Basic Classifiers

  • KNN classifier
  • SVM classifier
  • Softmax classifier
  • 2 layer nn classifier

CNN Layers + Implementations

  • FC networks, arbitrary depth + optimize with several popular update rules
  • Batchnorm
  • Dropout
  • Some CNN layers
  • Practice PyTorch + train CNN of own design on CIFAR-10
  • Salience Maps with pretrained models

Sequential + Generative Model

  • Vanilla RNN for image captioning
  • LSTM for image captioning
  • GAN + improve classifier performance

Attention

  • Transformer for image captioning
  • Use self-supervised pretraining for better performance