In the Maroney video What is ML, he argues that the main difference between traditional programming and machine learning is that in traditional programming, you have rules and data that go in, and answers that come out. In ML, programmers instead feed in answers and data to the computer, which then figures out what the rules or patterns are, and gives them as an output.
In the first basic script from the Maroney video, when I change the function to predict the output for 7, the script predicts 22.002396. When I re-fit the model on the same data and run the same predict function again, the function returns 21.998444. There are slight differences in the values. I think the machine learning model produces slightly different predictions despite being given the same input because it is stochastic, and therefore includes elements of randomness in how the model is fit, meaning that the same network trained on the same data could nonetheless still produce different predictions.
The script I made to produce housing price has two different models, one that predicts based on number of bedrooms and one that predicts based on number of bathrooms. The bedrooms model seems to predict somewhat closer to what the actual home prices were for the input data. Based on the bedroom model, for 4 bedrooms, the Mobjack house is a moderately good deal (10k less than what the model predicted) and the Church house is not a good deal (100k more than what the model predicted). For 3 bedrooms, the Hudgins house is a very good deal (138k less than the model’s prediction) and the New Point Comfort house is also a slightly good deal (6k less than what the model predicted). For 2 bedrooms, the Moon house is not a good deal (81k more than the model’s prediction), and for 5 bedrooms, the Mathews house is a good deal (17.5k less than what the model predicted).