hi everyone
now i am in the week 3 of advanced ml and i got tired from understanding how to choose the model of your ml project , dr ng in the lab compare the train mse and the validation mse with the baseline to check if it is high bias or high variance so from where i will get the baseline of performance in any project example the simple house price prediction ?
Hi @Bin_Thabit
Start with a simple approach that doesn’t require much tuning, like predicting the mean or most frequent value for the target. This gives you a performance benchmark. You can then compare more advanced models to see if they improve upon this simple baseline.
Hope it helps! Feel free to ask if you need further assistance.
sorry but i didnt get what you meant.
look here in the lab there is a baseline performance to compare it with the mse of both train and validation, so how can i get this baseline.
and thanks for responding to me
The baseline performance in ML is usually a simple heuristic model, like predicting the mean or median of the target variable from the training set. In your house price prediction example, you can use the mean house price as a constant prediction for all samples.
As I remember, Professor Ng also mentioned that a baseline can be set based on minimum human performance when applicable, meaning the simplest level of accuracy a human could achieve on the task, which helps assess whether a model is truly adding value.
oooh ok i got it so in the house prediction model the if i chose the mean it will be constant for all the degrees and then i will compare it with the mse of the train and validation. thanks alot @Alireza_Saei i appreciate that you helped me
You’re welcome! Happy to help