Trace how a model takes feature values as input and produces an output such as a price, category, score, or next word. Distinguish the trained model from the raw data it learned from and from the software that runs it.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Compare a learning algorithm with the fitted model it produces after training. Follow the basic training loop: use examples, make predictions, measure mistakes, and adjust the model so future predictions improve.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Recognize parameters as values the model learns during training, such as weights in a scoring rule. Contrast them with hyperparameters, which are choices people set before training, such as model size or learning rate.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Use a simple prediction mistake to calculate a loss value and see why training needs a numerical signal for “how wrong” the model was. Compare the idea of loss for numeric predictions and category predictions without going into advanced formulas.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Distinguish training from inference: training changes the model, while inference uses the trained model on new inputs. Recognize common serving patterns such as one-at-a-time predictions in an app and batch predictions over many records.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Sort supervised prediction tasks into classification, where the label is a category, and regression, where the label is a number. Reason through examples like spam/not spam, disease risk, delivery time, and home value.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Read model outputs that are not final decisions yet, such as confidence scores, probabilities, and ranked options. Practice choosing how a threshold can turn a prediction into an action, and why changing the threshold changes the outcome.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Separate training, validation, and test data by the job each split does. Use a small scenario to decide which data teaches the model, which data helps choose settings, and which data gives a final check on new-case performance.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Reason about whether a model will work on new examples, not just the examples it memorized. Compare underfitting, where the model misses real patterns, with overfitting, where it learns details that do not carry over.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Recognize unsupervised learning as pattern discovery without answer labels. Compare clustering similar items with finding unusual cases, using examples like customer groups, document themes, and suspicious transactions.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Distinguish decision-making systems from one-shot prediction systems by focusing on actions, consequences, and rewards over time. Trace the basic reinforcement learning vocabulary: agent, environment, state, action, policy, and reward.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Recognize generative models as systems that create new content rather than only selecting a label or predicting a number. Compare generation tasks such as writing text, making images, completing code, and producing audio from prompts.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Separate the four major machine learning goals: predicting known targets, discovering patterns, choosing actions, and generating new content. Classify everyday examples into these goals so the same word “AI” does not hide different technical jobs.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Review this chapter with practice based on your mistakes.