Machine Learning Interview Questions

Frank A. - "1) create the experimental and control groups. 2) Then calculate the proportion (mean) of the true conversion rates for both groups using the convert column which counts True as 1 and False as 0. This is their conversion rates 3) calculate the statistic of the two groups by subtracting the proportion and standardizing. 4) get the p-value and compare with 0.05. 5) conclude the difference is statistically significant if the p-value is less than 0.05 otherwise no statistical difference"See full answer

Zain R. - "I've worked on projects not quite like this, but very similar, in the past - I'll borrow from that to answer this: The Broader Context this problem doesn't specify the type of data we're working with, or how it's being ingested to align with my personal background, I'll assume a picture that lends this problem well to being a computer vision (abbreviated "CV") related question: let's say we have a conveyor belt in a waste facility, which sequentially carries a stream of waste w"See full answer

Dessalew A. - "Strength: simplicity + interpretability Weakness: rigidity (can’t model complexity well)"See full answer

Vips M. - "No discussion around better initialization of weights like Xavier etc.?"See full answer

Abinash S. - "I checked the unittest is giving a False assertion as you can see in the colab notebook below. F FAIL: testsimple (main_.Conv2dTest) Traceback (most recent call last): File "", line 19, in test_simple self.assertTrue(torch.equal(output, torch.tensor([[[[ 5., 1.], [ -2., -10.]]]]))) AssertionError: False is not true"See full answer

Chermaine Y. - "Zero in on the problem, the expectations of user are to find a restaurant but their feed is uninspired so they may bounce out of Yelp. Identify the impact size of user feeling like discovery is not personalised enough by seeing % of users that selected a restaurant from the homepage If large enough, I will look at who is likely the ones that want personalisation and why? Do they feel like they want to try new restaurants or are they finding it difficult to find restaurants they have been"See full answer

Nick S. - "[I'm not sure whether the answer below is the best, as I have not gotten result and feedback from my interview] Ans: I would solve by first using a VAE-style model, to create a latent space embedding that translates user description to generate images. Training would be done on the 1000 avatar images and 100000 descriptions, following this scheme: VAE: description -> encoder -> latent space -> decoder -> image Q: "OK, but that means you're limiting the generated images to be only the 1000 imag"See full answer

Ayush B. - "focus was on tradeoffs of diff object detection algorithms, data collection and labelling, foundational models, followup: not working well in production, retraining/active learning"See full answer