machinelearning_notebook/2_pytorch/1_NN/linear-regression-gradient-descend.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 线性模型和梯度下降\n",
    "这是神经网络的第一课，我们会学习一个非常简单的模型，线性回归，同时也会学习一个优化算法-梯度下降法，对这个模型进行优化。线性回归是监督学习里面一个非常简单的模型，同时梯度下降也是深度学习中应用最广的优化算法，我们将从这里开始我们的深度学习之旅"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 一元线性回归\n",
    "一元线性模型非常简单，假设我们有变量 $x_i$ 和目标 $y_i$，每个 i 对应于一个数据点，希望建立一个模型\n",
    "\n",
    "$$\n",
    "\\hat{y}_i = w x_i + b\n",
    "$$\n",
    "\n",
    "$\\hat{y}_i$ 是我们预测的结果，希望通过 $\\hat{y}_i$ 来拟合目标 $y_i$，通俗来讲就是找到这个函数拟合 $y_i$ 使得误差最小，即最小化\n",
    "\n",
    "$$\n",
    "\\frac{1}{n} \\sum_{i=1}^n(\\hat{y}_i - y_i)^2\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "那么如何最小化这个误差呢？\n",
    "\n",
    "这里需要用到**梯度下降**，这是我们接触到的第一个优化算法，非常简单，但是却非常强大，在深度学习中被大量使用，所以让我们从简单的例子出发了解梯度下降法的原理"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 梯度下降法\n",
    "在梯度下降法中，我们首先要明确梯度的概念，随后我们再了解如何使用梯度进行下降。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 梯度\n",
    "梯度在数学上就是导数，如果是一个多元函数，那么梯度就是偏导数。比如一个函数f(x, y)，那么 f 的梯度就是 \n",
    "\n",
    "$$\n",
    "(\\frac{\\partial f}{\\partial x},\\ \\frac{\\partial f}{\\partial y})\n",
    "$$\n",
    "\n",
    "可以称为 grad f(x, y) 或者 $\\nabla f(x, y)$。具体某一点 $(x_0,\\ y_0)$ 的梯度就是 $\\nabla f(x_0,\\ y_0)$。\n",
    "\n",
    "下面这个图片是 $f(x) = x^2$ 这个函数在 x=1 处的梯度\n",
    "\n",
    "![](https://ws3.sinaimg.cn/large/006tNc79ly1fmarbuh2j3j30ba0b80sy.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "梯度有什么意义呢？从几何意义来讲，一个点的梯度值是这个函数变化最快的地方，具体来说，对于函数 f(x, y)，在点 $(x_0, y_0)$ 处，沿着梯度 $\\nabla f(x_0,\\ y_0)$ 的方向，函数增加最快，也就是说沿着梯度的方向，我们能够更快地找到函数的极大值点，或者反过来沿着梯度的反方向，我们能够更快地找到函数的最小值点。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 梯度下降法\n",
    "有了对梯度的理解，我们就能了解梯度下降发的原理了。上面我们需要最小化这个误差，也就是需要找到这个误差的最小值点，那么沿着梯度的反方向我们就能够找到这个最小值点。\n",
    "\n",
    "我们可以来看一个直观的解释。比如我们在一座大山上的某处位置，由于我们不知道怎么下山，于是决定走一步算一步，也就是在每走到一个位置的时候，求解当前位置的梯度，沿着梯度的负方向，也就是当前最陡峭的位置向下走一步，然后继续求解当前位置梯度，向这一步所在位置沿着最陡峭最易下山的位置走一步。这样一步步的走下去，一直走到觉得我们已经到了山脚。当然这样走下去，有可能我们不能走到山脚，而是到了某一个局部的山峰低处。\n",
    "\n",
    "类比我们的问题，就是沿着梯度的反方向，我们不断改变 w 和 b 的值，最终找到一组最好的 w 和 b 使得误差最小。\n",
    "\n",
    "在更新的时候，我们需要决定每次更新的幅度，比如在下山的例子中，我们需要每次往下走的那一步的长度，这个长度称为学习率，用 $\\eta$ 表示，这个学习率非常重要，不同的学习率都会导致不同的结果，学习率太小会导致下降非常缓慢，学习率太大又会导致跳动非常明显，可以看看下面的例子\n",
    "\n",
    "![](https://ws2.sinaimg.cn/large/006tNc79ly1fmgn23lnzjg30980gogso.gif)\n",
    "\n",
    "可以看到上面的学习率较为合适，而下面的学习率太大，就会导致不断跳动\n",
    "\n",
    "最后我们的更新公式就是\n",
    "\n",
    "$$\n",
    "w := w - \\eta \\frac{\\partial f(w,\\ b)}{\\partial w} \\\\\n",
    "b := b - \\eta \\frac{\\partial f(w,\\ b)}{\\partial b}\n",
    "$$\n",
    "\n",
    "通过不断地迭代更新，最终我们能够找到一组最优的 w 和 b，这就是梯度下降法的原理。\n",
    "\n",
    "最后可以通过这张图形象地说明一下这个方法\n",
    "\n",
    "![](https://ws3.sinaimg.cn/large/006tNc79ly1fmarxsltfqj30gx091gn4.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "上面是原理部分，下面通过一个例子来进一步学习线性模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<torch._C.Generator at 0x112959630>"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "from torch.autograd import Variable\n",
    "\n",
    "torch.manual_seed(2017)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 读入数据 x 和 y\n",
    "x_train = np.array([[3.3], [4.4], [5.5], [6.71], [6.93], [4.168],\n",
    "                    [9.779], [6.182], [7.59], [2.167], [7.042],\n",
    "                    [10.791], [5.313], [7.997], [3.1]], dtype=np.float32)\n",
    "\n",
    "y_train = np.array([[1.7], [2.76], [2.09], [3.19], [1.694], [1.573],\n",
    "                    [3.366], [2.596], [2.53], [1.221], [2.827],\n",
    "                    [3.465], [1.65], [2.904], [1.3]], dtype=np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x11565e588>]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1146f2438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 画出图像\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "plt.plot(x_train, y_train, 'bo')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 转换成 Tensor\n",
    "x_train = torch.from_numpy(x_train)\n",
    "y_train = torch.from_numpy(y_train)\n",
    "\n",
    "# 定义参数 w 和 b\n",
    "w = Variable(torch.randn(1), requires_grad=True) # 随机初始化\n",
    "b = Variable(torch.zeros(1), requires_grad=True) # 使用 0 进行初始化"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 构建线性回归模型\n",
    "x_train = Variable(x_train)\n",
    "y_train = Variable(y_train)\n",
    "\n",
    "def linear_model(x):\n",
    "    return x * w + b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_ = linear_model(x_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "经过上面的步骤我们就定义好了模型，在进行参数更新之前，我们可以先看看模型的输出结果长什么样"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11577da58>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEOgAEIj/B0G15NO6D6dTAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1155f3be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(x_train.data.numpy(), y_train.data.numpy(), 'bo', label='real')\n",
    "plt.plot(x_train.data.numpy(), y_.data.numpy(), 'ro', label='estimated')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**思考：红色的点表示预测值，似乎排列成一条直线，请思考一下这些点是否在一条直线上？**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "这个时候需要计算我们的误差函数，也就是\n",
    "\n",
    "$$\n",
    "\\frac{1}{n} \\sum_{i=1}^n(\\hat{y}_i - y_i)^2\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 计算误差\n",
    "def get_loss(y_, y):\n",
    "    return torch.mean((y_ - y_train) ** 2)\n",
    "\n",
    "loss = get_loss(y_, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " 153.3520\n",
      "[torch.FloatTensor of size 1]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# 打印一下看看 loss 的大小\n",
    "print(loss)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "定义好了误差函数，接下来我们需要计算 w 和 b 的梯度了，这时得益于 PyTorch 的自动求导，我们不需要手动去算梯度，有兴趣的同学可以手动计算一下，w 和 b 的梯度分别是\n",
    "\n",
    "$$\n",
    "\\frac{\\partial}{\\partial w} = \\frac{2}{n} \\sum_{i=1}^n x_i(w x_i + b - y_i) \\\\\n",
    "\\frac{\\partial}{\\partial b} = \\frac{2}{n} \\sum_{i=1}^n (w x_i + b - y_i)\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 自动求导\n",
    "loss.backward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " 161.0043\n",
      "[torch.FloatTensor of size 1]\n",
      "\n",
      "Variable containing:\n",
      " 22.8730\n",
      "[torch.FloatTensor of size 1]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# 查看 w 和 b 的梯度\n",
    "print(w.grad)\n",
    "print(b.grad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 更新一次参数\n",
    "w.data = w.data - 1e-2 * w.grad.data\n",
    "b.data = b.data - 1e-2 * b.grad.data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "更新完成参数之后，我们再一次看看模型输出的结果"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11588b358>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW4AAAD8CAYAAABXe05zAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAFfJJREFUeJzt3X9w3HWdx/HXO2mgtOXwbDMIliR15uxZUhrKlinHiT1o\nsVJUGM4ZO1Gpdxi1lqung4PTP47DIo5z08rgDJIrikBAj6Ie0+E8BKr4gwO2NXDYSuuVpAbQxlYr\nbVraJu/7YzehDZvud5P97vf72X0+ZjLJbje77+5OX/1839/P5/M1dxcAIBx1SRcAACgNwQ0AgSG4\nASAwBDcABIbgBoDAENwAEBiCGwACQ3ADQGAIbgAIzKQ4nnTGjBne0tISx1MDQFXasmXLH9y9Mcpj\nYwnulpYWZbPZOJ4aAKqSmfVGfSytEgAITNHgNrPZZtZ93NefzeyzlSgOAPBmRVsl7v6ipDZJMrN6\nSS9L+n7MdQEAxlBqj/sySf/n7pF7McOOHj2qvr4+HT58uNRfxRgmT56smTNnqqGhIelSAFRQqcH9\nYUkPFPoDM+uQ1CFJTU1Nb/rzvr4+nX766WppaZGZlVonRnF37d27V319fZo1a1bS5QCooMgnJ83s\nFEkfkPRgoT939053z7h7prHxzTNaDh8+rOnTpxPaZWJmmj59OkcwQBp0dUktLVJdXe57V1esL1fK\niPt9kra6++/H+2KEdnnxfgIp0NUldXRIAwO52729uduS1N4ey0uWMh1wucZokwBAzVqz5o3QHjYw\nkLs/JpGC28ymSloi6XuxVZJyK1as0MaNG5MuA0Da7N5d2v1lECm43f2gu0939/2xVTJKnC0jd9fQ\n0FD5nhBA7SowGeOk95dBKldODreMensl9zdaRhMJ756eHs2ePVsf+9jH1NraqnvvvVcXXXSR5s+f\nrw996EM6cOCAJOnmm2/WggUL1Nraqo6ODrl7mf5WAKrSLbdIU6aceN+UKbn7Y5LK4I6rZbRz506t\nXLlSP/nJT3TXXXfpscce09atW5XJZLRu3TpJ0qpVq/Tss8/qhRde0KFDh7Rp06aJvSiA6tbeLnV2\nSs3Nklnue2dnbCcmpZg2mZqouFpGzc3NWrhwoTZt2qRt27bp4osvliQdOXJEF110kSRp8+bN+upX\nv6qBgQHt27dP5557rt7//vdP7IUBVLf29liDerRUBndTU649Uuj+iZg6daqkXI97yZIleuCBEyfJ\nHD58WCtXrlQ2m9U555yjm266iXnSAFInla2SuFtGCxcu1M9//nP95je/kSQdPHhQO3bsGAnpGTNm\n6MCBA8wiAZBKqRxxDx9xrFmTa480NeVCu1xHIo2Njbr77ru1fPlyvf7665KktWvX6p3vfKc+8YlP\nqLW1VW9729u0YMGC8rwgAJSRxTFrIpPJ+OgLKWzfvl3vete7yv5atY73FagOZrbF3TNRHpvKVgkA\nYGwENwAEhuAGgMAQ3AAQGIIbAAJDcANAYAjuAu6++2698sorI7evu+46bdu2bcLP29PTo/vvv7/k\n32NLWQDHS29wV/hSQMcbHdwbNmzQnDlzJvy84w1uADheOoM7jn1dJd1333268MIL1dbWpk9+8pMa\nHBzUihUr1Nraqrlz52r9+vXauHGjstms2tvb1dbWpkOHDmnRokUaXlA0bdo03XDDDTr33HO1ePFi\nPfPMM1q0aJHe8Y536OGHH5aUC+h3v/vdmj9/vubPn69f/OIXkqQbb7xRP/3pT9XW1qb169drcHBQ\nN9xwgxYsWKDzzjtPd955p6TcXiqrVq3S7NmztXjxYu3Zs2dCf28AVcbdy/51wQUX+Gjbtm17031j\nam52z0X2iV/NzdGfo8DrX3nllX7kyBF3d//0pz/tN910ky9evHjkMX/84x/d3f0973mPP/vssyP3\nH39bkj/yyCPu7n7VVVf5kiVL/MiRI97d3e3z5s1zd/eDBw/6oUOH3N19x44dPvx+bN682ZctWzby\nvHfeead/6Utfcnf3w4cP+wUXXOC7du3yhx56yBcvXuzHjh3zl19+2c844wx/8MEHx/x7AQifpKxH\nzNhU7lUSx76ujz/+uLZs2TKy/8ihQ4e0dOlS7dq1S9dff72WLVumyy+/vOjznHLKKVq6dKkkae7c\nuTr11FPV0NCguXPnqqenR5J09OhRrVq1St3d3aqvr9eOHTsKPtejjz6q559/fqR/vX//fu3cuVNP\nPvmkli9frvr6ep199tm69NJLx/33BlB90tkqieFSQO6ua6+9Vt3d3eru7taLL76o2267Tc8995wW\nLVqkb3zjG7ruuuuKPk9DQ8PI1dXr6up06qmnjvx87NgxSdL69et15pln6rnnnlM2m9WRI0fGrOn2\n228fqemll16K9J8HUPMSPAeWBukM7hj2db3sssu0cePGkX7xvn371Nvbq6GhIV1zzTVau3attm7d\nKkk6/fTT9dprr437tfbv36+zzjpLdXV1uvfeezU4OFjwed/73vfqjjvu0NGjRyVJO3bs0MGDB3XJ\nJZfou9/9rgYHB/Xqq69q8+bN464FqDoxnQMLSTpbJTHs6zpnzhytXbtWl19+uYaGhtTQ0KB169bp\n6quvHrlw8K233iopN/3uU5/6lE477TQ99dRTJb/WypUrdc011+iee+7R0qVLRy7gcN5556m+vl7z\n5s3TihUrtHr1avX09Gj+/PlydzU2NuoHP/iBrr76aj3xxBOaM2eOmpqaRq7OA0Anv7ZhBa9CkyS2\ndQ0c7ytqTl1dbqQ9mpmUH4SFiG1dAVSvGM6BhYbgBhCWuK9tGICKBnccbZlaxvuJmtTeLnV2Ss3N\nufZIc3Pudo30t6WIJyfN7C2SNkhqleSS/sHdSzprN3nyZO3du1fTp08fmU6H8XN37d27V5MnT066\nFKDy2ttrKqhHizqr5DZJP3T3vzezUyRNKfYLo82cOVN9fX3q7+8v9VcxhsmTJ2vmzJlJlwGgwooG\nt5mdIekSSSskyd2PSCq8ouQkGhoaNGvWrFJ/DQAwSpQe9yxJ/ZK+ZWa/NLMNZjY15roAAGOIEtyT\nJM2XdIe7ny/poKQbRz/IzDrMLGtmWdohABCfKMHdJ6nP3Z/O396oXJCfwN073T3j7pnGxsZy1ggA\nOE7R4Hb330n6rZnNzt91maSJXw4GADAuUWeVXC+pKz+jZJekj8dXEgDgZCIFt7t3S4q0hh4AEC+W\nvANAYAhuAAgMwQ0AgSG4ASAwBDcABIbgBoDAENwAEBiCGwACQ3ADQGAIbgAIDMENAIEhuAEgMAQ3\nAASG4AaAwBDcABAYghsAAkNwA0BgCG4ACAzBDQCBIbgBIDAENwAEhuAGgMAQ3AAQGIIbAAJDcANA\nYAhuIGRdXVJLi1RXl/ve1ZV0RaiASVEeZGY9kl6TNCjpmLtn4iwKQARdXVJHhzQwkLvd25u7LUnt\n7cnVhdiVMuL+O3dvI7SBlFiz5o3QHjYwkLsfVY1WCRCq3btLux9VI2pwu6THzGyLmXXEWRCAiJqa\nSrsfVSNqcP+tu7dJep+kz5jZJaMfYGYdZpY1s2x/f39ZiwRQwC23SFOmnHjflCm5+1HVIgW3u7+c\n/75H0vclXVjgMZ3unnH3TGNjY3mrBPBm7e1SZ6fU3CyZ5b53dnJisgYUnVViZlMl1bn7a/mfL5d0\nc+yVASiuvZ2grkFRpgOeKen7Zjb8+Pvd/YexVgUAGFPR4Hb3XZLmVaAWAEAETAcEksbqR5Qo0spJ\nADFh9SPGgRE3kCRWP2IcCG4gSax+xDgQ3ECSWP2IcSC4gSSx+hHjQHADSWL1I8aBWSVA0lj9iBIx\n4gaAwBDcABAYghsAAkNwA0BgCG5gPNhfBAliVglQKvYXQcIYcQOlYn8RJIzgBqI4vjXS21v4Mewv\nggqhVQIUM7o1Mhb2F0GFMOIGiinUGhmN/UVQQQQ3UMzJWiDsL4IE0CoBimlqKtzXbm6WenoqXg7A\niBuQTj4vm61XkTIENzB88rG3V3J/Y172cHiz9SpSxty97E+ayWQ8m82W/XmBWLS00ApB4sxsi7tn\nojyWETfAdR8RGIIb4LqPCAzBDXDyEYGJHNxmVm9mvzSzTXEWBFQcJx8RmFLmca+WtF3SX8RUC5Ac\nrvuIgEQacZvZTEnLJG2ItxwAQDFRWyVfk/QFSUMx1gIAiKBocJvZlZL2uPuWIo/rMLOsmWX7+/vL\nViAA4ERRRtwXS/qAmfVI+o6kS83svtEPcvdOd8+4e6axsbHMZQIAhhUNbnf/orvPdPcWSR+W9IS7\nfyT2yoBiuO4jahS7AyJMXPcRNYy9ShAm9hdBlWGvElQ/9hdBDSO4ESb2F0ENI7gRJvYXQQ0juBEm\n9hdBDWNWCcLF/iKoUYy4ASAwBDcABIbgBoDAENwAEBiCGwACQ3ADwARVer8zpgMCwAQksd8ZI24A\nmIA1a94I7WEDA7n740JwIz7sl40akMR+ZwQ34jF8/NjbK7m/cfxIeKMM0jQmSGK/M4Ib8Uji+BE1\nIW1jgiT2OyO4EQ/2y0ZM0jYmSGK/M4IbE1fouJX9shGTNI4J2ttzF14aGsp9j3vvM4IbEzPWcesV\nV7BfNmLBmIDgxkSNddz6yCPsl41YcA0NghsTdbLj1kofP05AmmYp4OS4hgYrJzFRTU2Fr7Ye0HFr\nEivfMDG1fg0NRtyYmCo4bk3bLAWgGIIbE1MFx61pnKUAnAytEkxc4MetVdDtQY1hxI2aVwXdHtSY\nosFtZpPN7Bkze87MfmVm/1qJwoBKqYJuD2pMlFbJ65IudfcDZtYg6Wdm9l/u/j8x1wZUTODdHtSY\nosHt7i7pQP5mQ/7L4ywKADC2SD1uM6s3s25JeyT9yN2fLvCYDjPLmlm2v7+/3HUCAPIiBbe7D7p7\nm6SZki40s9YCj+l094y7ZxobG8tdJ4ACWPFZm0qaVeLuf5K0WdLSeMoBEFXa9qVG5USZVdJoZm/J\n/3yapCWSfh13YUCapWGky4rP2hVlVslZkr5tZvXKBf1/uPumeMsC0iste5uw4rN2WW7SSHllMhnP\nZrNlf14gDVpaCq+0bG7ObYJYa3WgPMxsi7tnojyWlZNVIA2H7bUkLSNdVnzWLoI7cJygqry0XIGF\nFZ+1i1ZJ4DhcrrzRPW4pN9IlNDERtEpqyO7d0nJ16SW1aFB1ekktWq4uTlDFiJEuksa2roFb9dYu\n3bq3Q1OVG/61qFf/rg7NeKskkSRxYW8TJIkRd+C+rDUjoT1sqgb0ZVX/ZF5OyqJWMeIO3LR9hXsi\nY91fLdIylxpIAiPu0KVlikOFsWoQtYzgTrti/YAancyblrnU5UDLB6UiuNMsyiTtGp3iUC0HGszD\nx3gwjzvNmKQ9pmqZS81HjGHM464W1dQPKLNqOdDgI8Z4MKskzZqaCg/HQusHxKQa5lLzEWM8GHGn\nWY2eeKwlfMQYD4I7zaqlH4Ax8RFjPDg5CQApwMlJAKhiBDcABIbgBoDAENwAEBiCGwACQ3ADQGAI\nbgAIDMGN2LBdKRAP9ipBLLhCDRCfoiNuMzvHzDab2TYz+5WZra5EYQgbV6gB4hNlxH1M0ufdfauZ\nnS5pi5n9yN23xVwbAsZ2pUB8io643f1Vd9+a//k1SdslvT3uwhC2arlCDZBGJZ2cNLMWSedLejqO\nYlKNM20lYbtSID6Rg9vMpkl6SNJn3f3PBf68w8yyZpbt7+8vZ43J48KAJWO7UiA+kbZ1NbMGSZsk\n/be7ryv2+Krb1pULAwKIWVm3dTUzk3SXpO1RQrsqcaYNQIpEaZVcLOmjki41s+781xUx15UunGkD\nkCJRZpX8zN3N3c9z97b81yOVKK7ixjoByZk2ACnCyslhUZb6rVmTa480NeVCmzNtABLANSeHcQIS\nQIK45uR4cAISQCAI7mFFTkCy/qa28fkjTQjuYSc5Acn6m5Or9lDj80fa0OM+XldXwROQtL/HNvqc\nrpT7/66aVkny+aMSSulxE9wR1NXlRlqjmUlDQ5WvJ01qIdT4/FEJnJwssyTX36S9DVEL53RZf4W0\nIbgjSGr9TQi91VoINdZfIW0I7giS2ukuhKvI1EKosdMh0oYed4qF0lsd45wugBKU0uNmyXuKNTUV\nPvGXtjZEeztBDVQSrZIUq4U2BIDSEdwpRm8VQCG0SlKONgSA0RhxA0BgCO7jpH2xCwBItEpGRLmO\nAgCkASPuvBAWuwCAlKLgTrpNUQt7bgCoDqkI7q4u6bGPd+nHvS065nX6cW+LHvt4V0XDuxb23ABQ\nHVIR3E+v7tLXj3aoRb2qk6tFvfr60Q49vbpyyc1iFwChSEVwf27vGk3ViQ3mqRrQ5/ZWrsHMYhcA\noUjFJlNDVqc6vbmOIZnqPEW7KQFATIK7kMLA9MKN5LHuB4BalorgnnbbLTp2yokN5mOnTNG022gw\nA8BoRYPbzL5pZnvM7IXYqmhv16RvnthgnvRNGswAUEjRHreZXSLpgKR73L01ypNyIQUAKE1Ze9zu\n/qSkfROuCgBQFqnocQMAoitbcJtZh5llzSzb399frqcFAIxStuB29053z7h7prGxsVxPCwAYhVYJ\nAAQmyqySByQtkjRD0u8l/Yu731Xkd/olFbg++QlmSPpD5EprB+/L2HhvxsZ7U1hI70uzu0dqV8Sy\n5D3SC5tlo059qSW8L2PjvRkb701h1fq+0CoBgMAQ3AAQmCSDuzPB104z3pex8d6MjfemsKp8XxLr\ncQMAxodWCQAEpqLBbWbnmNlmM9tmZr8ys9WVfP0QmFm9mf3SzDYlXUuamNlbzGyjmf3azLab2UVJ\n15QGZvbP+X9LL5jZA2Y2OemaklJoJ1Mze6uZ/cjMdua//2WSNZZLpUfcxyR93t3nSFoo6TNmNqfC\nNaTdaknbky4ihW6T9EN3/2tJ88R7JDN7u6R/kpTJ79xZL+nDyVaVqLslLR11342SHnf3v5L0eP52\n8Coa3O7+qrtvzf/8mnL/+N5eyRrSzMxmSlomaUPStaSJmZ0h6RJJd0mSux9x9z8lW1VqTJJ0mplN\nkjRF0isJ15OYMXYy/aCkb+d//rakqypaVEwS63GbWYuk8yU9nVQNKfQ1SV+QxIU2TzRLUr+kb+Xb\nSBvMbGrSRSXN3V+W9G+Sdkt6VdJ+d3802apS50x3fzX/8+8knZlkMeWSSHCb2TRJD0n6rLv/OYka\n0sbMrpS0x923JF1LCk2SNF/SHe5+vqSDqpJD3onI92s/qNx/bGdLmmpmH0m2qvTy3BS6qphGV/Hg\nNrMG5UK7y92/V+nXT7GLJX3AzHokfUfSpWZ2X7IlpUafpD53Hz4626hckNe6xZJecvd+dz8q6XuS\n/ibhmtLm92Z2liTlv+9JuJ6yqPSsElOuT7nd3ddV8rXTzt2/6O4z3b1FuRNMT7g7oydJ7v47Sb81\ns9n5uy6TtC3BktJit6SFZjYl/2/rMnHSdrSHJV2b//laSf+ZYC1lU+kR98WSPqrcaLI7/3VFhWtA\nmK6X1GVmz0tqk/TlhOtJXP4IZKOkrZL+V7l/z1W5UjCK/E6mT0mabWZ9ZvaPkr4iaYmZ7VTuCOUr\nSdZYLqycBIDAsHISAAJDcANAYAhuAAgMwQ0AgSG4ASAwBDcABIbgBoDAENwAEJj/B4eTq//LFqBm\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1156dc6d8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_ = linear_model(x_train)\n",
    "plt.plot(x_train.data.numpy(), y_train.data.numpy(), 'bo', label='real')\n",
    "plt.plot(x_train.data.numpy(), y_.data.numpy(), 'ro', label='estimated')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "从上面的例子可以看到，更新之后红色的线跑到了蓝色的线下面，没有特别好的拟合蓝色的真实值，所以我们需要在进行几次更新"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 0, loss: 3.1357719898223877\n",
      "epoch: 1, loss: 0.3550889194011688\n",
      "epoch: 2, loss: 0.30295443534851074\n",
      "epoch: 3, loss: 0.30131956934928894\n",
      "epoch: 4, loss: 0.3006229102611542\n",
      "epoch: 5, loss: 0.29994693398475647\n",
      "epoch: 6, loss: 0.299274742603302\n",
      "epoch: 7, loss: 0.2986060082912445\n",
      "epoch: 8, loss: 0.2979407012462616\n",
      "epoch: 9, loss: 0.29727882146835327\n"
     ]
    }
   ],
   "source": [
    "for e in range(10): # 进行 10 次更新\n",
    "    y_ = linear_model(x_train)\n",
    "    loss = get_loss(y_, y_train)\n",
    "    \n",
    "    w.grad.zero_() # 记得归零梯度\n",
    "    b.grad.zero_() # 记得归零梯度\n",
    "    loss.backward()\n",
    "    \n",
    "    w.data = w.data - 1e-2 * w.grad.data # 更新 w\n",
    "    b.data = b.data - 1e-2 * b.grad.data # 更新 b \n",
    "    print('epoch: {}, loss: {}'.format(e, loss.data[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11598e0f0>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUdOFjLYuAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1158466d8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_ = linear_model(x_train)\n",
    "plt.plot(x_train.data.numpy(), y_train.data.numpy(), 'bo', label='real')\n",
    "plt.plot(x_train.data.numpy(), y_.data.numpy(), 'ro', label='estimated')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "经过 10 次更新，我们发现红色的预测结果已经比较好的拟合了蓝色的真实值。\n",
    "\n",
    "现在你已经学会了你的第一个机器学习模型了，再接再厉，完成下面的小练习。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**小练习：**\n",
    "\n",
    "重启 notebook 运行上面的线性回归模型，但是改变训练次数以及不同的学习率进行尝试得到不同的结果"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 多项式回归模型"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "下面我们更进一步，讲一讲多项式回归。什么是多项式回归呢？非常简单，根据上面的线性回归模型\n",
    "\n",
    "$$\n",
    "\\hat{y} = w x + b\n",
    "$$\n",
    "\n",
    "这里是关于 x 的一个一次多项式，这个模型比较简单，没有办法拟合比较复杂的模型，所以我们可以使用更高次的模型，比如\n",
    "\n",
    "$$\n",
    "\\hat{y} = w_0 + w_1 x + w_2 x^2 + w_3 x^3 + \\cdots\n",
    "$$\n",
    "\n",
    "这样就能够拟合更加复杂的模型，这就是多项式模型，这里使用了 x 的更高次，同理还有多元回归模型，形式也是一样的，只是出了使用 x，还是更多的变量，比如 y、z 等等，同时他们的 loss 函数和简单的线性回归模型是一致的。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "首先我们可以先定义一个需要拟合的目标函数，这个函数是个三次的多项式"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "y = 0.90 + 0.50 * x + 3.00 * x^2 + 2.40 * x^3\n"
     ]
    }
   ],
   "source": [
    "# 定义一个多变量函数\n",
    "\n",
    "w_target = np.array([0.5, 3, 2.4]) # 定义参数\n",
    "b_target = np.array([0.9]) # 定义参数\n",
    "\n",
    "f_des = 'y = {:.2f} + {:.2f} * x + {:.2f} * x^2 + {:.2f} * x^3'.format(\n",
    "    b_target[0], w_target[0], w_target[1], w_target[2]) # 打印出函数的式子\n",
    "\n",
    "print(f_des)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "我们可以先画出这个多项式的图像"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1158f5e48>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIpIk/j8hjlpJApitKgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1158f5dd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 画出这个函数的曲线\n",
    "x_sample = np.arange(-3, 3.1, 0.1)\n",
    "y_sample = b_target[0] + w_target[0] * x_sample + w_target[1] * x_sample ** 2 + w_target[2] * x_sample ** 3\n",
    "\n",
    "plt.plot(x_sample, y_sample, label='real curve')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "接着我们可以构建数据集，需要 x 和 y，同时是一个三次多项式，所以我们取了 $x,\\ x^2, x^3$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 构建数据 x 和 y\n",
    "# x 是一个如下矩阵 [x, x^2, x^3]\n",
    "# y 是函数的结果 [y]\n",
    "\n",
    "x_train = np.stack([x_sample ** i for i in range(1, 4)], axis=1)\n",
    "x_train = torch.from_numpy(x_train).float() # 转换成 float tensor\n",
    "\n",
    "y_train = torch.from_numpy(y_sample).float().unsqueeze(1) # 转化成 float tensor "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "接着我们可以定义需要优化的参数，就是前面这个函数里面的 $w_i$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 定义参数和模型\n",
    "w = Variable(torch.randn(3, 1), requires_grad=True)\n",
    "b = Variable(torch.zeros(1), requires_grad=True)\n",
    "\n",
    "# 将 x 和 y 转换成 Variable\n",
    "x_train = Variable(x_train)\n",
    "y_train = Variable(y_train)\n",
    "\n",
    "def multi_linear(x):\n",
    "    return torch.mm(x, w) + b"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "我们可以画出没有更新之前的模型和真实的模型之间的对比"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x115b8cc50>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEr4xxsQIS/jGGBMj/h9fRUynBRhSVgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115846320>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 画出更新之前的模型\n",
    "y_pred = multi_linear(x_train)\n",
    "\n",
    "plt.plot(x_train.data.numpy()[:, 0], y_pred.data.numpy(), label='fitting curve', color='r')\n",
    "plt.plot(x_train.data.numpy()[:, 0], y_sample, label='real curve', color='b')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "可以发现，这两条曲线之间存在差异，我们计算一下他们之间的误差"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " 413.9843\n",
      "[torch.FloatTensor of size 1]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# 计算误差，这里的误差和一元的线性模型的误差是相同的，前面已经定义过了 get_loss\n",
    "loss = get_loss(y_pred, y_train)\n",
    "print(loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 自动求导\n",
    "loss.backward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Variable containing:\n",
      " -34.1391\n",
      "-146.6133\n",
      "-215.9148\n",
      "[torch.FloatTensor of size 3x1]\n",
      "\n",
      "Variable containing:\n",
      "-27.0838\n",
      "[torch.FloatTensor of size 1]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# 查看一下 w 和 b 的梯度\n",
    "print(w.grad)\n",
    "print(b.grad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 更新一下参数\n",
    "w.data = w.data - 0.001 * w.grad.data\n",
    "b.data = b.data - 0.001 * b.grad.data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1164c6d30>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/JAtQgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115bb9438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 画出更新一次之后的模型\n",
    "y_pred = multi_linear(x_train)\n",
    "\n",
    "plt.plot(x_train.data.numpy()[:, 0], y_pred.data.numpy(), label='fitting curve', color='r')\n",
    "plt.plot(x_train.data.numpy()[:, 0], y_sample, label='real curve', color='b')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "因为只更新了一次，所以两条曲线之间的差异仍然存在，我们进行 100 次迭代"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 20, Loss: 73.67840\n",
      "epoch 40, Loss: 17.97097\n",
      "epoch 60, Loss: 4.94101\n",
      "epoch 80, Loss: 1.87171\n",
      "epoch 100, Loss: 1.12812\n"
     ]
    }
   ],
   "source": [
    "# 进行 100 次参数更新\n",
    "for e in range(100):\n",
    "    y_pred = multi_linear(x_train)\n",
    "    loss = get_loss(y_pred, y_train)\n",
    "    \n",
    "    w.grad.data.zero_()\n",
    "    b.grad.data.zero_()\n",
    "    loss.backward()\n",
    "    \n",
    "    # 更新参数\n",
    "    w.data = w.data - 0.001 * w.grad.data\n",
    "    b.data = b.data - 0.001 * b.grad.data\n",
    "    if (e + 1) % 20 == 0:\n",
    "        print('epoch {}, Loss: {:.5f}'.format(e+1, loss.data[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "可以看到更新完成之后 loss 已经非常小了，我们画出更新之后的曲线对比"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1164e8278>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\neQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115ad6128>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 画出更新之后的结果\n",
    "y_pred = multi_linear(x_train)\n",
    "\n",
    "plt.plot(x_train.data.numpy()[:, 0], y_pred.data.numpy(), label='fitting curve', color='r')\n",
    "plt.plot(x_train.data.numpy()[:, 0], y_sample, label='real curve', color='b')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "可以看到，经过 100 次更新之后，可以看到拟合的线和真实的线已经完全重合了"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "**小练习：上面的例子是一个三次的多项式，尝试使用二次的多项式去拟合它，看看最后能做到多好**\n",
    "\n",
    "**提示：参数 `w = torch.randn(2, 1)`，同时重新构建 x 数据集**"
   ]
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}