Introducing Alpha

Alluxia-F · Jul 8, 2017 · d4b0de5 · d4b0de5
1 parent 97bf757
commit d4b0de5
Showing 1 changed file with 144 additions and 0 deletions.
diff --git a/Chapter 4/8. Introducing Alpha.ipynb b/Chapter 4/8. Introducing Alpha.ipynb
@@ -0,0 +1,144 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "In fact, finding the appropriate alpha,even for state-of-the-art nerual networks,is often done simply by guessing.\n",
+    "\n",
+    "If learning is happening too slowly, then your alpha is too low and you increase it."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Iteration:0    Erroe: 0.04   Prediction: 1.0\n",
+      "Iteration:1    Erroe: 0.0144   Prediction: 0.92\n",
+      "Iteration:2    Erroe: 0.005184   Prediction: 0.872\n",
+      "Iteration:3    Erroe: 0.00186624   Prediction: 0.8432\n",
+      "Iteration:4    Erroe: 0.0006718464   Prediction: 0.82592\n",
+      "Iteration:5    Erroe: 0.000241864704   Prediction: 0.815552\n",
+      "Iteration:6    Erroe: 8.707129344e-05   Prediction: 0.8093312\n",
+      "Iteration:7    Erroe: 3.13456656384e-05   Prediction: 0.80559872\n",
+      "Iteration:8    Erroe: 1.12844396298e-05   Prediction: 0.803359232\n",
+      "Iteration:9    Erroe: 4.06239826674e-06   Prediction: 0.8020155392\n",
+      "Iteration:10    Erroe: 1.46246337603e-06   Prediction: 0.80120932352\n",
+      "Iteration:11    Erroe: 5.26486815369e-07   Prediction: 0.800725594112\n",
+      "Iteration:12    Erroe: 1.89535253533e-07   Prediction: 0.800435356467\n",
+      "Iteration:13    Erroe: 6.82326912719e-08   Prediction: 0.80026121388\n",
+      "Iteration:14    Erroe: 2.45637688579e-08   Prediction: 0.800156728328\n",
+      "Iteration:15    Erroe: 8.84295678884e-09   Prediction: 0.800094036997\n",
+      "Iteration:16    Erroe: 3.18346444398e-09   Prediction: 0.800056422198\n",
+      "Iteration:17    Erroe: 1.14604719983e-09   Prediction: 0.800033853319\n",
+      "Iteration:18    Erroe: 4.12576991939e-10   Prediction: 0.800020311991\n",
+      "Iteration:19    Erroe: 1.48527717099e-10   Prediction: 0.800012187195\n"
+     ]
+    }
+   ],
+   "source": [
+    "weight=0.5\n",
+    "goal_pred=0.8\n",
+    "inputs=2\n",
+    "alpha=0.1\n",
+    "\n",
+    "for iteration in range(20):\n",
+    "    prediction=inputs*weight\n",
+    "    error=(prediction-goal_pred)**2\n",
+    "    derivative=(prediction-goal_pred)*inputs\n",
+    "    weight-=alpha*derivative\n",
+    "    \n",
+    "    print 'Iteration:'+str(iteration)+'    Erroe: '+str(error)+'   Prediction: '+str(prediction)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "And despite all the crazy advancements of deep learning in the past few years, most people just try several orders of magnitude of alpha(10,1,0.1,0.01,0.001,0.0001)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Iteration:0    Erroe: 0.04   Prediction: 1.0\n",
+      "Iteration:1    Erroe: 0.036864   Prediction: 0.992\n",
+      "Iteration:2    Erroe: 0.0339738624   Prediction: 0.98432\n",
+      "Iteration:3    Erroe: 0.0313103115878   Prediction: 0.9769472\n",
+      "Iteration:4    Erroe: 0.0288555831594   Prediction: 0.969869312\n",
+      "Iteration:5    Erroe: 0.0265933054397   Prediction: 0.96307453952\n",
+      "Iteration:6    Erroe: 0.0245083902932   Prediction: 0.956551557939\n",
+      "Iteration:7    Erroe: 0.0225869324942   Prediction: 0.950289495622\n",
+      "Iteration:8    Erroe: 0.0208161169867   Prediction: 0.944277915797\n",
+      "Iteration:9    Erroe: 0.0191841334149   Prediction: 0.938506799165\n",
+      "Iteration:10    Erroe: 0.0176800973552   Prediction: 0.932966527198\n",
+      "Iteration:11    Erroe: 0.0162939777225   Prediction: 0.92764786611\n",
+      "Iteration:12    Erroe: 0.0150165298691   Prediction: 0.922541951466\n",
+      "Iteration:13    Erroe: 0.0138392339273   Prediction: 0.917640273407\n",
+      "Iteration:14    Erroe: 0.0127542379874   Prediction: 0.912934662471\n",
+      "Iteration:15    Erroe: 0.0117543057292   Prediction: 0.908417275972\n",
+      "Iteration:16    Erroe: 0.0108327681601   Prediction: 0.904080584933\n",
+      "Iteration:17    Erroe: 0.00998347913631   Prediction: 0.899917361536\n",
+      "Iteration:18    Erroe: 0.00920077437202   Prediction: 0.895920667075\n",
+      "Iteration:19    Erroe: 0.00847943366126   Prediction: 0.892083840392\n"
+     ]
+    }
+   ],
+   "source": [
+    "weight=0.5\n",
+    "goal_pred=0.8\n",
+    "inputs=2\n",
+    "alpha=0.01\n",
+    "\n",
+    "for iteration in range(20):\n",
+    "    prediction=inputs*weight\n",
+    "    error=(prediction-goal_pred)**2\n",
+    "    derivative=(prediction-goal_pred)*inputs\n",
+    "    weight-=alpha*derivative\n",
+    "    \n",
+    "    print 'Iteration:'+str(iteration)+'    Erroe: '+str(error)+'   Prediction: '+str(prediction)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "collapsed": true
+   },
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python [conda root]",
+   "language": "python",
+   "name": "conda-root-py"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 2
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython2",
+   "version": "2.7.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}