material/deep_learning/template_multi_numeric_sequence_data_classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "template_multi_numeric_sequence_data_classification.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mZAFnuGfwqy-"
      },
      "source": [
        "# 다중 숫자 순차열 데이터 분류 Template\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Hb0EOx6OyNIw"
      },
      "source": [
        "# 데이터 준비"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VBfCpbwh0h9R"
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n"
      ],
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "B5AtaEi6yPhX"
      },
      "source": [
        "## 기반 숫자열 생성"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V6wKtOwFwqXM"
      },
      "source": [
        "numbers = np.zeros((2000,2))\n",
        "numbers[:,0] = np.sin(np.arange(0,1000,0.5))\n",
        "numbers[:,1] = np.cos(np.arange(0,1000,0.5))"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "2ixw1zJVS69z",
        "outputId": "2d709203-b58f-4ba6-f6e6-c92b32edbc7d"
      },
      "source": [
        "plt.plot(numbers[:100,0])\n",
        "plt.plot(numbers[:100,1])\n",
        "plt.show()"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6RAkHiJbxNfE"
      },
      "source": [
        "SEQUENCE_LENGTH = 5\n",
        "OFFSET = 1"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "f-z-YNfNDBn5"
      },
      "source": [
        "## 입출력 데이터 생성\n",
        "\n",
        "[12345678901234567890, 34567890123456789012]의 숫자열 2개로 다음의 입출력 데이터를 생성한다.\n",
        "```\n",
        "[12345, 34567] -> T\n",
        "[23456, 45678] -> F\n",
        "[34567, 56789] -> T\n",
        "...\n",
        "[56789, 78901] -> F\n",
        "```"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yYxpmUQtwqTY"
      },
      "source": [
        "raw_x = []\n",
        "raw_y = []\n",
        "\n",
        "for i in range(len(numbers)):\n",
        "  if i+SEQUENCE_LENGTH+OFFSET >= len(numbers): break\n",
        "  raw_x.append(numbers[i:i+SEQUENCE_LENGTH,:])\n",
        "  next_value0 = numbers[i+SEQUENCE_LENGTH+OFFSET-1,0]\n",
        "  next_value1 = numbers[i+SEQUENCE_LENGTH+OFFSET-1,1]\n",
        "  raw_y.append([ int((next_value0+next_value1)>0) ])\n",
        "#              0         1              \n",
        "#              01234567890123456789\n",
        "# numbers[0] = 12345678901234567890\n",
        "#                 i = 3\n",
        "#                 <-x0>   = [3:8] = 45678\n",
        "#                      v0 = [8]   = 9\n",
        "# numbers[1] = 34567890123456789012\n",
        "#                 i = 3\n",
        "#                 <-x1>   = [3:8] = 67890\n",
        "#                      v1 = [9]   = 1\n",
        "#                      y = (v1+v2)>0 = (9+1)>0 = True -> 1\n",
        "\n"
      ],
      "execution_count": 36,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "15RBiNi9wqQU",
        "outputId": "558396cf-3497-4fae-e88e-b2d24644f206"
      },
      "source": [
        "print(len(raw_x))\n",
        "# len(raw_x) = 1994 = len(numbers) - SEQUENCE_LENGTH - OFFSET = 2000 - 5 - 1"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1994\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S7eFX6jIyrnx"
      },
      "source": [
        "## 데이터를 2차원 벡터화 하기\n",
        "\n",
        "RNN은 2차원 입력을 전제로 한다. 다중 숫자열은 그 자체로 2차원 데이터이다. 따로 해줄 것은 없다.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BagxNXc8MTuU",
        "outputId": "b42aa6d9-c8ff-47f6-ceb5-c3cc8a177815"
      },
      "source": [
        "print(raw_x[0])\n",
        "print(raw_y[0])"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[ 0.          1.        ]\n",
            " [ 0.47942554  0.87758256]\n",
            " [ 0.84147098  0.54030231]\n",
            " [ 0.99749499  0.0707372 ]\n",
            " [ 0.90929743 -0.41614684]]\n",
            "[0]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3D4HvrAwVzCI"
      },
      "source": [
        ""
      ],
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4mg54AkZ3boa",
        "outputId": "a6f7978b-7418-4bd1-ac99-9e8a8106ff4b"
      },
      "source": [
        "x = np.array(raw_x)\n",
        "y = np.array(raw_y)\n",
        "print(x.shape)\n",
        "print(y.shape)\n",
        "\n",
        "split_at = int(len(x)*0.9)\n",
        "(train_x, test_x) = x[:split_at], x[split_at:]\n",
        "(train_y, test_y) = y[:split_at], y[split_at:]\n",
        "\n",
        "print(\"Training Data:\")\n",
        "print(train_x.shape)\n",
        "print(train_y.shape)\n",
        "\n",
        "print(\"Test Data:\")\n",
        "print(test_x.shape)\n",
        "print(test_y.shape)\n"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1994, 5, 2)\n",
            "(1994, 1)\n",
            "Training Data:\n",
            "(1794, 5, 2)\n",
            "(1794, 1)\n",
            "Test Data:\n",
            "(200, 5, 2)\n",
            "(200, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lMq2Om3w8qNy"
      },
      "source": [
        "준비된 데이터는 다음과 같습니다."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "41UcKyn-q5Ez",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 427
        },
        "outputId": "0c425ebe-36e6-4256-8aca-f6f9f507948f"
      },
      "source": [
        "print(\"train_x[0]\\n\", train_x[0])\n",
        "print()\n",
        "print(\"train_y[0]\\n\", train_y[0])\n",
        "\n",
        "plt.plot(train_x[0].T[0], '.')\n",
        "plt.plot(train_x[0].T[1], '.')\n",
        "plt.plot([len(train_x[0])], train_y[0], 'x')\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "train_x[0]\n",
            " [[ 0.          1.        ]\n",
            " [ 0.47942554  0.87758256]\n",
            " [ 0.84147098  0.54030231]\n",
            " [ 0.99749499  0.0707372 ]\n",
            " [ 0.90929743 -0.41614684]]\n",
            "\n",
            "train_y[0]\n",
            " [0]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GeG-X2N3WmZz"
      },
      "source": [
        ""
      ],
      "execution_count": 40,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0-mHtIAk_qU9"
      },
      "source": [
        "# Template"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k3CG02lDAI73"
      },
      "source": [
        "## LSTM"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 819
        },
        "id": "GEffW4RR3bgJ",
        "outputId": "8db3aa95-f1e8-4947-e083-e054fa2fecf9"
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Dense, RepeatVector, Flatten\n",
        "from tensorflow.keras.layers import Bidirectional, LSTM, GRU\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "# input_shape = (5,2)\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(LSTM(128, input_shape=input_shape))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(2))\n",
        "model.compile(loss=\"sparse_categorical_crossentropy\", optimizer=\"adam\", metrics=[\"acc\"])\n",
        "model.summary()\n",
        "\n",
        "history = model.fit(train_x, train_y, epochs=10, verbose=0, validation_split=0.1, batch_size=128)\n",
        "\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.show()\n",
        "\n",
        "\n",
        "\n",
        "loss, acc = model.evaluate(test_x, test_y)\n",
        "print(\"loss=\", loss)\n",
        "print(\"acc=\", acc)\n",
        "\n",
        "\n",
        "y_ = model.predict(test_x)\n",
        "predicted = np.argmax(y_, axis=-1)\n",
        "plt.plot(test_y, 'o')\n",
        "plt.plot(predicted, ',')\n",
        "plt.show()"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_15\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "lstm_4 (LSTM)                (None, 128)               67072     \n",
            "_________________________________________________________________\n",
            "flatten_11 (Flatten)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_11 (Dense)             (None, 2)                 258       \n",
            "=================================================================\n",
            "Total params: 67,330\n",
            "Trainable params: 67,330\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "7/7 [==============================] - 0s 3ms/step - loss: 0.1906 - acc: 0.9250\n",
            "loss= 0.19062037765979767\n",
            "acc= 0.925000011920929\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "M1sV7tBtAWEz"
      },
      "source": [
        "## GRU"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 819
        },
        "id": "usltj8hI3bdL",
        "outputId": "b22c976f-c5e2-494b-b02d-7a22e3cdf4a8"
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Dense, RepeatVector, Flatten\n",
        "from tensorflow.keras.layers import Bidirectional, LSTM, GRU\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "# input_shape = (5,2)\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(GRU(128, input_shape=input_shape))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(2))\n",
        "model.compile(loss=\"sparse_categorical_crossentropy\", optimizer=\"adam\", metrics=[\"acc\"])\n",
        "model.summary()\n",
        "\n",
        "history = model.fit(train_x, train_y, epochs=10, verbose=0, validation_split=0.1, batch_size=128)\n",
        "\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.show()\n",
        "\n",
        "\n",
        "\n",
        "loss, acc = model.evaluate(test_x, test_y)\n",
        "print(\"loss=\", loss)\n",
        "print(\"acc=\", acc)\n",
        "\n",
        "\n",
        "y_ = model.predict(test_x)\n",
        "predicted = np.argmax(y_, axis=-1)\n",
        "plt.plot(test_y, 'o')\n",
        "plt.plot(predicted, ',')\n",
        "plt.show()"
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_16\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "gru_2 (GRU)                  (None, 128)               50688     \n",
            "_________________________________________________________________\n",
            "flatten_12 (Flatten)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_12 (Dense)             (None, 2)                 258       \n",
            "=================================================================\n",
            "Total params: 50,946\n",
            "Trainable params: 50,946\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "7/7 [==============================] - 0s 2ms/step - loss: 0.1248 - acc: 0.9600\n",
            "loss= 0.12476766854524612\n",
            "acc= 0.9599999785423279\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jCBiKCg9AeHT"
      },
      "source": [
        "## Bidirectional + LSTM"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "azGH_FDbwqHT",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 819
        },
        "outputId": "d0e04db1-1873-4007-84a1-c9ddd3497ccc"
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Dense, RepeatVector, Flatten\n",
        "from tensorflow.keras.layers import Bidirectional, LSTM, GRU\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "# input_shape = (5,2)\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(Bidirectional(LSTM(128), input_shape=input_shape))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(2))\n",
        "model.compile(loss=\"sparse_categorical_crossentropy\", optimizer=\"adam\", metrics=[\"acc\"])\n",
        "model.summary()\n",
        "\n",
        "history = model.fit(train_x, train_y, epochs=10, verbose=0, validation_split=0.1, batch_size=128)\n",
        "\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.show()\n",
        "\n",
        "\n",
        "\n",
        "loss, acc = model.evaluate(test_x, test_y)\n",
        "print(\"loss=\", loss)\n",
        "print(\"acc=\", acc)\n",
        "\n",
        "\n",
        "y_ = model.predict(test_x)\n",
        "predicted = np.argmax(y_, axis=-1)\n",
        "plt.plot(test_y, 'o')\n",
        "plt.plot(predicted, ',')\n",
        "plt.show()"
      ],
      "execution_count": 43,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_17\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "bidirectional_2 (Bidirection (None, 256)               134144    \n",
            "_________________________________________________________________\n",
            "flatten_13 (Flatten)         (None, 256)               0         \n",
            "_________________________________________________________________\n",
            "dense_13 (Dense)             (None, 2)                 514       \n",
            "=================================================================\n",
            "Total params: 134,658\n",
            "Trainable params: 134,658\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "7/7 [==============================] - 0s 4ms/step - loss: 0.1040 - acc: 0.9900\n",
            "loss= 0.10397318005561829\n",
            "acc= 0.9900000095367432\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TfY_p_SJA5Rf"
      },
      "source": [
        "## Bidirectional + GRU"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OcjycCe8wqEH",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 819
        },
        "outputId": "ddfe4050-03cc-49dd-b55b-36c6b1432fe4"
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Dense, RepeatVector, Flatten\n",
        "from tensorflow.keras.layers import Bidirectional, LSTM, GRU\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "# input_shape = (5,2)\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(Bidirectional(GRU(128), input_shape=input_shape))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(2))\n",
        "model.compile(loss=\"sparse_categorical_crossentropy\", optimizer=\"adam\", metrics=[\"acc\"])\n",
        "model.summary()\n",
        "\n",
        "history = model.fit(train_x, train_y, epochs=10, verbose=0, validation_split=0.1, batch_size=128)\n",
        "\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.show()\n",
        "\n",
        "\n",
        "\n",
        "loss, acc = model.evaluate(test_x, test_y)\n",
        "print(\"loss=\", loss)\n",
        "print(\"acc=\", acc)\n",
        "\n",
        "\n",
        "y_ = model.predict(test_x)\n",
        "predicted = np.argmax(y_, axis=-1)\n",
        "plt.plot(test_y, 'o')\n",
        "plt.plot(predicted, ',')\n",
        "plt.show()"
      ],
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_18\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "bidirectional_3 (Bidirection (None, 256)               101376    \n",
            "_________________________________________________________________\n",
            "flatten_14 (Flatten)         (None, 256)               0         \n",
            "_________________________________________________________________\n",
            "dense_14 (Dense)             (None, 2)                 514       \n",
            "=================================================================\n",
            "Total params: 101,890\n",
            "Trainable params: 101,890\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "7/7 [==============================] - 0s 3ms/step - loss: 0.4263 - acc: 0.9600\n",
            "loss= 0.42628583312034607\n",
            "acc= 0.9599999785423279\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "P3aO3WEwA-lR"
      },
      "source": [
        "## stacked"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xIYi9GUDwqA6",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 855
        },
        "outputId": "b318e850-84a6-44a4-b687-d0cf80e68d16"
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Dense, RepeatVector, Flatten\n",
        "from tensorflow.keras.layers import Bidirectional, LSTM, GRU\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "# input_shape = (5,2)\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(LSTM(128, input_shape=input_shape, return_sequences=True))\n",
        "model.add(LSTM(128))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(2))\n",
        "model.compile(loss=\"sparse_categorical_crossentropy\", optimizer=\"adam\", metrics=[\"acc\"])\n",
        "model.summary()\n",
        "\n",
        "history = model.fit(train_x, train_y, epochs=10, verbose=0, validation_split=0.1, batch_size=128)\n",
        "\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.show()\n",
        "\n",
        "\n",
        "\n",
        "loss, acc = model.evaluate(test_x, test_y)\n",
        "print(\"loss=\", loss)\n",
        "print(\"acc=\", acc)\n",
        "\n",
        "\n",
        "y_ = model.predict(test_x)\n",
        "predicted = np.argmax(y_, axis=-1)\n",
        "plt.plot(test_y, 'o')\n",
        "plt.plot(predicted, ',')\n",
        "plt.show()"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_19\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "lstm_6 (LSTM)                (None, 5, 128)            67072     \n",
            "_________________________________________________________________\n",
            "lstm_7 (LSTM)                (None, 128)               131584    \n",
            "_________________________________________________________________\n",
            "flatten_15 (Flatten)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_15 (Dense)             (None, 2)                 258       \n",
            "=================================================================\n",
            "Total params: 198,914\n",
            "Trainable params: 198,914\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "7/7 [==============================] - 0s 3ms/step - loss: 0.3058 - acc: 0.9000\n",
            "loss= 0.30578455328941345\n",
            "acc= 0.8999999761581421\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OVGRQYg5beg_"
      },
      "source": [
        "## CNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zIc_XMibgSiB",
        "outputId": "6cde1d52-78a5-4fc4-fe9f-fbcaf0bf8af8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 999
        }
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Dense, RepeatVector, Flatten\n",
        "from tensorflow.keras.layers import Bidirectional, LSTM, GRU\n",
        "from tensorflow.keras.layers import Conv2D, MaxPooling2D, Conv1D, MaxPooling1D, Input, Reshape\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "# input_shape = (5,2)\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(Conv1D(32, 3, padding=\"same\", input_shape=input_shape))\n",
        "model.add(Conv1D(32, 3, padding=\"same\"))\n",
        "model.add(MaxPooling1D())\n",
        "model.add(Conv1D(32, 3, padding=\"same\"))\n",
        "model.add(Conv1D(32, 3, padding=\"same\"))\n",
        "model.add(MaxPooling1D())\n",
        "model.add(Flatten())\n",
        "model.add(Dense(2))\n",
        "model.compile(loss=\"sparse_categorical_crossentropy\", optimizer=\"adam\", metrics=[\"acc\"])\n",
        "model.summary()\n",
        "\n",
        "history = model.fit(train_x, train_y, epochs=10, verbose=0, validation_split=0.1, batch_size=128)\n",
        "\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.show()\n",
        "\n",
        "\n",
        "\n",
        "loss, acc = model.evaluate(test_x, test_y)\n",
        "print(\"loss=\", loss)\n",
        "print(\"acc=\", acc)\n",
        "\n",
        "\n",
        "y_ = model.predict(test_x)\n",
        "predicted = np.argmax(y_, axis=-1)\n",
        "plt.plot(test_y, 'o')\n",
        "plt.plot(predicted, ',')\n",
        "plt.show()"
      ],
      "execution_count": 46,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_20\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv1d_16 (Conv1D)           (None, 5, 32)             224       \n",
            "_________________________________________________________________\n",
            "conv1d_17 (Conv1D)           (None, 5, 32)             3104      \n",
            "_________________________________________________________________\n",
            "max_pooling1d_8 (MaxPooling1 (None, 2, 32)             0         \n",
            "_________________________________________________________________\n",
            "conv1d_18 (Conv1D)           (None, 2, 32)             3104      \n",
            "_________________________________________________________________\n",
            "conv1d_19 (Conv1D)           (None, 2, 32)             3104      \n",
            "_________________________________________________________________\n",
            "max_pooling1d_9 (MaxPooling1 (None, 1, 32)             0         \n",
            "_________________________________________________________________\n",
            "flatten_16 (Flatten)         (None, 32)                0         \n",
            "_________________________________________________________________\n",
            "dense_16 (Dense)             (None, 2)                 66        \n",
            "=================================================================\n",
            "Total params: 9,602\n",
            "Trainable params: 9,602\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "7/7 [==============================] - 0s 2ms/step - loss: 0.1074 - acc: 0.9600\n",
            "loss= 0.10743852704763412\n",
            "acc= 0.9599999785423279\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAT4klEQVR4nO3de4xcZ3nH8e+TtUMMBAL1FgUnwaYyUcOtCauAxKVIQO0EiLmkYKe0paVEqASBoG6NUoUoVcXFKipV09LQRlwEhIRCakGQaWla1IrQbOLcnGBicmnshsSEWysMcZynf8yZMFnP5czsmZndl+9HsnbmzJlnnvPOOz+fPXNmJzITSdLyd9S0G5AkNcNAl6RCGOiSVAgDXZIKYaBLUiFWTOuBV69enWvXrp3Ww0vSsnTdddd9LzNnu902tUBfu3Yt8/Pz03p4SVqWIuLuXrd5yEWSCmGgS1IhDHRJKoSBLkmFMNAlqRADz3KJiEuBVwH3Z+azutwewEeAM4GfAG/OzOubbhTgyl372b5zD/t/eJAA2n9W7EmPXcn7Xv1MXnPqmqHu10udehfu2M0PDx4C4KiAh5NH1W4vW3PcKrZuOLlnrW71FtPXoO0ctq8m6vWqM+q4Nbmddce+6XqDnst2rbrjNuycXUxfTdYb92uzU3vcZiI4nLnoOTbqdtZ5/FHFoL+2GBEvAf4P+GSPQD8TeAetQH8+8JHMfP6gB56bm8thTlu8ctd+3vuFmzl46HDX21fOBNvPfu4RgzPofr30q7f1ihs59HD9v1K5auUM73/ds3tOnGHqNbmdg/pqot6o499kvV7bOcpz2WS9Xs9lu9aw29nUnO3XV5P1luprcxx99arX7zXYS0Rcl5lz3W4beMglM78OfL/PKptohX1m5jXAcRFxfO3uatq+c88jA3LXMec88rN9+dDhZPvOPUPfb+Hltn712hOm2327LTt46HDXWoPqDdvXsNs5qK8m6tWps3BZ26j16m5n3bGv09co9Xo9l53bOWjbFjtnh+2ryXqTfG32Grem5tgo29nr8Rdj4B46QESsBb7UYw/9S8AHMvM/qutfA/4kM4/Y/Y6Ic4FzAU466aTn3X13z/Pjj7Bu25cH/toTwJ0feOXQ95tEvW61Rq23VPvqVm8x499kvaU6N8Y9/k331WS9pTD+3WpNuq9+Y911/cXsoTcpMy/JzLnMnJud7frJ1Z6eetyqkdapc79J1Ot1v1HqLdW+ut1vMePfZL2lOjfGPf6j1up3n2nP2abrNTlnR+lrsa+RTk0E+n7gxI7rJ1TLGrV1w8msWjnT8/aVM8HWDScPfb9R6q08KoaqtWrlTNdao9RrcjsH9dVEvVHHv8l6vbZzlOeyyXq9nst2rWG3s6k526+vJust1dfmOPrqVa/fa3AUTfwtlx3AeRFxGa03RX+Umfc2UPdR2m8aDHuWS7/79VKnXlNnuXSrt5i+mjr7o6l6/eqMMm5NbucwY990vUFnRQw7bsPO2VH7arLeJF6bneqe5TKuvpbKWS6fBV4KrAbuA94HrATIzI9Wpy3+NbCR1mmLv9ft+PlCw57lIknqfwx94B56Zm4ZcHsCbx+xN0lSQ/ykqCQVwkCXpEIY6JJUCANdkgphoEtSIQx0SSqEgS5JhTDQJakQBrokFcJAl6RCGOiSVAgDXZIKYaBLUiEMdEkqhIEuSYUw0CWpEAa6JBXCQJekQhjoklQIA12SCmGgS1IhDHRJKoSBLkmFMNAlqRAGuiQVwkCXpEIY6JJUCANdkgphoEtSIQx0SSqEgS5JhagV6BGxMSL2RMTeiNjW5faTIuLqiNgVETdFxJnNtypJ6mdgoEfEDHAxcAZwCrAlIk5ZsNqfApdn5qnAZuBvmm5UktRfnT3004G9mXlHZj4IXAZsWrBOAk+oLj8R+J/mWpQk1VEn0NcA93Rc31ct63Qh8KaI2AdcBbyjW6GIODci5iNi/sCBAyO0K0nqpak3RbcAH8/ME4AzgU9FxBG1M/OSzJzLzLnZ2dmGHlqSBPUCfT9wYsf1E6plnd4CXA6Qmd8AjgFWN9GgJKmeOoF+LbA+ItZFxNG03vTcsWCd/wZeBhARv0or0D2mIkkTNDDQM/Mh4DxgJ3AbrbNZdkfERRFxVrXae4C3RsSNwGeBN2dmjqtpSdKRVtRZKTOvovVmZ+eyCzou3wq8sNnWJEnD8JOiklQIA12SCmGgS1IhDHRJKoSBLkmFMNAlqRAGuiQVwkCXpEIY6JJUCANdkgphoEtSIQx0SSqEgS5JhTDQJakQBrokFcJAl6RCGOiSVAgDXZIKYaBLUiEMdEkqhIEuSYUw0CWpEAa6JBXCQJekQhjoklQIA12SCmGgS1IhDHRJKoSBLkmFMNAlqRC1Aj0iNkbEnojYGxHbeqzzhoi4NSJ2R8Rnmm1TkjTIikErRMQMcDHwCmAfcG1E7MjMWzvWWQ+8F3hhZv4gIn55XA1Lkrqrs4d+OrA3M+/IzAeBy4BNC9Z5K3BxZv4AIDPvb7ZNSdIgdQJ9DXBPx/V91bJOzwCeERH/GRHXRMTGboUi4tyImI+I+QMHDozWsSSpq6beFF0BrAdeCmwBPhYRxy1cKTMvycy5zJybnZ1t6KElSVAv0PcDJ3ZcP6Fa1mkfsCMzD2XmncC3aQW8JGlC6gT6tcD6iFgXEUcDm4EdC9a5ktbeORGxmtYhmDsa7FOSNMDAQM/Mh4DzgJ3AbcDlmbk7Ii6KiLOq1XYCD0TErcDVwNbMfGBcTUuSjhSZOZUHnpuby/n5+ak8tiQtVxFxXWbOdbvNT4pKUiEMdEkqhIEuSYUw0CWpEAa6JBXCQJekQhjoklQIA12SCmGgS1IhDHRJKoSBLkmFMNAlqRAGuiQVwkCXpEIY6JJUCANdkgphoEtSIQx0SSqEgS5JhTDQJakQBrokFcJAl6RCGOiSVAgDXZIKYaBLUiEMdEkqhIEuSYUw0CWpEAa6JBXCQJekQhjoklSIWoEeERsjYk9E7I2IbX3We31EZETMNdeiJKmOgYEeETPAxcAZwCnAlog4pct6xwLvBL7ZdJOSpMHq7KGfDuzNzDsy80HgMmBTl/X+DPgg8NMG+5Mk1VQn0NcA93Rc31cte0REnAacmJlf7lcoIs6NiPmImD9w4MDQzUqSelv0m6IRcRTwYeA9g9bNzEsycy4z52ZnZxf70JKkDnUCfT9wYsf1E6plbccCzwL+LSLuAl4A7PCNUUmarDqBfi2wPiLWRcTRwGZgR/vGzPxRZq7OzLWZuRa4BjgrM+fH0rEkqauBgZ6ZDwHnATuB24DLM3N3RFwUEWeNu0FJUj0r6qyUmVcBVy1YdkGPdV+6+LYkScPyk6KSVAgDXZIKYaBLUiEMdEkqhIEuSYUw0CWpEAa6JBXCQJekQhjoklQIA12SCmGgS1IhDHRJKoSBLkmFMNAlqRAGuiQVwkCXpEIY6JJUCANdkgphoEtSIQx0SSqEgS5JhTDQJakQBrokFcJAl6RCGOiSVAgDXZIKYaBLUiEMdEkqhIEuSYUw0CWpELUCPSI2RsSeiNgbEdu63P7uiLg1Im6KiK9FxNOab1WS1M/AQI+IGeBi4AzgFGBLRJyyYLVdwFxmPgf4PPChphuVJPVXZw/9dGBvZt6RmQ8ClwGbOlfIzKsz8yfV1WuAE5ptU5I0SJ1AXwPc03F9X7Wsl7cAX+l2Q0ScGxHzETF/4MCB+l1KkgZq9E3RiHgTMAds73Z7Zl6SmXOZOTc7O9vkQ0vSL7wVNdbZD5zYcf2EatmjRMTLgfOBX8/MnzXTniSprjp76NcC6yNiXUQcDWwGdnSuEBGnAn8HnJWZ9zffpiRpkIGBnpkPAecBO4HbgMszc3dEXBQRZ1WrbQceD1wRETdExI4e5SRJY1LnkAuZeRVw1YJlF3RcfnnDfUmShuQnRSWpEAa6JBXCQJekQhjoklQIA12SCmGgS1IhDHRJKoSBLkmFMNAlqRAGuiQVwkCXpEIY6JJUCANdkgphoEtSIQx0SSqEgS5JhTDQJakQBrokFcJAl6RCGOiSVAgDXZIKYaBLUiEMdEkqhIEuSYUw0CWpEAa6JBXCQJekQhjoklQIA12SCmGgS1IhDHRJKsSKOitFxEbgI8AM8PeZ+YEFtz8G+CTwPOAB4I2ZeVezrXZ35a79XLhjNz88eOiI244KeDhhJoLDmaw5bhVbN5zMa05dM1K9Tk967Ere9+pn9qx15a79bN+5h/0/PEgAOWA7mqjX3t5+2zlMnTrj1lRf49zO9nb0qjdo7JvezrpzrOl6deZYZ5324486bk311XS9cczZQeM2TL3FiMz+URMRM8C3gVcA+4BrgS2ZeWvHOn8IPCcz3xYRm4HXZuYb+9Wdm5vL+fn5RTV/5a79bL3iRg49PCguf27Vyhne/7pn9wyBYeqtnAm2n/3cI2pduWs/7/3CzRw8dLh2X03X67ado/bVZL1B4z+t7ew19k321a417Jxtsl6/OTZKX03V6zf+TdYbx5xtMoPqiIjrMnOu2211DrmcDuzNzDsy80HgMmDTgnU2AZ+oLn8eeFlExEjdDmH7zj2PDORdx5zDXcec0/Vy58+Dhw6zfeeenvVuP3rLwHrty4cOZ9da23fueWTCDKpRt95tM2+sXa/Xdg7b1zjqDRr/g4cO1xqrUfrqV6/X2DfZV7tWnTnbvtzWr15Tc7ZXX03XG3b8m97Opudsr75GqbdYdfbQzwY2ZuYfVNd/G3h+Zp7Xsc4t1Tr7quvfqdb53oJa5wLnApx00knPu/vuuxfV/LptXx54KKObAO78wCsbqdet1qh9NV1vYa3F9NVkvSbHfzn1tZhaTdZbTn01XW8pz43a913kHnpjMvOSzJzLzLnZ2dlF13vqcasavd8o9brdZ9S+mq638H6L6avJek2Of7f7LdW+FlOryXrLqa+m6y3ludGEOoG+Hzix4/oJ1bKu60TECuCJtN4cHautG05m5VHDHdlZtXKGrRtObqTeypnoWmvrhpNZtXJmqL6artdtO0ftq8l6g8Z/WtvZa+yb7Ktda9g522S9fnNslL6aqtdv/JusN44522QGLVads1yuBdZHxDpawb0ZOGfBOjuA3wW+AZwN/GsOOpbTgPabCk2d5TKoXqd+76S3lzV1lkvdeoPeSR+2zqBxa6qvcW/nYs9yaXI7h5ljTderM8eaOsulqb6arjeuObtsznIBiIgzgb+kddripZn55xFxETCfmTsi4hjgU8CpwPeBzZl5R7+aTZzlIkm/aPodQ691HnpmXgVctWDZBR2Xfwr85mKalCQtjp8UlaRCGOiSVAgDXZIKYaBLUiFqneUylgeOOACM+lHR1cD3Bq41HUu1N/sajn0Nb6n2VlpfT8vMrp/MnFqgL0ZEzPc6bWfalmpv9jUc+xreUu3tF6kvD7lIUiEMdEkqxHIN9Eum3UAfS7U3+xqOfQ1vqfb2C9PXsjyGLkk60nLdQ5ckLWCgS1Ihll2gR8TGiNgTEXsjYtsU+zgxIq6OiFsjYndEvLNafmFE7I+IG6p/Z06ht7si4ubq8eerZU+OiH+OiNurn0+acE8nd4zJDRHx44h417TGKyIujYj7q2/bai/rOkbR8lfVnLspIk6bcF/bI+Jb1WN/MSKOq5avjYiDHWP30Qn31fO5i4j3VuO1JyI2jKuvPr19rqOvuyLihmr5RMasTz6Md45l5rL5R+vP934HeDpwNHAjcMqUejkeOK26fCytL9I+BbgQ+KMpj9NdwOoFyz4EbKsubwM+OOXn8bvA06Y1XsBLgNOAWwaNEXAm8BVaf+L6BcA3J9zXbwArqssf7Ohrbed6Uxivrs9d9Tq4EXgMsK56zc5MsrcFt/8FcMEkx6xPPox1ji23PfQ6X1g9EZl5b2ZeX13+X+A2YDx/tb4ZnV/k/QngNVPs5WXAdzJzcV8quwiZ+XVaf7u/U68x2gR8MluuAY6LiOMn1VdmfjUzH6quXkPrW8Mmqsd49bIJuCwzf5aZdwJ7ab12J95bRATwBuCz43r8Hj31yoexzrHlFuhrgHs6ru9jCYRoRKyl9eUe36wWnVf92nTppA9tVBL4akRcF60v5gZ4SmbeW13+LvCUKfTVtplHv8CmPV5tvcZoKc2736e1J9e2LiJ2RcS/R8SLp9BPt+duKY3Xi4H7MvP2jmUTHbMF+TDWObbcAn3JiYjHA/8IvCszfwz8LfArwK8B99L6dW/SXpSZpwFnAG+PiJd03pit3/Gmcr5qRBwNnAVcUS1aCuN1hGmOUS8RcT7wEPDpatG9wEmZeSrwbuAzEfGECba0JJ+7Bbbw6J2HiY5Zl3x4xDjm2HIL9DpfWD0xEbGS1pP16cz8AkBm3peZhzPzYeBjjPFXzV4yc3/1837gi1UP97V/hat+3j/pvipnANdn5n1Vj1Mfrw69xmjq8y4i3gy8CvitKgioDmk8UF2+jtax6mdMqqc+z93Uxwse+cL61wGfay+b5Jh1ywfGPMeWW6A/8oXV1Z7eZlpfUD1x1bG5fwBuy8wPdyzvPO71WuCWhfcdc1+Pi4hj25dpvaF2Cz//Im+qn/80yb46PGqPadrjtUCvMdoB/E51JsILgB91/No8dhGxEfhj4KzM/EnH8tmImKkuPx1YD/T9Lt+G++r13O0ANkfEY6L15fLrgf+aVF8dXg58KzP3tRdMasx65QPjnmPjfre36X+03g3+Nq3/Wc+fYh8vovXr0k3ADdW/M2l9WfbN1fIdwPET7uvptM4wuBHY3R4j4JeArwG3A/8CPHkKY/Y44AHgiR3LpjJetP5TuRc4ROt45Vt6jRGtMw8urubczcDchPvaS+v4anuefbRa9/XVc3wDcD3w6gn31fO5A86vxmsPcMakn8tq+ceBty1YdyJj1icfxjrH/Oi/JBViuR1ykST1YKBLUiEMdEkqhIEuSYUw0CWpEAa6JBXCQJekQvw//y5JtnaJv1QAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0MW7OTLyfph4"
      },
      "source": [
        "## CNN + LSTM"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RHQL-koAgiXH",
        "outputId": "2de00889-d5da-4013-be94-de46225f3d53",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 891
        }
      },
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras.layers import Dense, RepeatVector, Flatten\n",
        "from tensorflow.keras.layers import Bidirectional, LSTM, GRU\n",
        "from tensorflow.keras.layers import Conv2D, MaxPooling2D, Conv1D, MaxPooling1D, Input, Reshape\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "# input_shape = (5,2)\n",
        "\n",
        "model = keras.Sequential()\n",
        "model.add(Conv1D(32, 3, padding=\"same\", input_shape=input_shape))\n",
        "model.add(Conv1D(32, 3, padding=\"same\"))\n",
        "model.add(LSTM(128))\n",
        "model.add(Flatten())\n",
        "model.add(Dense(2))\n",
        "model.compile(loss=\"sparse_categorical_crossentropy\", optimizer=\"adam\", metrics=[\"acc\"])\n",
        "model.summary()\n",
        "\n",
        "history = model.fit(train_x, train_y, epochs=10, verbose=0, validation_split=0.1, batch_size=128)\n",
        "\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.show()\n",
        "\n",
        "\n",
        "\n",
        "loss, acc = model.evaluate(test_x, test_y)\n",
        "print(\"loss=\", loss)\n",
        "print(\"acc=\", acc)\n",
        "\n",
        "\n",
        "y_ = model.predict(test_x)\n",
        "predicted = np.argmax(y_, axis=-1)\n",
        "plt.plot(test_y, 'o')\n",
        "plt.plot(predicted, ',')\n",
        "plt.show()"
      ],
      "execution_count": 47,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_21\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv1d_20 (Conv1D)           (None, 5, 32)             224       \n",
            "_________________________________________________________________\n",
            "conv1d_21 (Conv1D)           (None, 5, 32)             3104      \n",
            "_________________________________________________________________\n",
            "lstm_8 (LSTM)                (None, 128)               82432     \n",
            "_________________________________________________________________\n",
            "flatten_17 (Flatten)         (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_17 (Dense)             (None, 2)                 258       \n",
            "=================================================================\n",
            "Total params: 86,018\n",
            "Trainable params: 86,018\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "7/7 [==============================] - 0s 3ms/step - loss: 0.0555 - acc: 0.9850\n",
            "loss= 0.05545246973633766\n",
            "acc= 0.9850000143051147\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "emE8KhR9BF-Z"
      },
      "source": [
        ""
      ],
      "execution_count": 45,
      "outputs": []
    }
  ]
}