+ "6000/6000 [==============================] - 116s 19ms/step - loss: 0.0432 - accuracy: 0.9875 - val_loss: 0.0325 - val_accuracy: 0.9906\n",
"Epoch 9/10\n",
- "6000/6000 [==============================] - 138s 23ms/step - loss: 0.0424 - accuracy: 0.9883 - val_loss: 0.0314 - val_accuracy: 0.9905\n",
+ "6000/6000 [==============================] - 103s 17ms/step - loss: 0.0430 - accuracy: 0.9886 - val_loss: 0.0329 - val_accuracy: 0.9898\n",
"Epoch 10/10\n",
- "6000/6000 [==============================] - 133s 22ms/step - loss: 0.0421 - accuracy: 0.9885 - val_loss: 0.0401 - val_accuracy: 0.9893\n"
+ "6000/6000 [==============================] - 97s 16ms/step - loss: 0.0448 - accuracy: 0.9884 - val_loss: 0.0318 - val_accuracy: 0.9902\n"
]
}
],
@@ -221,22 +220,33 @@
},
{
"cell_type": "code",
- "execution_count": 8,
+ "execution_count": 13,
"metadata": {},
"outputs": [
{
"data": {
+ "image/png": "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",
"text/plain": [
- "<keras.callbacks.History at 0x16bb95e80>"
+ "<Figure size 432x288 with 1 Axes>"
]
},
- "execution_count": 8,
- "metadata": {},
- "output_type": "execute_result"
+ "metadata": {
+ "needs_background": "light"
+ },
+ "output_type": "display_data"
}
],
"source": [
- "hist"
+ "from matplotlib import pyplot as plt\n",
+ "\n",
+ "\n",
+ "plt.plot(hist.history['accuracy'])\n",
+ "plt.plot(hist.history['val_accuracy'])\n",
+ "plt.title('model accuracy')\n",
+ "plt.ylabel('accuracy')\n",
+ "plt.xlabel('epoch')\n",
+ "plt.legend(['train', 'val'], loc='upper left')\n",
+ "plt.show()"
]
},
{
@@ -286,4 +296,4 @@
},
"nbformat": 4,
"nbformat_minor": 2
-}
\ No newline at end of file
+}