Response 4

Convolutions

9x9 Matrix and 2 filters

7x7 Matrix Filter 1

7x7 Matrix Filter 1

What is the purpose of using a 3x3 filter to convolve across a 2D image matrix?

This is important because when you are using an intentional filter, then it will highlight certain aspects of the image matrix that allows for the computer to process the data which helps with modeling.

Why would we include more than one filter? How many filters did you assign as part of your architecture when training a model to learn images of numbers from the mnist dataset?

You might include more than one filter so that you can highlight different parts of the image. When training a model to learn handwriting, I think we used one filter.

Mean Squared Error

MSE: From your 400+ observations of homes for sale, calculate the MSE for the following.

The 10 biggest over-predictions

The Mean Squared Error is: 227420.48874569248

The 10 biggest under-predictions

The Mean Squared Error is: 2624678.97170306

The 10 most accurate results (use absolute value)

The Mean Squared Error is: 2624678.97170306

In which percentile do the 10 most accurate predictions reside? Did your model trend towards over or under predicting home values?

My model trended towards under predicting home values.

Which feature appears to be the most significant predictor in the above cases?

In the most accurate predictions, it appears that number of bedrooms was the feature that correlates with the biggest prediction. The homes that were the most over predicted were locations with a lot of bathrooms, especially those with a high bathroom to bedroom ratio. The most underpredicted houses had very small square feet.