20 MORE Machine Learning Questions and Answer To Crush Interviews

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.

Checkout the First Interactive Beginner Machine Learning Quiz here!

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

Beginner Machine Learning Interview Quiz #2

 

Results

#1. Which of the following algorithms can be considered as a high-variance model?

Support Vector Machines

K-Nearest Neighbors

Logistic Regression

Decision Trees

#2. Which of the following statements is correct about type I errors?

A type I error occurs when the null hypothesis is true and is not rejected

A type I error occurs when the null hypothesis is true but is rejected

A type I error occurs when the null hypothesis is false but is not rejected

A type I error occurs when the null hypothesis is false and is rejected

#3. Which of these is not a variant of Boltzmann machines?

Energy-Based Models (EBM)

Contrastive Divergence (CD)

Deep Belief Networks (DBN)

Deep Boltzmann Machines (DBM)

#4. Which of the following is a good strategy to reduce bias from the results of an imbalanced dataset after modeling with a Decision Tree?

Fit the Decision Tree, then scale the results appropriately using the dataset as guidance

Fit the Decision Tree to each class separately, then combine the results proportionally to the initial dataset

Decision Trees will never work with imbalanced dataset

Balance the dataset, and only fit the Decision Tree

#5. What metric is used to identify the percentage of actual positives that were correctly identified?

Precision

Accuracy

Recall

F1

#6. Which of these is the default weight initialization for PyTorch?

Glorot/Xavier Uniform and Normal

Kaiming Initialization

Constant Initialization

LeCun Uniform and Normal

#7. In statistics, the number we’ve obtained when using a population is called?

Samples

Parameters

Population

Area

#8. Which of these can’t be used as a representation of Categorical variables?

Bar Charts

Pie Charts

Frequency Distribution

Histograms

#9. A model’s ability to correctly classify negative sample is known as

Specificity

Precision

Recall

Sensitivity

#10. After seeing the results of our prediction, we would like to trade-off Specificity and Sensitivity. What parameters can help us do this?

Decision Threshold

Breaking Point

Trade-off Accumulator

Precision-Recall Curve

#11. Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) are metrics used to evaluate a Regression Model. Which of the following is a correct statement about these metrics?

RMS penalizes larger differences between the predictions and the expected results

RMSE is significantly faster to compute than MAE

From both metrics, RMSE is the only one indifferent to the direction of the error

From both metrics, MAE is the only one indifferent to the direction of the error

#12. Overfitting and Underfitting are two fundamental concepts in Machine Learning. From the following statements, which ones are correct representations of Overfitting and Underfitting?

Select all that apply:

Overfitting happens when your model is too complex for your dataset. For example, a very deep neural network trying to learn a few dozen samples with a couple of features.

Underfitting happens when your model is too simple for your dataset. For example, a linear regression model trying to learn a large dataset with hundreds of features.

Overfitting happens when your model is too simple for your dataset. For example, a linear regression model trying to learn a large dataset with hundreds of features.

Underfitting happens when your model is too complex for your dataset. For example, a very deep neural network trying to learn a few dozen samples with a couple of features.Underfitting happens when your model is too complex for your dataset. For example, a very deep neural network trying to learn a few dozen samples with a couple of features.

#13. Assuming you are working with a severely imbalanced dataset. You want to split the data into two categories using a classification learning algorithm. Which of the following metrics should you avoid using when analyzing the algorithm’s performance?

Recall

Precision

F1-Score

Accuracy

#14. Occam ’s razor is the idea that, given two solutions with similar characteristics, the simplest and most direct one is the correct answer. In machine learning, Occam’s Razor is applied in many different scenarios. Which of the following examples are you comfortable justifying with Occam’s razor?

Given more than one learning algorithm with similar tradeoffs, we should use the least complex and most straightforward to interpret.

Feature selection and dimensionality reduction help simplify models to get better results.

Keeping the training process as fast as possible avoids overtraining and prevents overcomplicated results.

All of the above.

#15. All except one are attributes that represent a tensor?

The number of axes of the tensor.

The cardinality of a tensor, which represents the numerical relationship between the axes of the tensors.

The shape of the tensor, which represents how many dimensions does the tensor have along each axis.

The data type of the tensor, which represents the type of values contained in the tensor.

#16. Balancing the size of a dataset and number of features we use to train a model is always a problem we need to consider. Which of the following statements correctly summarizes your thoughts about the relationship between features and dataset size?

When training a model, as you add more features to the dataset, you often need to increase the dataset’s size to ensure the model learns reliably

When training a model, adding more features to the dataset increases the amount of information you can extract from the data. This allows you to use smaller datasets and still extract good performance from the data.

When training a learning algorithm, as you decrease the number of features in your dataset, you need to increase the number of training samples to make up the difference.

When training a learning algorithm, the number of features in your dataset is entirely independent of the number of training samples.

#17. How can we detect that variables are collinear?

Using k-Means to find similar clusters.

Using Variance Inflation factors.

Using Dimensionality reduction and extracting the standard deviation.

Using single linkage clustering.

#18. Assuming that you have sufficient data for each of the following problems, which of them would you address using Supervised Learning techniques?

Determine whether a website displays content for a mature audience.

Learn the best way to split a group of car buyers into different categories based on their buying pattern.

Given the medical records from patients suffering a specific illness, learn whether we can split them into different groups for better treatment.

Predict next year’s crop yields taking into account data of the past decade.

#19. Which of these is not true about regularization?

L1 regularization typically results in more zero-valued coefficients which means fewer features used.

L2 regularization results in small weights for many of the features that would have been zeroed out by L1 regularization

Too much regularization leads to overfitting.

Regularization helps us control our model capacity ensuring that our models are better at making correct classifications on unseen data points.

#20. What’s the goal of hyperparameter tuning?

To choose the set of optimal features from the data to train a model.

To choose the set of optimal samples from the data to train a model.

To choose the optimal parameters for a learning algorithm to train a model.

To choose the set of hypotheses that better fit the goal of the model.

Previous

Finish