A collection of Phython scripts resulting from the MIT deep leaning 2024 course.
http://introtodeeplearning.com/
Different models for different problem domains:
Perceptron : Fully connected layers of neurons (eg. hand writting detection)
Recurrent neural network : Data with a temporal component eg: speach, music
Variational auto encoder : Unsupervised learning (eg. debiasing a data set)
Convolutional neural network : Images (eg. facial recognition)
Generative adversarial network : To generate data (eg. for unsupervised learning)
Reinforcement learning : Gaming
Graph CNNs : Geomtetric features
Diffusion models (stable / latent) : Text to image generation, protein folding
Transformers (cross / self attention) : Text prediction / translation etc
