7-2.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iNFOeMfl3tIu"
      },
      "source": [
        "# 심층 신경망"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zKfwb8gS3tI2"
      },
      "source": [
        "<table align=\"left\">\n",
        "  <td>\n",
        "    <a target=\"_blank\" href=\"https://colab.research.google.com/github/rickiepark/hg-mldl/blob/master/7-2.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />구글 코랩에서 실행하기</a>\n",
        "  </td>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "HL0T-SUU8uSA"
      },
      "outputs": [],
      "source": [
        "# 실행마다 동일한 결과를 얻기 위해 케라스에 랜덤 시드를 사용하고 텐서플로 연산을 결정적으로 만듭니다. \n",
        "import tensorflow as tf\n",
        "\n",
        "tf.keras.utils.set_random_seed(42)\n",
        "tf.config.experimental.enable_op_determinism()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dPE5XsFhcfVD"
      },
      "source": [
        "## 2개의 층"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4sNOMcdaFVKa",
        "outputId": "7c64cfad-8b5b-485c-c1f4-c84a89559d50"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
            "32768/29515 [=================================] - 0s 0us/step\n",
            "40960/29515 [=========================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n",
            "26427392/26421880 [==============================] - 0s 0us/step\n",
            "26435584/26421880 [==============================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
            "16384/5148 [===============================================================================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n",
            "4423680/4422102 [==============================] - 0s 0us/step\n",
            "4431872/4422102 [==============================] - 0s 0us/step\n"
          ]
        }
      ],
      "source": [
        "from tensorflow import keras\n",
        "\n",
        "(train_input, train_target), (test_input, test_target) = keras.datasets.fashion_mnist.load_data()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "aJJiRMa6FkWx"
      },
      "outputs": [],
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "train_scaled = train_input / 255.0\n",
        "train_scaled = train_scaled.reshape(-1, 28*28)\n",
        "\n",
        "train_scaled, val_scaled, train_target, val_target = train_test_split(\n",
        "    train_scaled, train_target, test_size=0.2, random_state=42)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "MCZWQiEwF6MD"
      },
      "outputs": [],
      "source": [
        "dense1 = keras.layers.Dense(100, activation='sigmoid', input_shape=(784,)) # 은닉층, 출력값 100개는 스스로 판단해서 써야 함\n",
        "dense2 = keras.layers.Dense(10, activation='softmax') # 출력층"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Agy5gCVhcrm-"
      },
      "source": [
        "## 심층 신경망 만들기"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "xmWL7kOoGB4P"
      },
      "outputs": [],
      "source": [
        "model = keras.Sequential([dense1, dense2])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "em0xDl8qa12J",
        "outputId": "1e86323b-4a75-4728-cc73-493861a87166"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense (Dense)               (None, 100)               78500     \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 10)                1010      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 79,510\n",
            "Trainable params: 79,510\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model.summary() # non-trainable params : 경사 하강법으로 훈련되지 않는 파라미터, total params : (784 * 100 + 100) + (100 * 10 + 10)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qAi41rBTdk7k"
      },
      "source": [
        "## 층을 추가하는 다른 방법"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "0XeV6V4ha3I8"
      },
      "outputs": [],
      "source": [
        "model = keras.Sequential([\n",
        "    keras.layers.Dense(100, activation='sigmoid', input_shape=(784,), name='hidden'),\n",
        "    keras.layers.Dense(10, activation='softmax', name='output')\n",
        "], name='패션 MNIST 모델') # 층 변수를 따로 사용할 일이 없으므로 모델 안에 직접 층을 넣음"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bwXDLSOWbm3L",
        "outputId": "1a1ab55b-fb7a-4016-a56c-b271d4ecae12"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"패션 MNIST 모델\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " hidden (Dense)              (None, 100)               78500     \n",
            "                                                                 \n",
            " output (Dense)              (None, 10)                1010      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 79,510\n",
            "Trainable params: 79,510\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "yZSAxgZCbax7"
      },
      "outputs": [],
      "source": [
        "model = keras.Sequential() # add 메서드로 층을 추가하는 것이 일반적\n",
        "model.add(keras.layers.Dense(100, activation='sigmoid', input_shape=(784,)))\n",
        "model.add(keras.layers.Dense(10, activation='softmax'))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bW2coaNQboe5",
        "outputId": "a0ea6052-4032-4202-d23d-f0068580dc18"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense_2 (Dense)             (None, 100)               78500     \n",
            "                                                                 \n",
            " dense_3 (Dense)             (None, 10)                1010      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 79,510\n",
            "Trainable params: 79,510\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kkYrPJembpYk",
        "outputId": "2fedbee2-83e8-4439-bd6a-1f2f689b6dc0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1500/1500 [==============================] - 9s 2ms/step - loss: 0.5628 - accuracy: 0.8073\n",
            "Epoch 2/5\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.4075 - accuracy: 0.8522\n",
            "Epoch 3/5\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.3741 - accuracy: 0.8652\n",
            "Epoch 4/5\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.3509 - accuracy: 0.8732\n",
            "Epoch 5/5\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.3335 - accuracy: 0.8784\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f65203036d0>"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ],
      "source": [
        "model.compile(loss='sparse_categorical_crossentropy', metrics='accuracy')\n",
        "\n",
        "model.fit(train_scaled, train_target, epochs=5)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_dfXJDhwcyAK"
      },
      "source": [
        "## 렐루 활성화 함수"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "dG7yF8g6b062"
      },
      "outputs": [],
      "source": [
        "model = keras.Sequential()\n",
        "model.add(keras.layers.Flatten(input_shape=(28, 28)))\n",
        "model.add(keras.layers.Dense(100, activation='relu')) # 시그모이드는 함수의 왼쪽 끝과 오른쪽 끝에서 값이 1과 0으로 수렴하므로 변화가 점점 느려짐\n",
        "model.add(keras.layers.Dense(10, activation='softmax')) # 출력층에는 다중 분류일 경우 softmax로 고정 사용"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zHogWhu6g90a",
        "outputId": "82a88cec-cb39-49c5-9036-113a64540ab0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " flatten (Flatten)           (None, 784)               0         \n",
            "                                                                 \n",
            " dense_4 (Dense)             (None, 100)               78500     \n",
            "                                                                 \n",
            " dense_5 (Dense)             (None, 10)                1010      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 79,510\n",
            "Trainable params: 79,510\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "id": "JfPe_ruQdhqA"
      },
      "outputs": [],
      "source": [
        "(train_input, train_target), (test_input, test_target) = keras.datasets.fashion_mnist.load_data()\n",
        "\n",
        "train_scaled = train_input / 255.0\n",
        "\n",
        "train_scaled, val_scaled, train_target, val_target = train_test_split(\n",
        "    train_scaled, train_target, test_size=0.2, random_state=42)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9PGejuuhdvvk",
        "outputId": "7aca280b-11eb-4840-f2da-b34499572963"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1500/1500 [==============================] - 6s 3ms/step - loss: 0.5338 - accuracy: 0.8089\n",
            "Epoch 2/5\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.3917 - accuracy: 0.8593\n",
            "Epoch 3/5\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.3540 - accuracy: 0.8729\n",
            "Epoch 4/5\n",
            "1500/1500 [==============================] - 4s 3ms/step - loss: 0.3321 - accuracy: 0.8814\n",
            "Epoch 5/5\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.3160 - accuracy: 0.8872\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f64b2279cd0>"
            ]
          },
          "metadata": {},
          "execution_count": 15
        }
      ],
      "source": [
        "model.compile(loss='sparse_categorical_crossentropy', metrics='accuracy')\n",
        "\n",
        "model.fit(train_scaled, train_target, epochs=5)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lVYLpnjeep4y",
        "outputId": "d8683914-c48f-4180-f8ba-5cf792d9d118"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "375/375 [==============================] - 1s 2ms/step - loss: 0.3663 - accuracy: 0.8717\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.36633625626564026, 0.871749997138977]"
            ]
          },
          "metadata": {},
          "execution_count": 16
        }
      ],
      "source": [
        "model.evaluate(val_scaled, val_target)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3YtLsmGAoavz"
      },
      "source": [
        "## 옵티마이저"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "NxVj04Haocwa"
      },
      "outputs": [],
      "source": [
        "model.compile(optimizer='sgd', loss='sparse_categorical_crossentropy', metrics='accuracy') # 기본 미니 배치 경사하강법을 사용"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "id": "1426O4G8Hpfi"
      },
      "outputs": [],
      "source": [
        "sgd = keras.optimizers.SGD() # 위 코드와 동일\n",
        "model.compile(optimizer=sgd, loss='sparse_categorical_crossentropy', metrics='accuracy')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "id": "Sh-HDiULlp18"
      },
      "outputs": [],
      "source": [
        "sgd = keras.optimizers.SGD(learning_rate=0.1) # 파라미터를 바꾸고 싶다면 직접 객체를 만들어 사용"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "metadata": {
        "id": "uF1XolBXsl3a"
      },
      "outputs": [],
      "source": [
        "sgd = keras.optimizers.SGD(momentum=0.9, nesterov=True) # 모멘텀 : 이전 그레이디언트를 가속도처럼 사용"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "id": "Hy2MENbL170j"
      },
      "outputs": [],
      "source": [
        "adagrad = keras.optimizers.Adagrad() # 모델이 최적점에 갈수록 학습률을 낮추는 적응적 학습률\n",
        "model.compile(optimizer=adagrad, loss='sparse_categorical_crossentropy', metrics='accuracy')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 22,
      "metadata": {
        "id": "KkpbSMXWtakr"
      },
      "outputs": [],
      "source": [
        "rmsprop = keras.optimizers.RMSprop() # 옵티마이저 매개변수의 기본값, 적응적 학습률 사용\n",
        "model.compile(optimizer=rmsprop, loss='sparse_categorical_crossentropy', metrics='accuracy')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "metadata": {
        "id": "Gdu0hQIAz4JW"
      },
      "outputs": [],
      "source": [
        "model = keras.Sequential() \n",
        "model.add(keras.layers.Flatten(input_shape=(28, 28))) # 2차원 이미지 데이터를 1차원 배열로 만들어주는 Flatten 층\n",
        "model.add(keras.layers.Dense(100, activation='relu'))\n",
        "model.add(keras.layers.Dense(10, activation='softmax'))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2tcxIfILoi1t",
        "outputId": "cbe2c75f-9c9b-4fc7-d51e-c0ce40fe3f44"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/5\n",
            "1500/1500 [==============================] - 4s 2ms/step - loss: 0.5227 - accuracy: 0.8167\n",
            "Epoch 2/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.3909 - accuracy: 0.8605\n",
            "Epoch 3/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.3537 - accuracy: 0.8707\n",
            "Epoch 4/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.3285 - accuracy: 0.8794\n",
            "Epoch 5/5\n",
            "1500/1500 [==============================] - 3s 2ms/step - loss: 0.3074 - accuracy: 0.8858\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f649c3cfd50>"
            ]
          },
          "metadata": {},
          "execution_count": 24
        }
      ],
      "source": [
        "model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics='accuracy') # 모멘텀과 적응적 학습률을 모두 사용\n",
        "\n",
        "model.fit(train_scaled, train_target, epochs=5)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8gxAWehsv9Gi",
        "outputId": "f61f20ce-fd10-42ad-98e1-85fad6e83d7d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "375/375 [==============================] - 1s 2ms/step - loss: 0.3612 - accuracy: 0.8688\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.36117008328437805, 0.8688333630561829]"
            ]
          },
          "metadata": {},
          "execution_count": 25
        }
      ],
      "source": [
        "model.evaluate(val_scaled, val_target) # 일반적으로 rmsprop보다 조금 나은 결과"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [],
      "name": "7-2 심층 신경망.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "default:Python",
      "language": "python",
      "name": "conda-env-default-py"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.9.10"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}