Discuss Google Professional Machine Learning Engineer Exam Topic 6 Question 107

You are building a linear model with over 100 input features, all with values between -1 and 1. You suspect that many features are non-informative. You want to remove the non-informative features from your model while keeping the informative ones in their original form. Which technique should you use?

AUse Principal Component Analysis to eliminate the least informative features.

BUse L1 regularization to reduce the coefficients of uninformative features to 0.

CAfter building your model, use Shapley values to determine which features are the most informative.

DUse an iterative dropout technique to identify which features do not degrade the model when removed.

Suggested Answer: B

L1 regularization, also known as Lasso regularization, adds the sum of the absolute values of the model's coefficients to the loss function1.It encourages sparsity in the model by shrinking some coefficients to precisely zero2. This way, L1 regularization can perform feature selection and remove the non-informative features from the model while keeping the informative ones in their original form. Therefore, using L1 regularization is the best technique for this use case.

Regularization in Machine Learning - GeeksforGeeks

Regularization in Machine Learning (with Code Examples) - Dataquest

L1 And L2 Regularization Explained & Practical How To Examples

L1 and L2 as Regularization for a Linear Model

Get Premium Professional Machine Learning Engineer Questions as Interactive Web-Based Practice Test or PDF Real Exam Format Detailed Explanations Instant Access Get Your Practice Test

29 comments · newest first

Iraida 7 months ago

PCA can lose interpretability. I’d stick with L1 for clarity.

Johnetta 7 months ago

L1 regularization directly sets uninformative features to zero. Very efficient!

Olive 7 months ago

Iterative dropout sounds interesting, but it might take too long.

Stephen 7 months ago

Shapley values are great for understanding feature importance after modeling.

Jerry 7 months ago

PCA is good, but it transforms features. I prefer keeping them as is.

Miesha 8 months ago

I think L1 regularization is the best choice. It simplifies the model.

Jill 8 months ago

Wait, iterative dropout? That sounds a bit unconventional for this scenario!

Lucia 8 months ago

Definitely agree with L1 regularization! It's effective for feature selection.

Luz 8 months ago

This question is like a buffet of feature selection methods - I'll have one of each, please!

Jacquline 9 months ago

B) L1 regularization, the feature pruning ninja technique. Snip, snip!

Cammy 9 months ago

C) Shapley values, the secret sauce for feature importance. Mmm, tasty.

Genevieve 9 months ago

D) Iterative dropout sounds like a fun way to play feature selection roulette.

Nu 9 months ago

B) L1 regularization is the way to go! Gotta love that sparsity.

Temeka 9 months ago

I practiced a question similar to this where we used Shapley values to evaluate feature importance after model training.

Tawna 9 months ago

I'm not entirely sure, but I think PCA is more about transforming features rather than just eliminating them.

Rosina 10 months ago

I'm a bit confused here. A) PCA is more for dimensionality reduction, not really feature selection, right? I think I'll stick with B) L1 as the most straightforward option.

Blythe 10 months ago

D) Iterative dropout could work, but that seems a bit more complicated than I'd want to try on an exam. I'm leaning towards B) L1 regularization - it's a classic feature selection method that I'm comfortable with.

Benedict 10 months ago

Shapley values sound interesting, but is it really practical for 100+ features?

Marion 10 months ago

I think L1 regularization is the way to go here!

Louann 10 months ago

I remember discussing L1 regularization in class, and it seems like a good option since it can shrink some coefficients to zero.

Tamesha 11 months ago

PCA won't really help keep features in their original form.

Winfred 11 months ago

I feel like the iterative dropout technique could be useful, but I can't recall if it's commonly used for feature selection specifically.

Victor 11 months ago

Hmm, I'm not sure. C) using Shapley values seems interesting, but I'm not super familiar with that technique. I'll have to look into it more.

Chantell 11 months ago

I think I'd go with B) L1 regularization. That seems like a good way to automatically shrink the coefficients of the less important features to zero, keeping the important ones.

Rebbeca 6 months ago

↳ replying to Chantell

PCA is interesting, but I think B is more practical for this case.

Naomi 6 months ago

↳ replying to Chantell

L1 regularization helps simplify the model, which is great.

Rima 6 months ago

↳ replying to Chantell

I like the idea of using Shapley values too, but B feels more straightforward.

Denny 6 months ago

↳ replying to Chantell

B is definitely a solid choice!

Francesco 6 months ago

↳ replying to Chantell

I agree, L1 regularization is effective for feature selection.

Google Professional Machine Learning Engineer Exam - Topic 6 Question 107 Discussion