20 Machine Learning Questions and Answer To Destroy Your Interview

Other Topics — Machine Learning Interview Questions

Introduction

In this article you will find essential machine learning interview questions that are geared towards beginners preparing for job or internship interviews. The questions in this article are general and cover a large breadth of information.

Without further ado, let’s get into the quiz!

Beginner Machine Learning Interview Quiz

 

#1. Are classifications and regressions supervised learning? Is community detection categorized as unsupervised learning or supervised learning?

No, Supervised

Yes, Unsupervised

No, Unsupervised

Yes, Supervised

#2. Does dimensionality reduction reduce the number of features or the number of samples in a dataset? Can reducing dimension help us save some computational cost?

Samples, Yes

Samples, No

Features, Yes

Features, No

#3. Are “for loops” efficient in Python Machine Learning? What alternative could be use?

Yes, None

Yes, While Loops

No, Switch Statements

No, Array Broadcasting

#4. Which statement is correct when comparing numpy arrays to python lists?

NumPy data structures take up less space

SciPy and NumPy have optimized functions such as linear algebra operations built in

NumPy arrays have a need for speed and are faster than lists

All of the above

#5. If a square matrix A has dimensions of (n X n) and linearly independent rows and columns, which of the following is False?

A is a full rank matrix

Rank of A is n

Inverse of A does not exist

Determinant of A is non zero

#6. Does dot-product represent the correlation between any two vectors? Is element-wise multiplication of two vectors the same as the dot-product of two vectors?

No, No

No, Yes

Yes, No

Yes, Yes

#7. How many parameters does a Gaussian Distribution have? If two Gaussians are independent of each other, is the summation of the two Gaussians a Gaussian?

1, No

2, No

1, Yes

2, Yes

#8. Which statement is incorrect?

1. Supervised learning needs both the Dataset (x) and the Labels (y)
2. Unsupervised learning finds the relationship between a data point and its label
3. Supervised learning needs a relationship between input (x) and output (y)
4. Unsupervised learning finds similarity among the data points

2

4

1

3

#9. Which of the following equations are correct?

p(X) = Σₓ p(X, Y)
p(X, Y) = p(X | Y) p(X)
p(X) = Σᵧ p(X, Y)

1 and 2

None

2 and 3

1 and 3

#10. Which of the following solutions are still a Gaussian Random Variable? In the choices X₁ and X₂ are independent Gaussian random variables, while A, B and b are just constant matrices or vectors.

1. X₁ + X₂
2. AX₁ + b
3. AX₁ + BX₂
4. AX₁ + BX₂ + b

1, 2 and 3

1, 2, 3 and 4

1, 3 and 4

1 and 2

#11. You are trying to send a telegram about the weather in San Fransisco. The weather has an equal probability of being Sunny, Cloudy, Rainy, and Windy (uniform each with 1/4 probability). What are the expected amount of bits needed to encode this message / what is the entropy limit?

1

2

4

8

#12. Mutual information quantifies what?

The difference between two distributions

Reduction in uncertainty after seeing another feature

Increase in uncertainty after seeing another feature

The similarity between two distributions

#13. What is measured by Mutual Information? Do we usually look for a feature with high mutual information or a low one?

Reduction in uncertainty, Low

Reduction in uncertainty, High

Increase in uncertainty, Low

Increase in uncertainty, High

#14. How can we optimize the parameters of a non-probabilistic model with equality constraints? How to optimize it for inequality constraints? What approach would we take to optimize a probabilistic model

Lagrangian function, KKT, MLE

MLE, Lagrangian function, KKT

KKT, MLE, Lagrangian function

Lagrangian function, MLE, KKT

#15. When implementing an EM algorithm for a Gaussian Mixture Model (GMM), in the E step we:

Re-estimate parameters using the current responsibilities

Evaluate the responsibilities using the current parameter values

Initialize our parameters to arbitrary values

Initialize our parameters all to zero

#16. Is the area under a curve of a mixture model equal to 1? Is GMM a soft assignment task?

No, Yes

No, No

Yes, No

Yes, Yes

#17. Is the Gaussian Mixture Model (GMM) a supervised or unsupervised learning method? In a GMM, each data point may belong to ___?

Supervised, Only a single cluster (hard clustering)

Unsupervised, Only a single cluster (hard clustering)

Supervised, Multiple clusters (each with respective probability)

Unsupervised, Multiple clusters (each with respective probability)

#18. Assume the points in the image above have been manually assigned to their correct cluster. Which clustering algorithm, if any, would result in the same or similar classification of points?

Gaussian Mixture Model

K-Means

All of the above

None of the above

#19. Which of the following is true for Single-link distance measurement?

A. It avoids chaining
B. It is sensitive to outliers

Only A

Only B

Both A and B

Neither A nor B

#20. In the figure below, if you draw a horizontal line for y = 2, what would be the total number of clusters formed?

2

3

4

5

Previous

Finish

Results