Depth

Machine learning quiz

All 10 questions. After you submit, each question shows the right answer and why. Back to the quick check

ExplorerQuestion 1 of 10

To check whether a machine-learning model really works, it should be tested on examples it did not learn from.

True

False

ExplorerQuestion 2 of 10

A computer is shown thousands of photos, each tagged 'cat' or 'dog', and learns to tag new photos. What kind of learning is this?

Supervised learning

No learning at all

Reinforcement learning

Unsupervised learning

InvestigatorQuestion 3 of 10

Data points (1, 2), (2, 4) and (3, 6). The model y = 1.5x predicts 1.5, 3 and 4.5. What is its mean squared error? (Give your answer to 2 decimal places.)

Your answer

InvestigatorQuestion 4 of 10

A shop groups its customers into types based on what they buy, without anyone saying in advance what the types are. Which kind of learning fits best?

Unsupervised learning

Rule-based programming

Supervised learning

Reinforcement learning

InvestigatorQuestion 5 of 10

In a decision tree, what does a leaf do?

It splits the examples into two new groups

It stores a copy of every training example

It changes the questions while the tree is being used

It gives one output for every input that reaches it

ScientistQuestion 6 of 10

According to the Deep Learning textbook, which trend in the history of deep learning was made possible by faster CPUs, general-purpose GPUs and better distributed-computing software?

The increase in model size over time

The replacement of neural networks by hand-written rules

The end of the need for training data

The disappearance of overfitting

ScientistQuestion 7 of 10

A model has 1% error on its training set but 30% error on its test set. What is the best diagnosis?

Overfitting

The test set is too small to ever matter

Underfitting

Perfect generalisation

AdvancedQuestion 8 of 10

In reinforcement learning, the learner is told the correct action for each situation, just as in supervised learning.

True

False

AdvancedQuestion 9 of 10

In Mitchell's definition of learning, for a spam filter, which item is the experience E?

The percentage of emails sorted correctly

The collection of emails people have already labelled as spam or not spam

The job of sorting incoming email

The computer the filter runs on

ExpertQuestion 10 of 10

Name the 1959 paper whose title contains the phrase 'machine learning', and say what task its program learned.

Your answer