Match the following definitions with the appropriate term for each.

In practice session 5, we applied a square root to every BoVW vector we computed.

Knowing that the Hellinger kernel is defined as:

,

what is the appropriate term to describe the technique we applied during practice session?

Among the following families of functions, select the ones a neural network with

• 1 hidden layer and an unrestricted (but finite) number of hidden units

• a nonconstant, bounded, monotonically increasing activation function

can learn.

Associate the appropriate term to each definition.

Which of those training methods can have a parallel implementation?

Among the following affirmations, check the ones which are true regarding Decision Trees.

How to limit the risks of over-fitting with a Random Forest Classifier?

The figure above depicts a distribution of "false" and "true" classes given a certain feature.

The figure above depicts a distribution of "false" and "true" classes given a certain feature.

Select among the following curves the one which represents the right "Receiver Operating Characteristic" (ROC) curve corresponding to this distribution and thresholds.

What is the expected accuracy of a truly random classifier (which picks one of the possible classes with a uniform probability) if we have 50 classes with equal probabilities.

Please provide 2 significant digits.