Exercises

• Examine the time taken for models to converge. Is there a big difference between different models?
• Examine the training and validation losses between the models. What do you notice?
• Experiment with downscaling and upscaling the architecture, note how this affects learning.
• Experiment with different optimizers and loss metrics and note how this affects learning.
• Implement an LSTM on the IMBD dataset for sentiment classification.
• Implement an LSTM on the Wikimovies dataset to build a character/word-level language model and generate artificial movie plots.