{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Numpy\n",
    "\n",
    "## Numpy特征和导入\n",
    "\n",
    "1. 用于多维数组的第三方Python包\n",
    "2. 更接近于底层和硬件 (高效)\n",
    "3. 专注于科学计算 (方便)\n",
    "4. 导入包：`import numpy as np`\n",
    "\n",
    "## list转为数组\n",
    "\n",
    "1. `a = np.array([0,1,2,3])`\n",
    "2. 输出为：`[0 1 2 3]`\n",
    "3. 数据类型：`<type 'numpy.ndarray'>`\n",
    "\n",
    "## 一维数组\n",
    "\n",
    "1. a = np.array([1,2,3,4])属性\n",
    "  `a.ndim`–>维度为1\n",
    "  `a.shape`–>形状，返回(4,)\n",
    "  `len(a)`–>长度，4\n",
    "2. 访问数组`a[1:5:2]`下标1-5，下标关系+2\n",
    "3. 逆序`a[::-1]`\n",
    "\n",
    "## 多维数组\n",
    "\n",
    "1. 二维：`a = np.array([[0,1,2,3],[1,2,3,4]])`\n",
    "  输出为：\n",
    "  ```\n",
    " [[0 1 2 3]\n",
    "  [1 2 3 4]]\n",
    "  ```\n",
    "  `a.ndm` –>2\n",
    "  `a.shape` –>(2,4)–>行数，列数\n",
    "  `len(a)` –>2–>第一维大小\n",
    "\n",
    "2. 三维：`a = np.array([[[0],[1]],[[2],[4]]])`\n",
    "  `a.shape`–>(2,2,1)\n",
    "\n",
    "## 用函数创建数组\n",
    "\n",
    "1. np.arange()`\n",
    "    ```\n",
    "    a = np.arange(0, 10)\n",
    "    b = np.arange(10)\n",
    "    c = np.arange(0,10,2)\n",
    "    ```\n",
    "    输出：\n",
    "    ```\n",
    "    [0 1 2 3 4 5 6 7 8 9]\n",
    "    [0 1 2 3 4 5 6 7 8 9]\n",
    "    [0 2 4 6 8]\n",
    "    ```\n",
    "2. `np.linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None)`\n",
    "    等距离产生num个数\n",
    "3. `np.logspace(start, stop, num=50, endpoint=True, base=10.0, dtype=None)`\n",
    "    以log函数取\n",
    "\n",
    "## 常用数组\n",
    "\n",
    "1. `a = np.ones((3,3))`\n",
    "    输出：\n",
    "    ```\n",
    "    [[ 1. 1. 1.]\n",
    "    [ 1. 1. 1.]\n",
    "    [ 1. 1. 1.]]\n",
    "    ```\n",
    "2. `np.zeros((3,3))`\n",
    "3. `np.eye(2)`单位矩阵\n",
    "4. `np.diag([1,2,3],k=0)`对角矩阵，k为对角线的偏移\n",
    "\n",
    "## 随机数矩阵\n",
    "\n",
    "1. `a = np.random.rand(4)`\n",
    "    输出：`[ 0.99890402 0.41171695 0.40725671 0.42501804]`范围在`[0,1]`之间\n",
    "2. `a = np.random.randn(4) `Gaussian函数，\n",
    "3. 生成100个0-m的随机数: `[t for t in [np.random.randint(x-x, m) for x in range(100)]]` \n",
    "    也可以\n",
    "    ```\n",
    "    m_arr = np.arange(0,m)      # 生成0-m-1\n",
    "    np.random.shuffle(m_arr)    # 打乱m_arr顺序\n",
    "    ```\n",
    "    然后取前100个即可\n",
    "\n",
    "## 查看数据类型\n",
    "\n",
    "1. `a.dtype`\n",
    "\n",
    "## 数组复制\n",
    "\n",
    "1. 共享内存\n",
    "    ```\n",
    "    a = np.array([1,2,3,4,5])\n",
    "    b = a\n",
    "    print np.may_share_memory(a,b)\n",
    "    ```\n",
    "    输出：`True`\n",
    "    说明使用的同一个存储区域，修改一个数组同时另外的也会修改\n",
    "\n",
    "2. 不共享内存\n",
    "    ```\n",
    "    b = a.copy()\n",
    "    ```\n",
    "## 布尔型\n",
    "\n",
    "```\n",
    "a = np.random.random_integers(0,20,5)\n",
    "print a\n",
    "print a%3==0\n",
    "print a[a % 3 == 0]\n",
    "```\n",
    "输出：\n",
    "```\n",
    "[14 3 6 15 4]\n",
    "[False True True True False]\n",
    "[ 3 6 15]\n",
    "```\n",
    "## 中间数、平均值\n",
    "\n",
    "1. 中间数np.median(a)\n",
    "2. 平均值np.mean(a),\n",
    "    若是矩阵，不指定axis默认求所有元素的均值\n",
    "    axis=0,求列的均值\n",
    "    axis=1，求行的均值\n",
    "    \n",
    "## 矩阵操作\n",
    "\n",
    "1. 乘积`np.dot(a,b)`\n",
    "```\n",
    "a = np.array([[1,2,3],[2,3,4]])\n",
    "b = np.array([[1,2],[2,3],[2,2]])\n",
    "print np.dot(a,b)\n",
    "```\n",
    "\n",
    "或者使用`np.matrix()`生成矩阵，相乘需要满足矩阵相乘的条件\n",
    "\n",
    "2. 内积`np.inner(a,b)`\n",
    "  行相乘\n",
    "3. 逆矩阵`np.linalg.inv(a)`\n",
    "4. 列的最大值`np.max(a[:,0])`–>返回第一列的最大值\n",
    "5. 每列的和`np.sum(a,0)`\n",
    "6. 每行的平均数`np.mean(a,1)`\n",
    "7. 求交集`p.intersect1d(a,b)`，返回一维数组\n",
    "8. 转置：`np.transpose(a)`\n",
    "9. 两个矩阵对应对应元素相乘（点乘）：`a*b`\n",
    "\n",
    "## 文件操作\n",
    "\n",
    "1. 保存：tofile()\n",
    "    ```\n",
    "    a = np.arange(10)\n",
    "    a.shape=2,5\n",
    "    a.tofile(\"test.bin\")\n",
    "    ```\n",
    "    读取：（需要注意指定保存的数据类型）\n",
    "    ```\n",
    "    a = np.fromfile(\"test.bin\",dtype=np.int32)\n",
    "    print a\n",
    "    ```\n",
    "2. 保存：`np.save(\"test\",a)`–>会保存成test.npy文件\n",
    "    读取：`a = np.load(\"test\")`\n",
    "    \n",
    "## 组合两个数组\n",
    "\n",
    "1. 垂直组合\n",
    "```\n",
    "a = np.array([1,2,3])\n",
    "b = np.array([[1,2,3],[4,5,6]])\n",
    "c = np.vstack((b,a))\n",
    "```\n",
    "2. 水平组合\n",
    "```\n",
    "a = np.array([[1,2],[3,4]])\n",
    "b = np.array([[1,2,3],[4,5,6]])\n",
    "c = np.hstack((a,b))\n",
    "```\n",
    "## 读声音Wave文件\n",
    "\n",
    "1. wave\n",
    "```\n",
    "import wave\n",
    "from matplotlib import pyplot as plt\n",
    "import numpy as np\n",
    "# 打开WAV文档\n",
    "f = wave.open(r\"c:\\WINDOWS\\Media\\ding.wav\", \"rb\")\n",
    "# 读取格式信息\n",
    "# (nchannels, sampwidth, framerate, nframes, comptype, compname)\n",
    "params = f.getparams()\n",
    "nchannels, sampwidth, framerate, nframes = params[:4]\n",
    "# 读取波形数据\n",
    "str_data = f.readframes(nframes)\n",
    "f.close()\n",
    "#将波形数据转换为数组\n",
    "wave_data = np.fromstring(str_data, dtype=np.short)\n",
    "wave_data.shape = -1, 2\n",
    "wave_data = wave_data.T\n",
    "time = np.arange(0, nframes) * (1.0 / framerate)\n",
    "# 绘制波形\n",
    "plt.subplot(211) \n",
    "plt.plot(time, wave_data[0])\n",
    "plt.subplot(212) \n",
    "plt.plot(time, wave_data[1], c=\"g\")\n",
    "plt.xlabel(\"time (seconds)\")\n",
    "plt.show()\n",
    "```\n",
    "## where\n",
    "\n",
    "找到y数组中=1的位置：np.where(y==1)\n",
    "\n",
    "## np.ravel(y)\n",
    "\n",
    "将二维的转化为一维的，eg:(5000,1)-->(5000,)\n",
    "\n",
    "## ndarray.flat函数\n",
    "\n",
    "将数据展开对应的数组，可以进行访问\n",
    "\n",
    "应用：0/1映射\n",
    "```\n",
    "def dense_to_one_hot(label_dense,num_classes):\n",
    "    num_labels = label_dense.shape[0]\n",
    "    index_offset = np.arange(num_labels)*num_classes\n",
    "    labels_one_hot = numpy.zeros((num_labels, num_classes))\n",
    "    labels_one_hot.flat[index_offset + labels_dense.ravel()] = 1\n",
    "    return labels_one_hot\n",
    "```\n",
    "## 数组访问\n",
    "```\n",
    "X = np.array([[1,2],[3,4]])\n",
    "```\n",
    "`X[0:1]`和`X[0:1,:]`等价，都是系那是第一行数据\n",
    "\n",
    "## `np.c_()`\n",
    "\n",
    "按照第二维度，即列拼接数据\n",
    "\n",
    "`np.c_[np.array([[1,2,3]]), 0, 0, np.array([[4,5,6]])]`\n",
    "\n",
    "输出：`array([[1, 2, 3, 0, 0, 4, 5, 6]])`\n",
    "\n",
    "两个列表list拼接，长度要一致\n",
    "\n",
    "`np.c_[[1,2,3],[2,3,4]]`\n",
    "\n",
    "`np.c_[range(1,5),range(2,6)]`"
   ]
  }
 ],
 "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"
  },
  "toc": {
   "colors": {
    "hover_highlight": "#DAA520",
    "navigate_num": "#000000",
    "navigate_text": "#333333",
    "running_highlight": "#FF0000",
    "selected_highlight": "#FFD700",
    "sidebar_border": "#EEEEEE",
    "wrapper_background": "#FFFFFF"
   },
   "moveMenuLeft": true,
   "nav_menu": {
    "height": "390px",
    "width": "252px"
   },
   "navigate_menu": true,
   "number_sections": true,
   "sideBar": true,
   "threshold": 4,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": false,
   "widenNotebook": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}