Day 12 and 13 - Pet Breeds Classifier, Multi-Classification, Sizing and Test Time Augmentation
Nov 13 2024
(Fast.ai)
- Lesson 4 - reading Chapter 10, but also chapter 5, 6, and 7 since these are not covered in the lessons
- Finished reading Chapter 5 - Pet Breeds Classifier
- Talked about fine tuning and freezing parameters of pre-trained models at first; and having different learning rates for the pre-trained layers vs the new layers you added on.
- Finished Reading Chapter 6 Multi-Classification Google “python for data analysis free pdf” to get the free book
- Finished reading Chapter 7 Sizing and Test Time Augmentation
- Started Chapter 10 - NPM Deep Dive RNNs - Recurrent Neural Networks
- Self-supervised learning: Training a model using labels that are embedded in the independent variable, rather than requiring external labels. For instance, training a model to predict the next word in a text
Nov 14 2024
(Stanford)
-
[AI/ML Learning - Neural Networks Artificial Intelligence & Machine Learning 8 - Neural Networks Stanford CS221: AI (Autumn 2021) - How Machines originally couldn’t do XOR (one layer) but now they can by having a non-linear function between layers.
(Fast.ai)
- Completed Chapter 10 - NPM Deep Dive RNNs - Recurrent Neural Networks
- IMDB problem of categorizing movie reviews into positive or negative. But first, we need to fine tuning pretrained NLP models on the language used for movie reviews, then we can do a better job at categorizing the reviews.