A model works well in training but fails on validation data. Which change most directly reduces overfitting?
Why A is the answer
Add regularization or simplify the model
Regularization and simpler models reduce overfitting. More training usually makes it worse.
In detail
- Regularization or a simpler model reduces overfitting.
- More training epochs usually make overfitting worse.
Use case
An illustrative scenario, not a real incident.
A demand model scores highly in training and poorly in validation. Adding regularization narrows the gap.
See it in different ways
flowchart TD
A{"Results"} -->|"Poor on training"| B["Underfitting: more features or a bigger model"]
A -->|"Good on training, poor on validation"| C["Overfitting: regularize or add data"]stateDiagram-v2 [*] --> Train Train --> Validate: tune Validate --> Train: adjust Validate --> Test: choose the final model Test --> [*]
flowchart TD
Q{"Validation compared with training?"}
Q -->|"Much worse"| A0["Regularize or simplify"]
Q -->|"Both poor"| A1["Model is underfitting"]Key point: Regularize or simplify to reduce overfitting.
Why the other options are wrong
- B. Train for more epochs until training loss stops falling
- More epochs make it worse.
- C. Add more features from the raw data
- More features can make it worse.
- D. Remove the validation set and train on everything
- Without validation, you cannot see the problem.