Quiz 1 Doubt regarding regularisation

  1. Question 8 Increasing the parameter keep_prob from (say) 0.5 to 0.6 will likely cause the following: (Check the two that apply)
    Increasing the regularization effect
    Reducing the regularization effect
    Causing the neural network to end up with a higher training set error
    Causing the neural network to end up with a lower training set error

Could anyone Please explain how can it lead to decrease in regularisation and how can it lead to low training set error?

2 Likes

Hi @ajaykumar3456, from your question you seem to have a good understanding of the working of keep_prob. What will happen when you decrease keep_prob from 0.5 to say 0.4?
regards Stephanus

1 Like

when keep_prob decreases to 0.4, then we will be eliminating more number of neurons in that particular layer
Correct me If i am wrong.
if we are eliminating more number of neurons, then more regularisation is being done as the network is being simplified.

1 Like

So what will happen the other way around, when you increase?

1 Like

If you are increasing the keep_prob, You are allowing more number of neurons to stay in the layer,and less amount of regularisation is being done. And at the same time as the network is more complex and as we are not applying any kind of regularisation, training set error will be low.

@sjfischer Please correct me If I am wrong.
and Thanks for the way you are clarifying doubts by giving hints.