CSCI 1109 - M38 - What is ML? tasks; train/val/test splits скачать в хорошем качестве

CSCI 1109 - M38 - What is ML? tasks; train/val/test splits

In this module, we zoom out from individual analyses and ask a new question: what happens when we want the computer to learn patterns from data and make predictions on new cases? You’ll meet the basic vocabulary of machine learning, see where it fits into the data science pipeline you’ve been building, and learn why train/validation/test splits are the backbone of honest model evaluation. Contrast “just writing rules” with learning from examples, using simple supervised tasks like predicting high water-use days or tipping percentages. Distinguish core ML task types—classification vs regression (and a taste of clustering)—and connect each one to concrete decision-making contexts. Set up and interpret train/validation/test splits in Python, and see how they help you spot overfitting and keep your test set truly “unseen.” Compute and compare basic metrics such as accuracy, MAE, and RMSE, and tie them back to what kinds of errors actually matter in a real problem. Recognize common pitfalls like data leakage and unbalanced splits, and practice quick sanity checks that keep your models more honest and responsible. By the end of the module, you should be able to explain what a machine learning model is in plain language, choose an appropriate basic task type and split strategy for a small tabular dataset, and read simple model results with a sharper sense of what they do—and don’t—tell you about performance in the real world. Course module page: https://web.cs.dal.ca/~rudzicz/Teaching/CS...