Question Bank

Machine Learning interview questions

Naming a model or a metric is only the start of an interview answer. Explain why it fits the task, and connect training, regularization, validation and evaluation to a concrete example. Start with your own explanation, then state its assumptions and limitations.

1635
questions
16
categories
50
companies

Showing 20 of 746 questions

MediumMachine Learning

What is cross-validation and why is it important in machine learning?

cross-validationvalidation
Practice in bot
MediumMachine Learning

Explain the difference between bias and variance and how this trade-off affects a model.

bias-variance tradeoffmodel evaluation
Practice in bot
MediumMachine Learning

When should Dropout be used in neural networks and how does it affect training?

dropoutneural networks
Practice in bot
HardMachine Learning

What is early stopping and how does it help in model training?

early stoppingmodel training
Practice in bot
HardMachine Learning

How does the Bagging ensemble method work and what are its main advantages?

ensemble methodsbagging
Practice in bot
HardMachine Learning

Explain what PCA is and how it can be used for dimensionality reduction.

pcadimensionality reduction
Practice in bot
EasyMachine LearningNNS

Why disable gradient computation during inference?

ml-general
Practice in bot
MediumMachine LearningNNS

What does model.eval() do in PyTorch and why is it needed?

ml-general
Practice in bot
MediumMachine LearningNNS

What is the difference between Batch Normalization and Layer Normalization?

ml-general
Practice in bot
MediumMachine LearningQuantum One

How to build metrics for demand forecasting model evaluation? What features are important for demand forecasting?

ml-general
Practice in bot
MediumMachine LearningQuantum One

How to handle missing values in sales data?

ml-general
Practice in bot
MediumMachine LearningQuantum One

How to properly impute data considering seasonality? Is it correct to fill December missing values with yearly average?

ml-general
Practice in bot
MediumMachine LearningВТБ

Why do banking recommendation systems use interpretable models (trees, LGBM) instead of deep learning?

ml-general
Practice in bot
MediumMachine Learning

Tell me in detail about one ML task: approach, beautiful idea under the hood.

ml-general
Practice in bot
MediumMachine LearningQuantum One

How to encode high-cardinality categorical feature (100 ports) for models that do not support categories?

ml-general
Practice in bot
MediumMachine LearningQuantum One

What error will we make if we calculate target encoding on the entire train set for time series?

ml-general
Practice in bot
EasyMachine LearningWildberries

Tell me about main binary classification metrics: precision, recall, accuracy, F1, ROC-AUC.

ml-general
Practice in bot
MediumMachine LearningWildberries

Why does F1-score use harmonic mean instead of arithmetic mean? What is the advantage?

ml-general
Practice in bot
EasyMachine LearningWildberries

A random classifier outputs 1 with probability 60%. Sample: 30 ones, 70 zeros. What are precision and recall?

ml-general
Practice in bot
MediumMachine LearningWildberries

Which model extrapolates better beyond the training set: decision tree, random forest, gradient boosting, KNN, linear regression, neural network?

ml-general
Practice in bot

Prepare for your next interview with Vibe Interview.

Download Vibe Interview
Machine Learning interview questions — page 3 | Vibe Interview