Week 6: CST383 Introduction to Data Science

    I can’t believe we’re already done with Week 6. It feels like the challenging part just started getting started, and now we only have two weeks left. The pace of this class has been very intense, but I am really enjoying it as the pieces are starting to connect...

    This week, one concept that stood out to me was hyperparameter tuning. It seems simple conceptually  just try different values and validate, but the number of combinations increases very quickly. I can see how grid search becomes inefficient and why random search can sometimes be more practical. I’d like to understand more deeply when increasing model complexity will actually help versus when it will affect performance.

    Overall, this week made me realize that building a model is not the hard part. The hard is choosing and validating it properly. It seems that machine learning is more of a careful process of testing assumptions and diagnosing errors.