{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "##### 교재 페이지 (172-175) #####" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "13) Softmax Cross entropy with logits - 실습" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [], "source": [ "import numpy as np\n", "import tensorflow as tf" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "(101, 16) (101, 1)\n", "one_hot: Tensor(\"one_hot:0\", shape=(?, 1, 7), dtype=float32)\n", "reshape one_hot: Tensor(\"Reshape:0\", shape=(?, 7), dtype=float32)\n", "WARNING:tensorflow:From C:\\Users\\jsdata00010\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\framework\\op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n", "Instructions for updating:\n", "Colocations handled automatically by placer.\n", "Step: 0\tCost: 10.559\tAcc: 4.95%\n", "Step: 100\tCost: 0.638\tAcc: 85.15%\n", "Step: 200\tCost: 0.357\tAcc: 89.11%\n", "Step: 300\tCost: 0.247\tAcc: 95.05%\n", "Step: 400\tCost: 0.192\tAcc: 96.04%\n", "Step: 500\tCost: 0.158\tAcc: 97.03%\n", "Step: 600\tCost: 0.135\tAcc: 98.02%\n", "Step: 700\tCost: 0.118\tAcc: 98.02%\n", "Step: 800\tCost: 0.105\tAcc: 100.00%\n", "Step: 900\tCost: 0.095\tAcc: 100.00%\n", "Step: 1000\tCost: 0.087\tAcc: 100.00%\n", "Step: 1100\tCost: 0.080\tAcc: 100.00%\n", "Step: 1200\tCost: 0.074\tAcc: 100.00%\n", "Step: 1300\tCost: 0.069\tAcc: 100.00%\n", "Step: 1400\tCost: 0.065\tAcc: 100.00%\n", "Step: 1500\tCost: 0.061\tAcc: 100.00%\n", "Step: 1600\tCost: 0.058\tAcc: 100.00%\n", "Step: 1700\tCost: 0.055\tAcc: 100.00%\n", "Step: 1800\tCost: 0.052\tAcc: 100.00%\n", "Step: 1900\tCost: 0.050\tAcc: 100.00%\n", "Step: 2000\tCost: 0.047\tAcc: 100.00%\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 4 Real Y: 4\n", "[True] Prediction: 4 Real Y: 4\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 4 Real Y: 4\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 2 Real Y: 2\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 2 Real Y: 2\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 2 Real Y: 2\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 4 Real Y: 4\n", "[True] Prediction: 2 Real Y: 2\n", "[True] Prediction: 2 Real Y: 2\n", "[True] Prediction: 3 Real Y: 3\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 1 Real Y: 1\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 5 Real Y: 5\n", "[True] Prediction: 0 Real Y: 0\n", "[True] Prediction: 6 Real Y: 6\n", "[True] Prediction: 1 Real Y: 1\n" ] } ], "source": [ "#파워포인트 37페이지(softmax)\n", "# Predicting animal type based on various features\n", "xy = np.loadtxt('zoo1.csv', delimiter=',')\n", "x_data = xy[:101, :-1]\n", "y_data = xy[:101, [-1]]\n", "\n", "# Make sure the shape and data are OK\n", "print(x_data.shape, y_data.shape)\n", "nb_class = 7\n", "\n", "X = tf.placeholder(tf.float32, [None, 16])\n", "Y = tf.placeholder(tf.int32, [None, 1])\n", "\n", "Y_one_hot = tf.one_hot(Y, nb_class)\n", "print(\"one_hot:\", Y_one_hot)\n", "Y_one_hot = tf.reshape(Y_one_hot, [-1, nb_class])\n", "print(\"reshape one_hot:\", Y_one_hot)\n", "\n", "\n", "W = tf.Variable(tf.random_normal([16, nb_class]), name='weight')\n", "b = tf.Variable(tf.random_normal([nb_class]), name='bias')\n", "# tf.nn.softmax computes softmax activations\n", "# softmax = exp(logits) / reduce_sum(exp(logits), dim)\n", "logits = tf.matmul(X, W) + b\n", "hypothesis = tf.nn.softmax(logits)\n", "# Cross entropy cost/loss\n", "cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(logits=logits,\n", " labels=tf.stop_gradient([Y_one_hot])))\n", "optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(cost)\n", "\n", "prediction = tf.argmax(hypothesis, 1)\n", "correct_prediction = tf.equal(prediction, tf.argmax(Y_one_hot, 1))\n", "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n", "\n", "\n", "# Launch graph\n", "with tf.Session() as sess:\n", " sess.run(tf.global_variables_initializer())\n", " for step in range(2001):\n", " _, cost_val, acc_val = sess.run([optimizer, cost, accuracy], feed_dict={X: x_data, Y: y_data})\n", " if step % 100 == 0:\n", " print(\"Step: {:5}\\tCost: {:.3f}\\tAcc: {:.2%}\".format(step, cost_val, acc_val))\n", "# Let's see if we can predict\n", " pred = sess.run(prediction, feed_dict={X: x_data})\n", "# y_data: (N,1) = flatten => (N, ) matches pred.shape\n", " for p, y in zip(pred, y_data.flatten()):\n", " print(\"[{}] Prediction: {} Real Y: {}\".\n", " format(p == int(y), p, int(y)))\n", " " ] } ], "metadata": { "kernelspec": { "display_name": "Python 3", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.3" } }, "nbformat": 4, "nbformat_minor": 2 }