Pen and Paper HW

EECE490
How would you define Machine Learning?
Machine learning is the process of building machines that lear from data.
Learning:
- Get better at tasks
- Given a performance measure.
Machine learning is good for complex problems:
- for which we don’t have algorithmic solutions.
- Replace long lists of hand written rules
- Adapt to fluctuating environments
- Help humans understand the world

What is a labeled training set?

It is a set that contains the desired solution.

What are the two most common supervised tasks?

- Regression
- Classification 


Can you name two common unsupervised tasks?

- Clustering
- Visualisation
- Dimensionality reduction
- Association rule learning


What type of Machine Learning algorithm would

you use to allow a robot to walk in various
unknown terrains?
Reinforcement learning.

What type of algorithm would you use to segment

your customers into multiple groups?
2 cases:
Know how to segment customers: classification
Don’t know how to segment customers: clustering


Chadi Abi Fadel

Would you frame the problem of spam detection
as a supervised learning problem or an
unsupervised learning problem?
Supervised, it is fed the emails and their labels


What is the difference between a model

parameter and a learning algorithm’s
hyperparameter?
Parameter: to determine new predictions
Hyperparameter: parameter of the learning model itself not the algorithm

What do model-based learning algorithms search

for? What is the most common strategy they use
to succeed? How do they make predictions?
Optimal values for parameter so that it generalises well.
Ex: minimise the cost function


If your model performs great on the training data

but generalizes poorly to new instances, what is 

happening? Can you name three possible
solutions?
Overfitting
Solutions:
- More data
- Simplifying the model
- Reduce noise in training data


What is a test set and why would you want to use

it?
To estimate how the model will generalise.

What is the purpose of a validation set?

Compare models 


What can go wrong if you tune hyperparameters

using the test set?
Overfitting the test set


Chadi Abi Fadel

What is cross-validation and why would you prefer
it to a validation set?
Compare without validation set.
Saves training data

True/False. Explain your reasoning in 1 sentence.

a. A classifier that attains 100% accuracy on the training set
and 70% accuracy on test set is better than a classifier that
attains 70% accuracy on the training set and 75% accuracy on
test set.
False, Overfitting

Provide short answers.

What is a loss function? Give an example.
A loss function measures the discrepancy between the predicted values and the real ones

What is “generalization”?
Generalisation is the ability to perform well on data not used in training
What is overfitting?
Learning the training set very well and not performing on new data

Chadi Abi Fadel