W2 Practice quiz: Activation Functions - Query Answer clarification

Course Q&A Machine Learning Specialization Advanced Learning Algorithms
,
1

Hi Everyone,

In the explanation video ‘Choosing activation functions‘, it was shared that for regression problem like predicting house prices scenario, we will pick the Activation Function which can result the 0 or positive values in the output Layer.

However, in the same week ‘Practice quiz: Activation Functions‘ quiz, in the below query, one of the answer is ‘linear’. Isn’t it contradicting the choice of picking the activation model as linear activation can server the negative values/number as well where predicting the housing prices should result only 0 or positive values???

2.

Question 2

1407×680 50.7 KB

For the task of predicting housing prices, which activation functions could you choose for the output layer? Choose the 2 options that apply.

ReLU

Correct

Yes! ReLU outputs values 0 or greater, and housing prices are positive values.

Sigmoid

linear

Correct

Yes! A linear activation function can be used for a regression task where the output can be both negative and positive, but it’s also possible to use it for a task where the output is 0 or greater (like with house prices).

2

Personal opinion:
I would never use ReLU as an output activation. It provides no advantage over using a simple linear activation.

In the case of this specific example, where the expected outputs are all assumed to be positive, ReLU could be used. But that’s more of a party trick than a strategy.

1 Like