Exam AWS Certified Machine Learning - Specialty topic 1 question 39 discussion

DROPOUT HELPS PREVENT OVERFITTING https://keras.io/layers/core/#dropout THE BEAUTIFUL ANSER SHOULD BE B.

https://kharshit.github.io/blog/2018/05/04/dropout-prevent-overfitting Answer is B 100%

Increasing dropout rate will reduce complexity of the model which inturn reduces overfitting

This is clearly B, dont get why the answer is marked as D.

Regularization will seek to obtain similar accuracies in train and test sets. Anything else will make the overfitting worse

B is correct, D so stup*d answer

A. NO - accuracy on training set is high B. YES - increased dropout rate => reduce model complexity => less overfitting C. NO - we want to reduce model complexity D. NO - the model converged

I don't understand why the highlighted "right" answer is D. To increase the number of epochs will make the situation even worse than it is; dropout is the right action to take in this case

B is correct

agree, B makes more sense here

Definitely B because overfitting comes from complex model that captures patterns of training data well. But D is getting this model more complex, worsening overfitting.

Increasing Epoch only makes things worse on a overfitting model. You should perform regularization by introducing drop outs to generalize the model.

Option B is the correct answer because increasing the dropout rate at the flatten layer helps prevent overfitting by randomly dropping out units during training, effectively creating a more robust model that can generalize better to new data. Dropout is a regularization technique that helps prevent overfitting by forcing the model to learn redundant representations of the data. By increasing the dropout rate at the flatten layer, the model becomes more generalized, which should help to improve the testing accuracy.

Overfitting occurs when a model is too complex and memorizes the training data instead of learning the underlying pattern. As a result, the model performs well on the training data but poorly on new, unseen data. Increasing the dropout rate, a regularization technique, can help combat overfitting by randomly dropping out some neurons during training, which prevents the model from relying too heavily on any single feature.

Model is overfitting, I will go with option B, increasing epoch will cause more overfitting

it should answer B.

12-sep exam