ML Review Quiz: Big Recap

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

3 main types of learning *

You were introduced to 3 different learning strategies utilized by machine learning algorithms: unsupervised learning, supervised learning, and reinforcement learning. Match each type of learning to one of the goals below:

3 points

unsupervised

supervised

reinforcement

Learn how to optimally behave in your environment

Discover patterns in your data

Predict Y from X

This type of ML produces no right or wrong answers, but a human has to interpret the results and make the best of it: *

1 point

Unsupervised Learning

Supervised Learning

Reinforcement Learning

The main ML types also have subcategories... *

...match each subcategory to its main learning type.

5 points

Unsupervised Learning

Supervised Learning

Anomaly Detection

Dimensionality Reduction

Classification

Clustering

Regression

Over- and Underfitting *

A model's performance on new data points can be bad for two reasons: underfitting (= the model has a high bias) or overfitting (= high variance). Which is which?

2 points

underfitting

overfitting

When you evaluate the model on the data it was trained on, the performance is close to that of a human, but on new data points it performs poorly.

No matter on what data (train or test) you evaluate the model, the performance is always far below that of a human.

Data vs. Concept Drift *

We want to predict the price of a house from its size. Which of the scenarios results in either a data or a concept drift? Remember: Data Drift is when the input distribution changes, Concept Drift is when the input/output relation changes.

2 points

Data Drift

Concept Drift

Due to inflation, the same size house now costs 20% more than 5 years ago.

On average, people built larger houses in 2020 than in 2010.

Machine learning is an “iterative” process, meaning that an AI team often has to try many ideas before arriving at a solution that’s good enough, rather than have the first thing they try work. *

1 point

True

False

Which of these are reasons that it’s often unrealistic to expect an ML system to be 100% accurate? *

1 point

You might not have enough data

Data can be mislabeled

Data can be ambiguous

All of the above

Submit

Clear form

This form was created inside of Dr. Franziska Horn. Report Abuse

Forms