Miscellaneous projects:

1. The file IMDB_sentiment.ipynb is a somewhat detailed notebook containing some collections of approaches to sentiment analysis using the IMDB dataset.

   • We first define a simple baseline and then compare various deep learning and machine learning approaches. The most successful model is a combination of various models achieving test accuracy of ~92%.

2. The file ResNet.ipynb is a toy model investigating image classification of the 102flowers dataset using a fine-tuned ResNet50. We look at how "discriminative learning rates" and "progressive resizing" impact training and accuracies of the fine-tuned models.