The Attention Mechanism and GPT's Transformer Architecture Behind LLMs and something about Graph Neural Network (GNN)

This quarter I’m taking Professor Candogan’s Frontiers in AI for Operational Decision-Making (BUSN 40915). We swept through the fundamentals of AI for the first two weeks. My opinion may be biased because I am already preloaded with all the AI terminologies, but the course infused extreme clarity into my understanding.

The following are key slides that I found informative/inspiring. Courtesy to Professor Candogan — I don’t own copyrights! All mistakes are on my own.

Something about GNN: it’s essentially a generalization of CNN.

Something about Transformer: (I tend to think…) it’s a much more powerful version of RNN: