CS 582

CS 582 - Introduction to Deep Learning

Credits: (3)

Prerequisites: MATH 440 or (MATH 260 and [STAT 200 or STAT 301])

This course introduces the foundational principles of deep learning and modern large-scale AI systems. The course covers learning as optimization, neural networks, convolutional and recurrent architectures, attention mechanisms, transformers, and large language models (LLMs). Students examine how deep learning models are trained, evaluated, and deployed, with emphasis on model behavior, inference-time control, and real-world AI system considerations. Students will cover deep reinforcement learning algorithms including deep Q-learning and policy gradient methods. Deep unsupervised models such as auto-encoders, and deep generative models including variational auto-encoders and generative adversarial networks. Encoder-decoder architectures and their applications in real-world problems such as machine translation, image captioning, visual question answering, and text summarization.

 

Detailed Description of Content of Course

• Neural Networks
  • Linear
  • Convolutional
  • Recurrent
  • Modern
  • Adversarial
• Multilayer Perceptrons
• Attention Mechanisms and Transformers
• Optimization
• Computer Vision
• Natural Language Processing
• Reinforcement learning (Q-learning/policy gradient)
• Unsupervised (variational encoders and real world usage).

 

Detailed Description of Conduct of Course

Primarily a lecture based class. Programming projects are assigned to give students experience in implementing existing algorithms. Presentations and group work may be used.

 

Student Learning Outcomes

Students will be able to:

• Explain core principles of machine learning and deep learning
• Understand learning as an optimization process
• Design and train neural networks for vision and sequence-based tasks
• Explain attention mechanisms and Transformer architectures
• Analyze how large language models generate and control outputs
• Understand practical challenges in deploying and serving LLMs
• Apply different training metholodiges (re-enforcement/unsupervised)

 

Assessment Measures

Student achievement is measured by written tests and evaluation of homework and programming assignments. Quizzes and presentations may be used.

 

Other Course Information

None.

 

Review and Approval

March 18, 2026