Ultimate Secret of Cost Function. 😎
the secret of model’s accuracy. 📕
What is Cost Function ?
The 🔵 blue line on the left fits the data nicely for θ values 4 and 5x10⁵ compared to lines 📗green and 🔴red.
How can we know which values of theta(θ) will make our model perform best ?
In order to answer our above question we need some performance function which measures how good or bad our model is performing.
“Cost Function measures how bad our model is performing.”
💭How to find Cost Function ? 😄
📃Types of Cost Function:
Linear Regression :
Linear Regression Cost Function is given by distance between the models prediction and training samples. There are mainly two main types of regression cost function
→Mean Absolute Error (MAE) :
The sum of differences between actual and predicted values.
→Mean Squared Error (MSE) :
The sum of square of differences between actual and predicted values
Logistic Regression:
Logistic Regression’s Cost Function (Log Loss) tries to capture the value of θ for which the model predicts high probability for positive instances and is given by: