In this lab **how** was **initial parameters** decided?

**How to calculate** these values?

This is ungraded lab.

optimal here would be the best fit for the model. So in this case they must have selected w and b based on the number of features for getting the desired housing price with the most usual features people would look for while purchasing a house.

Note this statement in the notebook

You will build a linear regression model using these values so you can then predict the price for other houses. For example, a house with 1200 sqft, 3 bedrooms, 1 floor, 40 years old.

If you notice in this image

they have selected the maximum to minimum square feet and varying price range with different number of bedroom to check variability and randomness in the analysis.

Based on these criteria they have choosen these initial parameters.

Hello @Debatreyo_Roy,

I don’t know exactly how the maker of this optional lab did that, but it is possible for them to just run gradient descent on the dataset once and get the optimal weights. Since it is a multiple regression problem, it is also possible to solve for the optimal weights by the normal equation.

However, I want to emphasize that usually we initialize the weights to a random set of values, and in a general neural network where normal equation doesn’t apply, we can only train it to find the optimal set of weights, so there is no tricks here - either someone did it or we do it.

Cheers,

Raymond