A simple neural network’s validation performance improves initially but then worsens after a certain number of epochs. Which of the following changes would help prevent this degradation? (Choose two.)

The scenario describes overfitting, where the model learns the training data too well, leading to poor generalization on unseen validation data. Enabling early stopping directly addresses this by monitoring validation performance and stopping training when it begins to degrade, preventing the model from overfitting further. Increasing dropout rates is a regularization technique that randomly deactivates a percentage of neurons during training. This forces the network to learn more robust features and prevents co-adaptation of neurons, thereby reducing overfitting. Adding more layers or increasing neurons per layer would increase model complexity, making overfitting more likely, not less. Investigating model bias is important for fairness and accuracy but doesn't directly prevent overfitting during training.