- Core Concepts
Short notes on attention, cross-entropy loss, and native vision transformers.
- KNN K Nearest Neighbors
Strategies for keeping K-Nearest Neighbors robust against outliers, from preprocessing and tuning K to weighting neighbors and choosing distance metrics.
- RNN And LSTM
Explains how recurrent neural networks process sequences, why they suffer from vanishing gradients, and how LSTM gates fix long-term memory.
- Transfer Learning
Explains how transfer learning reuses a pre-trained source model through fine-tuning for a new task, and why it works so well with limited data.
- Transformer Architecture
A summary of the "Attention is All You Need" paper covering the encoder-decoder design, scaled dot-product and multi-head attention, positional encoding, and major Transformer variants.