Program Structure


  • Intro to program
  • Cover tools
  • Vocabulary
  • Where DL fits into ML

Neural Networks

  • Build NN w/ Python. Algos covered:
    • Gradient descent
    • Backpropogation
  • First project: Predict bike use
  • Model evaluation and validation
  • Define and train deep models in PyTorch

Convolutional Networks

  • Used for computer vision
  • Build one w/ PyTorch
  • Second project: Classify dog breeds from pictures.
  • Build autoencoder (Used for image compression)
  • Transfer learning

Recurrent Neural Networks

  • Used on data that comes in sequence, like text
  • Build a text generating RNN
  • NLP
    • Word2Vec
    • Sentiment prediction
  • Third project: Generate TV scripts
  • Generative Adversarial Networks

Generative Adversarial Networks

  • Used to understand real world data
  • Generate images w/ a CycleGAN
  • Fourth project: Generate faces w/ a ‘deep convolutional GAN’

Deploying ML Models

  • Deploy and make accessible via web app
  • Monitor models w/ AWS SageMaker
  • Fifth project: Deploy sentiment analysis model accessible from website


  • Projects have ‘suggested’ deadlines
    • Designed to guide progress
    • No penalty for missing
  • Course has final, term end deadline
  • To pass course, all projects must be completed to passing degree
  • If all projects are not completed by term end, a 4 week extension is automatically applied.


  • Knowledge platform
    • Similar to SO
    • Peers and staff answer questions
  • Study groups
    • Group chat with cohort
    • New rooms open for each project
  • Reviews
    • Each project submission receives feedback
    • Changes might be requested
    • No limit on submissions


  • Required: Intermediate Python Experience (Ruh roh)
  • Option: Multivaribale calc, linear algebra