A wildlife club gathers 4,000 bird photos before anyone marks their species. Which workflow stage is the club doing?

Volunteers examine each weather record and add either "storm" or "no storm" to it. What stage is this?

A program repeatedly compares its predicted plant disease with the expert-marked answer and adjusts internal values to reduce mistakes. What is happening?

After training, a team checks a model using 500 labeled receipts that were never used to adjust it. Which stage is this?

A city puts its trained pothole detector into a phone app so residents can submit road photos. What stage has begun?

A team wants to train a model to learn which saved messages are scams. What must happen to the saved messages before supervised training can begin?

A leaf classifier was trained only with healthy green leaves. It is then given yellow, damaged leaves. What is the most likely result?

A team adds many correctly labeled night-time road images to training data that previously contained only daylight images. What change is most likely?

A model is trained with 9,000 examples labeled "cat" and only 50 labeled "dog." Which risk does this imbalance create?

A fruit model was trained on photos taken in one orchard. It works poorly on fruit photos from a different orchard with different lighting. What would most directly help?

An audio model learned "yes" and "no" from recordings by adults only. Why might it make more errors on children's voices?

A team discovers that 15% of its training photos were labeled "recyclable" when they actually showed trash. What can those wrong labels do?

A recommendation model was trained on music choices from only one age group. What is a reasonable concern when it recommends music to other age groups?

Each row in a data set includes study hours, attendance rate, prior quiz score, and whether the student passed the final exam. What is the label?

A model predicts whether a package will arrive late. Each row has distance, shipping method, weather, and "late" or "on time." Which column is the label?

A school data set is meant to predict lunch choice. It lists grade level, favorite cuisine, dietary restriction, and the meal selected. Which value is the label rather than a feature?

For a model that predicts a house's sale price, rows include bedrooms, square footage, neighborhood, and sale price. What should the model be trained to output?

A gardener wants a model to identify whether a tomato is ripe. A row contains color, firmness, diameter, and an expert's ripe/unripe mark. Which item is NOT a feature?

A program receives thousands of emails already marked spam or not spam and learns to predict those marks for new emails. What kind of learning is this?

A music app groups songs by similarities in tempo and instruments, but no song was given a category beforehand. Which approach is it using?

Which situation best illustrates supervised learning?

A store gives a program purchase records with no "customer type" column. The program finds groups with similar buying patterns. Why is this not supervised learning?

A class wants a model that predicts whether homework will be submitted on time. Which data set is appropriate for supervised learning?

A team trains a model on 8,000 labeled messages and keeps 2,000 labeled messages out of training. What is the main purpose of the 2,000 held-out messages?

A classifier gets 180 correct answers out of 200 held-out test examples. What does its 90% accuracy report?

Why should a team avoid reporting accuracy only on the same examples used to train a model?

Two models are tested on the same 1,000 held-out images. Model A gets 920 correct and Model B gets 870 correct. Which conclusion is supported?

After a training run ends, what is a trained model?

A team saves its trained model and later gives it a new photo. What does the model do with that photo?

Which statement correctly distinguishes an algorithm from a trained model?