ShreyanshIITK20 / Regression-models

Regression models

Motive

This repository is created as a part of recuitment task for institue's programming club. It mainly focusses on basic understanding of machine learning and implementation of regression algorithm from scratch.

Task

To train regression models (linear and polynomial) on the basis of non-standard loss functions

| y - y_p |³

| y - y_p |

| y - y_p |⁷

| y - y_p |4

Linear regression

Original parameters for dataset :

Slope : 2.34

Intercept : 1.67

Gaussian noise included

1. Optimization by cubic loss function

Slope : 2.36

Intercept : 0.05

2. Optimization by linear loss function

Slope : 2.37

Intercept : 0.03

Polynomial regression

Original parameters for dataset :

Coefficient of x² : 2

Coefficient of x : 3

Constant term : 8

Gaussian noise included

1. Optimization by 4^th degree loss function

Coefficient of x² : 1.97

Coefficient of x : 0.02

Constant term : 0.0002

2. Optimization by 7^th degree loss function

Coefficient of x² : 2.13

Coefficient of x : 0.02

Constant term : 0.0002