Skip to main content

SC4001 - Neural Networks and Deep Learning

Lecture Clarity:
(4/5)
Content Relevance:
(5/5)
Content Difficulty:
(4/5)
Overall Workload:
(4/5)
Team Dependency:
(4/5)

Course Summary​

This course introduces students to Neural Networks and their applications in problem-solving. The first half is taught by Prof Alvin, while the second half is led by Prof Chen Change Loy, a highly cited researcher at MMLab@NTU. Prof Loy offers deep insights into the material, especially since parts of the course relate closely to his research areas. The topics covered include:

  1. Neurons and Activation Functions
  2. Regression and Classification
  3. Neuron Layers
  4. Deep Networks
  5. Model Selection
  6. Convolutional Neural Networks (CNNs)
  7. Recurrent Neural Networks (RNNs)
  8. Attention
  9. Autoencoders
  10. Generative Adversarial Networks (GANs)
  11. Diffusion Models

Workload​

The workload is fairly manageable overall but becomes heavier toward the end of the semester due to the project. Prof Jagath provides tutorial answers along with sample code, which is helpful for reviewing how solutions are derived. There are no tutorials for the second half of the course, but example materials are provided.

Projects​

There are two projects in this course. The first involves designing simple neural networks to solve basic classification and regression tasks. This project is straightforward and shouldn’t take long if done carefully. Since the dataset changes every year, some questions may be confusing, but TAs are available to clarify any doubts. The second project is more demanding and self-driven. You must choose a project from a provided list or propose your own. You can build models in any deep learning field, as long as your project demonstrates sufficient depth and your model shows adequate complexity.

Tips to Do Well​

The final exam requires a lot of manual matrix multiplication, so it’s important to get comfortable using your calculator and entering values quickly. Time will be tight, so plan your approach carefully. Questions from the second half of the course tend to be more qualitative, making them easier to score. Also, consider which other modules you take alongside this one—project-heavy semesters can consume more time than expected since training models is a slow process.

Written by BS