Overfitting in Deep Learning: Dropout Regularization
A deep learning model for image classification achieves 99% training accuracy but only 85% validation accuracy. The model has millions of parameters. Which technique is most likely to reduce overfitting while maintaining high accuracy?
Quick Answer
The answer is adding dropout layers with a rate of 0.5 after each convolutional block. This technique directly addresses overfitting by randomly deactivating a portion of neurons during each training pass, which forces the network to learn more robust, redundant representations rather than relying on specific neurons—a process known as dropout regularization. On the CompTIA AI+ AI0-001 exam, this scenario tests your understanding of how overfitting manifests when a model with millions of parameters memorizes training data (99% accuracy) but fails to generalize to validation data (85% accuracy). A common trap is confusing overfitting remedies with optimization tweaks: increasing epochs or reducing batch size does not directly combat overfitting caused by excessive model capacity, while dropout acts as a structural regularizer. Remember the mnemonic "Dropout Drops Dependency"—it cuts reliance on any single neuron, forcing the network to spread its learning across all paths.
⚠ Common exam trap
The AI0-001 exam often tests the misconception that reducing learning rate or batch size is a primary method to combat overfitting, when in fact these are optimization adjustments, not regularization techniques designed to reduce model capacity.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Add dropout layers with a rate of 0.5 after each convolutional block
Dropout is a regularization technique that randomly drops a fraction of neurons during training, which prevents co-adaptation of features and forces the network to learn more robust representations. With 99% training accuracy and 85% validation accuracy, the model is clearly overfitting, and adding dropout layers with a rate of 0.5 after each convolutional block directly addresses this by reducing the model's capacity to memorize the training data, while still allowing high accuracy on the validation set.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Reduce batch size from 32 to 8
Why it's wrong here
Smaller batches add gradient noise, which can slightly regularise, but they do not constrain the millions of free parameters causing the 99% versus 85% gap. Batch size governs optimisation step variance and hardware throughput, so it is the right lever when tuning convergence speed or memory, not capacity-driven overfitting.
- ✗
Decrease the learning rate by a factor of 10
Why it's wrong here
Lowering the learning rate only shrinks each gradient step; it does not constrain the millions of parameters that let the network memorise training data, so the train-validation gap persists. It tempts because it stabilises training and is correct when loss oscillates or diverges rather than when capacity causes overfitting.
- ✓
Add dropout layers with a rate of 0.5 after each convolutional block
Why this is correct
Dropout randomly deactivates neurons during training, preventing co-adaptation and forcing distributed representations. A rate of 0.5 after each convolutional block substantially regularises the millions of parameters, narrowing the training-validation gap while retaining capacity for high accuracy.
- ✗
Increase the number of training epochs to 500
Why it's wrong here
Training longer lets the network memorise the training set further, widening the 99% versus 85% gap rather than closing it. Epoch count controls how many passes over the data occur, so increasing it suits underfitting, where both training and validation accuracy remain low.
About these practice questions
One of 962 original AI0-001 practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
Same concept, more angles
1 more way this is tested on AI0-001
These questions test the same concept from different angles. Work through them to make sure you can recognise it however the exam phrases it.
Variation 1. A model's training accuracy is 99% but validation accuracy drops to 60%. What is the most likely issue?
easy- A.Data leakage
- ✓ B.Overfitting
- C.Multicollinearity
- D.Underfitting
Why B: A training accuracy of 99% with a validation accuracy of only 60% is a classic symptom of overfitting. The model has memorized the training data, including noise and outliers, rather than learning generalizable patterns, causing it to perform poorly on unseen validation data.
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AI0-001 practice question is part of Courseiva's free CompTIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI0-001 exam.