📚 Learning Guide
Degradation Problem in Deep Networks
medium

Why does increasing the depth of a neural network sometimes lead to worse performance, despite having more parameters?

Master this concept with our detailed explanation and step-by-step learning approach

Learning Path
Learning Path

Question & Answer
1
Understand Question
2
Review Options
3
Learn Explanation
4
Explore Topic

Choose the Best Answer

A

It causes the model to overfit the training data too quickly.

B

Deeper networks can suffer from vanishing gradients, making training ineffective.

C

More layers always improve the model's capacity to learn.

D

Increased depth requires more data, which is not always available.

Understanding the Answer

Let's break down why this is correct

Answer

Adding more layers gives the network more capacity, but it also makes training harder because the signal that tells each layer how to change must travel through many weights. As depth grows, gradients can shrink or explode, so the early layers learn very slowly or not at all, a problem called vanishing or exploding gradients. Moreover, deeper models are more flexible and can fit the training data too well, leading to overfitting and poor generalization. For example, a 3‑layer network might achieve 90 % accuracy, while a 20‑layer version can get stuck at 70 % because its gradients vanish and it overfits. Techniques like batch normalization or residual connections are often needed to keep deep networks trainable.

Detailed Explanation

When a network has many layers, the error signal that tells the model how to change its weights travels through each layer during training. Other options are incorrect because People think more layers mean the model will fit the training data too fast, but that happens when the data is too small, not just because the network is deeper; It is tempting to think that more layers always give a better model, but deeper networks are harder to train.

Key Concepts

Degradation problem in deep networks
Vanishing gradients
Overfitting
Topic

Degradation Problem in Deep Networks

Difficulty

medium level question

Cognitive Level

understand

Practice Similar Questions

Test your understanding with related questions

Ready to Master More Topics?

Join thousands of students using Seekh's interactive learning platform to excel in their studies with personalized practice and detailed explanations.