NLP

From scratch:

• Character-level language models:
  • Recurrent Neural Network (RNN)
  • Bigram

NLP and Computer Vision:

• Image Captioning Transformer (CNN Encoder, Attention, LSTM Decoder): Notebook

Using libraries:

"On the shoulders of giants" - Isaac Newton

• Sentiment Analysis (LSTM, Bidirectional, Pre-trained GloVe): Notebook
• Word Embeddings (Pre-trained GloVe: Arithmetic, t-SNE, Clustering): Notebook
• Positional Encodings (Sine, Cosine): Notebook
• Azure AI (Sentiment Analysis) + Python Backend with Django: Web App
• BERT (+15 languages): Fine-tuning a Large Language Model (LLM) with TensorFlow: Notebook

Data:

• dinos.txt: 1.5k dinosaur names.
• names_3M_ARG.txt: 3M Argentine names.
• glove.6B.50d.txt: 6B tokens, 400k vocab, uncased, 50d vectors.
• IMDb_50k.csv: 50k IMDb movie reviews (positive/negative).