https://www.youtube.com/watch?v=4b4MUYve_U8&list=PLoROMvodv4rMiGQp3WXShtMGgzqpfVfbU&t=1s

In this Lecture, Andrew starts of with the most common algorithm known as Linear Regression. He here explains the idea behind linear regression is to make a model that fits the data (i.e line of best fit but with many parameters). However, to fit the line, the Linear Regression method helps the concepts of Gradient Descent. In this lecture, he gives examples of housing prices and parameters (also called weights), such as bedrooms, living rooms and more. The key takes away for me were:

• Linear Regression algorithms should be used for analysing linear relationships rather than relationships that require classification. Some examples include fuel prices with time, product prices with time and more.

Theta — parameters or Weights

h - hypothesis

Which can also be represented in terms of matrix (or called the Inception Term)

• Got introduced to the idea of what a cost function:

J - Cost function which is an Ordinary Least square

h - predicted answer for that sample

y - actual answer for that sample

Note: 1/2 is used for ease of differentiation.

• How the cost function is optimised using Gradient Descent.

alpha - learning rate.

j - represents the theta’s number

A little bit of differentiation to J(theta) :