How do I approach Machine Learning interview questions?

Machine Learning questions require understanding of core concepts and practice. PracHub provides solutions with explanations to help you master machine learning interviews.

What difficulty level is this interview question?

This is a medium difficulty Machine Learning question, commonly asked during Technical Screen rounds at J.P. Morgan.

What role is this question designed for?

This question is commonly asked for Data Scientist candidates at J.P. Morgan during technical interviews.

Explain Overfitting and Transformer Basics | J.P. Morgan Interview Question

Explain Overfitting and Transformer Basics

Last updated: Apr 11, 2026

Quick Overview

This question evaluates proficiency in core machine learning competencies such as overfitting and generalization, selection and regularization of loss functions for classification and regression, encoder-decoder sequence architectures, and self-attention mechanisms including queries, keys, and values, as well as considerations like bias–variance tradeoffs, masking, and attention computational cost. It is commonly asked in technical interviews for Machine Learning and Data Scientist roles because it probes both conceptual understanding and practical application of training dynamics, model architecture choices, and scalability trade-offs within the Machine Learning domain.

J.P. Morgan

Apr 7, 2026, 12:00 AM

Data Scientist

Technical Screen

Machine Learning

Answer the following machine learning questions in a self-contained way:

What is overfitting? How would you recognize it from training and validation behavior, and what are several practical ways to reduce it?
Write down appropriate loss functions for common supervised learning settings, such as binary classification and regression, and explain when each is suitable. Include how regularization can be added.
Explain the encoder-decoder architecture used in sequence models.
Explain self-attention, including the roles of queries, keys, and values, and how it differs from older recurrent sequence models.

A strong answer should also discuss tradeoffs such as bias-variance, model capacity, optimization vs generalization, masking in sequence generation, and the computational cost of attention for long sequences.

Solution

Show

Comments (0)

Loading comments...

Browse More Questions

More Machine Learning•More J.P. Morgan•More Data Scientist•J.P. Morgan Data Scientist•J.P. Morgan Machine Learning•Data Scientist Machine Learning