2023 September Newsletter

Hey!

MY QUICK UPDATES

THOUGHTS, INSIGHTS, AND ADVICE

  • Be careful how deep you go into theory when you first start on your Machine Learning journey. Fully understanding a single algorithm can involve reading papers, referencing books, deeply understanding math, and analyzing code. I think it’s better to understand the basics of how the algorithm works and then move on to practical projects. Check out this post how to learn a machine learning algorithm by machinelearningmaster.com.

RANDOM LEARNINGS, KNOWLEDGE, AND REMINDERS

  • Train Set vs Validation Set vs Test Set: For Supervised Learning problems you often divide your data into three chunks. You create a Train Set to find parameters for your model, a Validation Set to choose between different models or choose between different hyperparameters, and a Test Set to assess Generalization Error on data your model hasn’t seen before.
  • Cross Validation: This is a common technique to choose between algorithms or choose between different hyperparameters. It’s performed by creating many different splits of your data to make many different combinations of a Train Set and Validation Set. You then assess different models/hyperparameters by using each of these different splits and averaging the results to choose the one that performs the best. After this you can then train the best model/hyperparameters on the entire dataset that you used for cross-validation. Keep in mind that before any of this you should still set aside a chunk of data for the Test Set to assess Generalization Error at the end.

MY NEW CONTENT

OTHER USEFUL/INTERESTING RESOURCES

Michael Hammer

Read all my Newsletters here: https://michaelphammer.com/newsletter/

Consider Sharing:

Facebook
Twitter
Pinterest
LinkedIn

More Articles You Might Enjoy