Final Exam Information Inroduction to Deep Learning Oliver
- Slides: 7
Final Exam Information Inroduction to Deep Learning Oliver Schulte
Topics Included • From advanced RNNs (LSTM, GRU) to generative models, transformer • Both material from the slides and the book • Check lecture schedule for details on the book
Topics Excluded • Reinforcement Learning
Sample Topics Overview (I) • RNNs with gates (LSTM, GRUs), about 15% • Encoder-Decoder architectures, about 20% – Transformer, attention, self-attention • Embeddings, about 15% • Auto-Encoders, about 20% – Basic auto-associative, PCA (from slides) – Convolutional • Generative Models, about 30% – Variational auto-encoder – Generative adversarial network
Format: Long Answer • • On-line, should take about 1 hour 2 parts, subset and short answer type Subset: I post 5 or 6 questions. Your exam selects 2. You upload the answers to each question. Questions about the level of hard mid-term question. Open book but not open neighbour. Copying will lead to a mark of 0, you must give your own answers. • Kind of like assignment • Write answer in word processor, upload file. – Maybe review equation editor in Office, or Latex package.
Format: Short Answer • Time budget, should take about 1 hour • Similar to most midterm questions, mix of – Multiple choice – Match the concept – Compute an answer – Short answer – Recognition (e. g. what does this piece of code do? ) • Enter answers in Canvas, similar to inclass quiz
Review Session • May 8, 2 -3 pm as on course schedule • May have to change if social distancing is still in place