Overfitting and Prevention Techniques

You are training a deep neural network for Pinterest’s visual search engine. During training you observe that the training loss keeps decreasing while the validation loss starts to increase after epoch 8. Your task is to (1) explain why this is happening, (2) implement a Python function that detects the onset of overfitting by monitoring the per-epoch training and validation losses, and (3) apply at least two distinct regularization techniques to mitigate the problem. Your solution must be compatible with PyTorch and should not require re-labeling any data. You may assume the model is a ResNet-50 fine-tuned on 1.2 M Pinterest images (224×224 RGB) with 32 GB GPU memory and a 12-hour training budget.