swap %pylab to explict imports and %matplotlib inline, add model coeffs

DrSkippy · Jun 5, 2015 · fa33560 · fa33560
1 parent 21008d4
commit fa33560
Showing 1 changed file with 35 additions and 16 deletions.
diff --git a/ml-basis-expansion-101/Basis Expansion.ipynb b/ml-basis-expansion-101/Basis Expansion.ipynb
@@ -1,7 +1,7 @@
 {
  "metadata": {
   "name": "",
-  "signature": "sha256:3549a4752602045c42ea70e7f5579fd904ef56de95d9ab983974fef1b10ea8f9"
+  "signature": "sha256:b8a3206d7df375d88d2ccbc5123abca03ea42d8513444b68a23f253f854f9dfb"
  },
  "nbformat": 3,
  "nbformat_minor": 0,
@@ -12,7 +12,9 @@
      "cell_type": "code",
      "collapsed": false,
      "input": [
-      "%pylab inline"
+      "import numpy as np\n",
+      "import matplotlib.pyplot as plt \n",
+      "%matplotlib inline"
      ],
      "language": "python",
      "metadata": {},
@@ -95,10 +97,9 @@
      "input": [
       "def f(n=100, sigma=1, mu=1):\n",
       "    # random normal vector\n",
-      "    return sigma * random.randn(n) + mu\n",
+      "    return sigma * np.random.randn(n) + mu\n",
       "    \n",
-      "plt.scatter(f(),f())\n",
-      "plt.show()"
+      "plt.scatter(f(),f())"
      ],
      "language": "python",
      "metadata": {},
@@ -235,8 +236,7 @@
       "plt.plot(z,sigmoid(z,2,3), color=\"cyan\", label=\"compress, shift\")\n",
       "plt.plot(z,sigmoid(z,-2), color=\"yellow\", label=\"shift\")\n",
       "plt.plot(z,sigmoid(z,0,0.5), color=\"green\", label=\"expand\")\n",
-      "plt.legend(loc=2)\n",
-      "plt.show()"
+      "plt.legend(loc=2)"
      ],
      "language": "python",
      "metadata": {},
@@ -332,7 +332,7 @@
       "\n",
       "gx, gy = xx.ravel(), yy.ravel()\n",
       "\n",
-      "Z = ext_m.decision_function(array([gx, gy, gx*gx, gy*gy, gx*gy]).transpose())\n",
+      "Z = ext_m.decision_function(np.array([gx, gy, gx*gx, gy*gy, gx*gy]).transpose())\n",
       "Z = Z.reshape(xx.shape)\n",
       "plt.imshow(Z, interpolation='nearest',\n",
       "           extent=(xx.min(), xx.max(), yy.min(), yy.max()), aspect='auto',\n",
@@ -374,8 +374,7 @@
       "def b1(x, pars):\n",
       "    return np.exp(-(x-pars[0])*(x-pars[0])/(pars[1]*pars[1]))\n",
       "\n",
-      "plt.plot(x, b1(x, (0,1)), color=\"green\")\n",
-      "plt.show()"
+      "plt.plot(x, b1(x, (0,1)), color=\"green\")"
      ],
      "language": "python",
      "metadata": {},
@@ -398,8 +397,7 @@
       "    return np.sum(c(x,pars))\n",
       "\n",
       "plt.plot(x,target(x)        , color=\"red\")\n",
-      "#plt.plot(x,  10*b1(x, (1,1)), color=\"green\")\n",
-      "plt.show()"
+      "#plt.plot(x,  10*b1(x, (1,1)), color=\"green\")"
      ],
      "language": "python",
      "metadata": {},
@@ -420,8 +418,7 @@
       "a = minimize(cost,[1,1,3])\n",
       "print(a.x)\n",
       "plt.plot(x,target(x), color=\"red\")\n",
-      "plt.plot(x, a.x[2]*b1(x, a.x), color=\"green\")\n",
-      "plt.show()"
+      "plt.plot(x, a.x[2]*b1(x, a.x), color=\"green\")"
      ],
      "language": "python",
      "metadata": {},
@@ -480,8 +477,7 @@
       "                    1,2,1])\n",
       "print(a.x)\n",
       "\n",
-      "plt.scatter(x, gx(x, a.x), color=\"green\")\n",
-      "plt.show()"
+      "plt.scatter(x, gx(x, a.x), color=\"green\")"
      ],
      "language": "python",
      "metadata": {},
@@ -628,6 +624,29 @@
      "metadata": {},
      "outputs": []
     },
+    {
+     "cell_type": "markdown",
+     "metadata": {},
+     "source": [
+      "Given this particular set of basis functions, we can also reach into the estimator above and find out what the relative weights were in fitting the model to this data:"
+     ]
+    },
+    {
+     "cell_type": "code",
+     "collapsed": false,
+     "input": [
+      "# nb: must be same order as the X array that went into the model fit (ext_X)\n",
+      "feature_names = ['v11', 'v_11', 'v1_1', 'v_1_1', 'v22', 'v_22', 'v2_2', 'v_2_2']\n",
+      "\n",
+      "x = np.arange(len(feature_names))\n",
+      "plt.bar(x, ext_m.coef_.ravel() )\n",
+      "_ = plt.xticks(x + 0.5, feature_names, rotation=30)\n",
+      "_ = plt.ylabel(\"model coefficients\")"
+     ],
+     "language": "python",
+     "metadata": {},
+     "outputs": []
+    },
     {
      "cell_type": "markdown",
      "metadata": {},