Machine Learning Toward AI Session 2

• 0-45 mins: Notebooks and Tutorials (as much as we get to)

  • Our goal (which we will supersede): https://playground.tensorflow.org/

  • Learning a Line Tutorial:

• 45-90 mins: From Neurons to networks

  • Loss Function and Training Loop Intuition: Loss Surface

  • Expanding from one neuron to many: Forward pass, Parameter Space,

  • Backwards Pass: Hand-wavey explanation involving "attribution"

  • Data and Training: MNIST

  • Encoding visual "concepts" and connection to LLM Embeddings/embedding space

• Break for Lunch

• 0-45 mins: Coding A Network!

• 45-60 mins: Review and Recap

• 60-75 mins: Pitch for Next Meeting

• 90-as long as you want to go: Working on "Creating a Network", Project

Things to watch and read:

• Fantastic Resource from 3Blue1Brown's Grant Sanderson (equivalent YouTube playlist)

• Also fantastic from Welch Labs