Daily Notes: 2025-12-11

ML Notes

Introduction to Classification (Using scikit-learn)

• Given the “no free lunch theorem,” it is always recommended to try a handful of different learning algorithms to choose the best for a particular problem.
• Factors include:
  • The amount of training examples or features
  • The amount of noise in the dataset
  • Whether the classes are linearly separable

5 main steps involved in training a supervised ML algorithm

1. Selecting features + collecting (labeled) training examples.
2. Choosing a performance metric.
3. Choosing a learning algorithm + training a model.
4. Evaluating the performance based on the performance metric.
5. Changing the settings of the algorithm + tuning the model. (Adjusting hyperparameters like learning rate, # of epochs, etc.)

Personal Notes

Questions I still have

Tomorrow’s plan