notes

Personal notes
git clone git://git.laack.co/notes.git
Log | Files | Refs

GradientDescentCode.md (1263B)

      1 # Gradient Descent Implementation 
      2 
      3 This approach implements a [LearningRate](LearningRate.md) parameter to narrow in upon a local minimum of the given third degree polynomial.
      4 
      5 ## Code
      6 ```python
      7 import sys
      8 import random
      9 
     10 RECURSION_LIMIT = 1500
     11 
     12 sys.setrecursionlimit(RECURSION_LIMIT)
     13 print("ax^3 + bx^2 + cx + d")
     14 a = float(input("Enter a: "))
     15 b = float(input("Enter b: "))
     16 c = float(input("Enter c: "))
     17 d = float(input("Enter d: "))
     18 learningRate = float(input("Learning Rate: "))
     19 
     20 def calculateValue(x):
     21     return x**3 * a + x**2 * b + x * c + d
     22 
     23 def printResult(x, y):
     24     print("x: " + str(x) + "\ty: " + str(y))
     25 
     26 def limit(x):
     27     rightX = x + .00000001
     28     leftX = x - .00000001
     29 
     30     rightY = calculateValue(rightX)
     31     leftY = calculateValue(leftX)
     32 
     33     return ((rightY - leftY) / .00000002)
     34 
     35 def descend(x, depth):
     36     # Need - 15 because recursion includes other function calls...
     37     if depth >= RECURSION_LIMIT - 15:
     38         return x
     39     lim = limit(x)
     40     printResult(x, calculateValue(x))
     41     depth += 1
     42 
     43     if lim > 0:
     44         return descend(x - learningRate * lim, depth)
     45     else:
     46         return descend(x + learningRate * lim, depth)
     47 
     48 currSearch = random.random() * 10
     49 xVal = descend(currSearch, 0)
     50 
     51 printResult(xVal, calculateValue(xVal))
     52 ```