What is underfitting and what strategies fix it?
This question requires an explanation of underfitting, its causes, and methods to increase model capacity or feature richness.
Why Interviewers Ask This
While overfitting gets more attention, underfitting indicates a model that is too simple to capture underlying patterns. Interviewers ask this to verify you can diagnose both sides of the bias-variance spectrum. They want to know if you understand that simply adding more data isn't always the solution and that model architecture or feature engineering might be the bottleneck.
How to Answer This Question
Define underfitting as a state where the model is too simple to capture the data's structure, leading to poor performance on both training and test sets. Contrast this with overfitting. Then, propose solutions such as choosing a more complex model (e.g., deeper trees or more layers), adding relevant features, reducing regularization strength, or training for more epochs. Emphasize that the goal is to find the right balance.
Key Points to Cover
- Underfitting means the model is too simple to capture patterns.
- Solutions include increasing model complexity.
- Adding relevant features can help the model learn better.
- Reducing regularization allows the model more freedom to fit.
Sample Answer
Underfitting happens when a model is too simplistic to learn the underlying patterns in the data, resulting in low accuracy on both training and test datasets. This often occurs due to insufficient model complexity or excessive regularization. To fix it, I would first consider switching to a more complex algorithm, such as a deeper decision tree or a neural network with more layers. I would also review the feature set to add meaningful variables that better represent the problem. Finally, I might reduce the regularization penalty or increase the number of training epochs to allow the model more opportunity to learn.
Common Mistakes to Avoid
- Confusing underfitting with overfitting.
- Suggesting only more data without addressing model complexity.
- Not mentioning feature engineering as a solution.
Practice This Question with AI
Answer this question orally or via text and get instant AI-powered feedback on your response quality, structure, and delivery.
Related Interview Questions
How do you handle missing or inconsistent data in a dataset?
Medium
AmazonWhat are the steps involved in the typical lifecycle of a data science project?
Medium
AmazonWhat is Elastic Net and when should it be used?
Hard
Can you explain the difference between supervised and unsupervised learning?
Easy
AmazonWhat are the main differences between precision and recall?
Medium
What are the common loss functions used in regression?
Medium