mlr-gd 0.2.0

A package for multiple linear regression by gradient descent.

This is the dev branch, where development usually happens

Multiple linear regression by gradient descent.

Disclaimer:

This code is very early on and my first proper attempt to create a package so things may be a bit weird/not up to standard.

Installation

To install mlr-gd you can use pip:

$ python -m pip install mlr-gd

Alternatively, you can install it by cloning the GitHub repository:

$ git clone https://github.com/DrSolidDevil/mlr-gd.git
$ cd mlr-gd
$ pip install .

To import the package into your script:

import melar

Example

import numpy as np
import melar

# y = x1 + 0.5*x2
x = np.array([[1, 3, 5, 8], [1, 2, 3, 6]])
y = np.array([1.5, 4, 6.5, 11])

learning_rate = 0.01
generations = 100


model = melar.LinearRegression(weights_amount=2)
model.train(x, y, learning_rate, generations, do_print=True)
print(f"Weights: {model.weights}, Bias: {model.bias}")

Gen: 0, Cost: 95.4852602406095
Gen: 1, Cost: 5.593624864417041
Gen: 2, Cost: 0.3286224504551768
Gen: 3, Cost: 0.020244781001893267
...
Gen: 99, Cost: 0.0007438760098695897
Training Complete
Weights: [0.94643617 0.57630021], Bias: -0.003265101149422934