NCA-GENM Question 20
Single answerYou are training a deep neural network for a multimodal generative AI model, and the model is experiencing vanishing gradient issues. To mitigate this, you decide to use residual connections. What is the primary purpose of introducing residual connections in such a scenario?
- A
To enable the model to skip irrelevant data during training
- B
To allow the gradient to flow through the network more effectively during backpropagation
- C
To reduce the number of parameters in the neural network
- D
To prevent overfitting by adding regularization
Show answer and explanation
Correct answer: B
Explanation
Residual connections, a key component of nonsequential neural networks, address the issue of vanishing gradients by introducing shortcut pathways that allow the gradient to bypass certain layers during backpropagation. This ensures better gradient flow and enables the training of very deep neural networks, which is crucial for multimodal generative AI models.
- A. Incorrect.
Residual connections do not directly help in skipping irrelevant data; they are designed to address issues like vanishing gradients.
- B. Correct.
Residual connections help in mitigating vanishing gradient issues by allowing the gradient to flow through the network more effectively, especially in very deep neural networks.
- C. Incorrect.
Residual connections do not reduce the number of parameters in a neural network; instead, they introduce additional pathways to improve training stability.
- D. Incorrect.
While residual connections may indirectly affect overfitting, their primary purpose is not to act as a regularization technique but to improve gradient flow.