Logistic Regression with SymPy

logistic_regression_sympy.py

from sympy import *
import pandas as pd

m, b, i, n = symbols('m b i n')
x, y = symbols('x y', cls=Function)

joint_likelihood = Sum(log((1.0 / (1.0 + exp(-(b + m * x(i)))))**y(i) * (1.0 - (1.0 / (1.0 + exp(-(b + m * x(i))))))**(1-y(i))), (i, 0, n))

points = list(pd.read_csv("https://tinyurl.com/y2cocoo7").itertuples())

d_m = lambdify([m,b], diff(joint_likelihood, m).subs(n, len(points) - 1).doit() \
      .replace(x, lambda i: points[i].x) \
      .replace(y, lambda i: points[i].y))

d_b = lambdify([m,b], diff(joint_likelihood, b).subs(n, len(points) - 1).doit() \
      .replace(x, lambda i: points[i].x) \
      .replace(y, lambda i: points[i].y))

# Perform Gradient Descent 
m = 0.01
b = 0.01
L = .01

for j in range(10_000):
    m += d_m(m,b) * L
    b += d_b(m,b) * L

print(m, b)