April 2020
Intermediate to advanced
438 pages
12h 2m
English
What is likely to happen if you select more convolutional layers to retrain (fine-tune)? For example, what if you select the last four (or six) convolutional layers and retrain them by tweaking the preceding code? What is the impact on the test accuracy? Does it improve? Change the optimizer to Adam and tune the (hyper-) parameters' epoch, learning rate, and batch_size. Note the impact of the hyperparameter tuning on test accuracy. Can you get a model with better accuracy only by using transfer learning? Try to implement transfer learning/fine-tuning using PyTorch and TensorFlow as well.
Read now
Unlock full access