过拟合与欠拟合.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "authorship_tag": "ABX9TyMyTZsRncTnX9AKS3f9cEhg",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/DENGCPU/TensorFlow-Study/blob/main/%E8%BF%87%E6%8B%9F%E5%90%88%E4%B8%8E%E6%AC%A0%E6%8B%9F%E5%90%88.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "w-sE5erW_wUy",
        "outputId": "6f0deca3-761f-4468-d9af-1b5245144008"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2.15.0\n"
          ]
        }
      ],
      "source": [
        "import  tensorflow as tf\n",
        "\n",
        "from tensorflow.keras import layers\n",
        "from tensorflow.keras import regularizers\n",
        "\n",
        "print(tf.__version__)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install git+https://github.com/tensorflow/docs"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jx1xISwtH5lg",
        "outputId": "b5413952-6837-410f-aca0-db30f7c421d7"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting git+https://github.com/tensorflow/docs\n",
            "  Cloning https://github.com/tensorflow/docs to /tmp/pip-req-build-a11dwzws\n",
            "  Running command git clone --filter=blob:none --quiet https://github.com/tensorflow/docs /tmp/pip-req-build-a11dwzws\n",
            "  Resolved https://github.com/tensorflow/docs to commit 75b2672b5bed8ca0995663536db84bd9a39b8896\n",
            "  Preparing metadata (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "Collecting astor (from tensorflow-docs==2024.5.3.31743)\n",
            "  Downloading astor-0.8.1-py2.py3-none-any.whl (27 kB)\n",
            "Requirement already satisfied: absl-py in /usr/local/lib/python3.10/dist-packages (from tensorflow-docs==2024.5.3.31743) (1.4.0)\n",
            "Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from tensorflow-docs==2024.5.3.31743) (3.1.4)\n",
            "Requirement already satisfied: nbformat in /usr/local/lib/python3.10/dist-packages (from tensorflow-docs==2024.5.3.31743) (5.10.4)\n",
            "Requirement already satisfied: protobuf>=3.12 in /usr/local/lib/python3.10/dist-packages (from tensorflow-docs==2024.5.3.31743) (3.20.3)\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.10/dist-packages (from tensorflow-docs==2024.5.3.31743) (6.0.1)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->tensorflow-docs==2024.5.3.31743) (2.1.5)\n",
            "Requirement already satisfied: fastjsonschema>=2.15 in /usr/local/lib/python3.10/dist-packages (from nbformat->tensorflow-docs==2024.5.3.31743) (2.19.1)\n",
            "Requirement already satisfied: jsonschema>=2.6 in /usr/local/lib/python3.10/dist-packages (from nbformat->tensorflow-docs==2024.5.3.31743) (4.19.2)\n",
            "Requirement already satisfied: jupyter-core!=5.0.*,>=4.12 in /usr/local/lib/python3.10/dist-packages (from nbformat->tensorflow-docs==2024.5.3.31743) (5.7.2)\n",
            "Requirement already satisfied: traitlets>=5.1 in /usr/local/lib/python3.10/dist-packages (from nbformat->tensorflow-docs==2024.5.3.31743) (5.7.1)\n",
            "Requirement already satisfied: attrs>=22.2.0 in /usr/local/lib/python3.10/dist-packages (from jsonschema>=2.6->nbformat->tensorflow-docs==2024.5.3.31743) (23.2.0)\n",
            "Requirement already satisfied: jsonschema-specifications>=2023.03.6 in /usr/local/lib/python3.10/dist-packages (from jsonschema>=2.6->nbformat->tensorflow-docs==2024.5.3.31743) (2023.12.1)\n",
            "Requirement already satisfied: referencing>=0.28.4 in /usr/local/lib/python3.10/dist-packages (from jsonschema>=2.6->nbformat->tensorflow-docs==2024.5.3.31743) (0.35.1)\n",
            "Requirement already satisfied: rpds-py>=0.7.1 in /usr/local/lib/python3.10/dist-packages (from jsonschema>=2.6->nbformat->tensorflow-docs==2024.5.3.31743) (0.18.1)\n",
            "Requirement already satisfied: platformdirs>=2.5 in /usr/local/lib/python3.10/dist-packages (from jupyter-core!=5.0.*,>=4.12->nbformat->tensorflow-docs==2024.5.3.31743) (4.2.2)\n",
            "Building wheels for collected packages: tensorflow-docs\n",
            "  Building wheel for tensorflow-docs (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for tensorflow-docs: filename=tensorflow_docs-2024.5.3.31743-py3-none-any.whl size=182531 sha256=7f0b426cbf5dae601b11d55e407d78056c744d2e17e6811a81f97165d0f3268c\n",
            "  Stored in directory: /tmp/pip-ephem-wheel-cache-0rcj6kp7/wheels/86/0f/1e/3b62293c8ffd0fd5a49508e6871cdb7554abe9c62afd35ec53\n",
            "Successfully built tensorflow-docs\n",
            "Installing collected packages: astor, tensorflow-docs\n",
            "Successfully installed astor-0.8.1 tensorflow-docs-2024.5.3.31743\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow_docs as tfdocs\n",
        "import tensorflow_docs.modeling\n",
        "import tensorflow_docs.plots\n",
        "\n",
        "from  IPython import display\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "import numpy as np\n",
        "\n",
        "import pathlib\n",
        "import shutil\n",
        "import tempfile\n",
        "\n",
        "logdir = pathlib.Path(tempfile.mkdtemp())/\"tensorboard_logs\"\n",
        "shutil.rmtree(logdir, ignore_errors=True)"
      ],
      "metadata": {
        "id": "1F5FSnkdG5Po"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 希格斯数据集：11,000,000个样本，每个样本有28个特征和 1个二元类标签\n",
        "gz = tf.keras.utils.get_file('HIGGS.csv.gz', 'http://mlphysics.ics.uci.edu/data/higgs/HIGGS.csv.gz')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bYDjhRgXHpdy",
        "outputId": "92003804-6625-4140-d790-583bbcc4cf86"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from http://mlphysics.ics.uci.edu/data/higgs/HIGGS.csv.gz\n",
            "2816407858/2816407858 [==============================] - 41s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "FEATURES = 28\n",
        "# 从Gzip文件读取CSV记录, 返回一个标量列表\n",
        "ds = tf.data.experimental.CsvDataset(gz,[float(),]*(FEATURES+1),compression_type='GZIP')\n",
        "\n",
        "# 将标量列表重新打包为（feature_vector,label)对\n",
        "def pack_row(*row):\n",
        "  label = row[0]\n",
        "  features = tf.stack(row[1:],1)\n",
        "  return features , label\n",
        "\n",
        "# 接收10000个样本为单位的批次，将pack_row函数应用于每个批次，然后将批次重新拆分为单个记录\n",
        "packed_ds = ds.batch(10000).map(pack_row).unbatch()\n",
        "\n",
        "# 检查这个新的packed_ds中的一些记录\n",
        "for features, label in packed_ds.batch(1000).take(1):\n",
        "  print(features[0])\n",
        "  plt.hist(features.numpy().flatten(),bins=101)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 554
        },
        "id": "_f6pQqsMI4J4",
        "outputId": "5cc62060-7a4e-40a5-c36c-16b86334d80e"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tf.Tensor(\n",
            "[ 0.8692932  -0.6350818   0.22569026  0.32747006 -0.6899932   0.75420225\n",
            " -0.24857314 -1.0920639   0.          1.3749921  -0.6536742   0.9303491\n",
            "  1.1074361   1.1389043  -1.5781983  -1.0469854   0.          0.65792954\n",
            " -0.01045457 -0.04576717  3.1019614   1.35376     0.9795631   0.97807616\n",
            "  0.92000484  0.72165745  0.98875093  0.87667835], shape=(28,), dtype=float32)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 使用前1000个样本进行验证，再用接下来的10,000个样本机械能训练\n",
        "N_VALIDATION = int(1e3) #1，000\n",
        "N_TRAIN = int(1e4)  #10,000\n",
        "BUFFER_SIZE = int(1e4)\n",
        "BATCH_SIZE = 500\n",
        "STEPS_PER_EPOCH = N_TRAIN  # batch_size\n",
        "\n",
        "validate_ds = packed_ds.take(N_VALIDATION).cache()\n",
        "train_ds = packed_ds.skip(N_VALIDATION).take(N_TRAIN).cache()\n",
        "\n",
        "train_ds"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MxmsAevnLBnB",
        "outputId": "86386d69-a9b3-4836-973b-ba2d14d53121"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<CacheDataset element_spec=(TensorSpec(shape=(28,), dtype=tf.float32, name=None), TensorSpec(shape=(), dtype=tf.float32, name=None))>"
            ]
          },
          "metadata": {},
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 使用Dataset.batch方法创建适当大小的批次进行训练\n",
        "validate_ds = validate_ds.batch(BATCH_SIZE)\n",
        "train_ds = train_ds.shuffle(BUFFER_SIZE).repeat().batch(BATCH_SIZE)"
      ],
      "metadata": {
        "id": "REj1XqwdMVgk"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 演示过拟合\n",
        "\n",
        "# 训练过程\n",
        "# 使用tf.keras.optimizers.schedules 随着时间的推移减小学习率\n",
        "lr_schedule = tf.keras.optimizers.schedules.InverseTimeDecay(\n",
        "    0.001,\n",
        "    decay_steps = STEPS_PER_EPOCH * 1000,\n",
        "    decay_rate=1,\n",
        "    staircase = False\n",
        ")\n",
        "def get_optimier():\n",
        "  return tf.keras.optimizers.Adam(lr_schedule)"
      ],
      "metadata": {
        "id": "sELkOjXoMyJg"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "step = np.linspace(0,100000)\n",
        "lr = lr_schedule(step)\n",
        "plt.figure(figsize=(8,6))\n",
        "plt.plot(step / STEPS_PER_EPOCH,lr)\n",
        "plt.ylim([0,max(plt.ylim())])\n",
        "plt.xlabel('Epoch')\n",
        "_ = plt.ylabel('Learning Rate')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 482
        },
        "id": "aNwxFz_4ZhCx",
        "outputId": "d4d9a2db-f979-48da-a27e-bebe50eb312d"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 800x600 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def get_callbacks(name):\n",
        "  return [\n",
        "      tfdocs.modeling.EpochDots(),\n",
        "      tf.keras.callbacks.EarlyStopping(monitor='val_binary_crossentropy',patience=200),\n",
        "      tf.keras.callbacks.TensorBoard(logdir/name)\n",
        "  ]\n",
        "\n",
        "# 每个模型使用相同的Model.compile和 Model.fit设置\n",
        "def compile_and_fit(model,name,optimizer=None,max_epochs=10000):\n",
        "  if optimizer is None:\n",
        "    optimizer = get_optimier()\n",
        "    model.compile(optimizer=optimizer,\n",
        "                  loss=tf.keras.losses.BinaryCrossentropy(from_logits=True),\n",
        "                  metrics=[\n",
        "                      tf.keras.metrics.BinaryCrossentropy(\n",
        "                          from_logits=True, name='binary_crossentropy'),\n",
        "                          'accuracy'\n",
        "\n",
        "                  ])\n",
        "    model.summary()\n",
        "\n",
        "    history=model.fit(\n",
        "    train_ds,\n",
        "    steps_per_epoch=STEPS_PER_EPOCH,\n",
        "    epochs=max_epochs,\n",
        "    validation_data=validate_ds,\n",
        "    callbacks=get_callbacks(name),\n",
        "    verbose=0\n",
        "    )\n",
        "    return history\n",
        "\n",
        "# 微模型\n",
        "tiny_model = tf.keras.Sequential([\n",
        "    layers.Dense(16, activation='relu',input_shape=(FEATURES,)),\n",
        "    layers.Dense(1)\n",
        "])\n",
        "size_histories = {}\n",
        "\n",
        "size_histories['Tiny'] = compile_and_fit(tiny_model, 'sizes/Tiny')\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ax95vWHTdMfM",
        "outputId": "8bccc092-a3e7-4f6b-fa67-b6aad2f3d47e"
      },
      "execution_count": null,
      "outputs": [
        {
          "metadata": {
            "tags": null
          },
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " dense (Dense)               (None, 16)                464       \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 1)                 17        \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 481 (1.88 KB)\n",
            "Trainable params: 481 (1.88 KB)\n",
            "Non-trainable params: 0 (0.00 Byte)\n",
            "_________________________________________________________________\n",
            "\n",
            "Epoch: 0, accuracy:0.6374,  binary_crossentropy:0.6011,  loss:0.6011,  val_accuracy:0.6610,  val_binary_crossentropy:0.5885,  val_loss:0.5885,  \n",
            "................................................."
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 看下模型的表现\n",
        "plotter = tfdocs.plots.HistoryPlotter(metric= 'binary_crossentropy',smoothing_std=10)\n",
        "plotter.plot(size_histories)\n",
        "plt.ylim([0.5,0.7])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 245
        },
        "id": "4ZBvk_xJHdYg",
        "outputId": "53396702-bd18-463f-bcf8-9e432b91880a"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "name 'tfdocs' is not defined",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-1-e6c8e836fff3>\u001b[0m in \u001b[0;36m<cell line: 2>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;31m# 看下模型的表现\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mplotter\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtfdocs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplots\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mHistoryPlotter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmetric\u001b[0m\u001b[0;34m=\u001b[0m \u001b[0;34m'binary_crossentropy'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0msmoothing_std\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m10\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0mplotter\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msize_histories\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mylim\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0.5\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m0.7\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mNameError\u001b[0m: name 'tfdocs' is not defined"
          ]
        }
      ]
    }
  ]
}