Machine Learning and Neural Networks

What is a feature in machine learning?

• A program that runs the dataset

• A characteristic that describes an object

• A rule that controls the computer

• A picture that shows the results

In a tabular dataset, what does each column represent?

• A place to store images only

• A particular feature of the data

• A single example in the dataset

• A step in a computer program

In a table of a dataset, what does each row correspond to?

• A computer hardware part

• An example in the dataset

• A random number generator

• A feature name list

In the ML4K pipeline, what is the first step shown in the diagram?

• Use the model

• Test the model

• Train a classifier

• Collect your data

Which step comes immediately after you label each item to a class in the ML4K pipeline?

• Use the model in a system

• Design a feature set

• Train a classifier

• Collect your data

A team has collected pictures and chosen features like body type and fin color. What should they do next to build the dataset, according to the diagram?

• Skip straight to training

• Use the model in a game

• Immediately test the model

• Code the data using the feature set

Which source is a valid way to collect data shown in the image?

• Stories remembered without checking

• Random numbers made by guessing

• Opinions guessed without asking

• Sensors in a car tracking speed

A dataset is split into two parts before training a model. Which pair correctly names these parts and shows their typical sizes?

• Training set ninety percent, test set ten percent

• Training set ten percent, test set ninety percent

• Training set half, test set half

• Training set seventy percent, test set thirty percent

What does the algorithm do with the training data during model building?

• Finds patterns called statistical correlations

• Randomly guesses without using features

• Deletes half of the examples to simplify

• Changes labels to match the features

After a model is trained, what is the next step to check how well it works?

• Test the model using the test set for accuracy

• Add more features without checking results

• Split the training set again into two sets

• Ignore testing because training is enough