CSCA 5642: Introduction to Deep Learning
ÌýÌýPreview this courseÌýin the non-credit experience today!Ìý
Start working toward program admission and requirements right away.ÌýWork you complete in the non-credit experience will transfer to the for-credit experience when you upgrade and pay tuition. See How It Works for details.
Cross-listed with DTSA 5511
ÌýÌýImportant Update: Machine Learning Specialization ChangesÌýÌý
We are excited to inform you the current Machine Learning: Theory and Hands-On Practice with Python Specialization (taught by Professor Geena Kim) is being retired and will be replaced with a new and improved version (to be taught by Professor Daniel Acuna) that reflects the latest advancements in the field. The last opportunity to sign up for the current version will now be November 28, 2025. The new version will be available Spring 1, 2026.
- Course Type:ÌýBreadth (MS-CS) Pathway|Breadth (MS-AI)
- Specialization: Machine Learning: Theory & Hands-On Practice with Python
- Instructor:ÌýDr. Daniel Acuna
- Prior knowledge needed:Ìý
- Programming languages: Basic to intermediate experience with Python, Jupyter Notebook
- Math: Intermediate level experience with Probability and Statistics, Differential Equations
- Technical requirements: Windows or Mac, Linux, Jupyter Notebook
Learning Outcomes
- Explain the mathematical foundations of neural networks and how they learn from data.
- Train and regularize deep neural networks for effective generalization.
- Apply transformer-based and multimodal models to real-world scenarios.
- Design and apply specialized neural network architectures for images and sequences.
Course Grading Policy
Course Grading and AI Usage Policy
Check this course's reading on Assessment Expectations for what is allowed by the AI Usage Policy classification(s) listed below. If you are unsure of whether a particular use is approved, please reach out to your Course Facilitator before submitting your assignment.
| Assignment | Percentage of Grade | AI Usage Policy |
|---|---|---|
| Quizzes (5) | 40% (8% each) | Conditional |
| Programming Assignments (5) | 40% (8% each) | Conditional |
| Final Exam | 20% | No AI Use |
Course Content
Duration: 6Ìýhours
Welcome to Introduction to Deep Learning. This module builds the mathematical foundations of neural networks. Starting from linear models, you will learn about the artificial neuron and develop the mathematics of gradient descent and backpropagation. The focus is on understanding how and why neural networks work through the underlying math—covering the forward pass, loss functions, and the chain rule to show how information flows through networks and how they learn from data.
Duration: 4Ìýhours
This module focuses on training neural networks effectively. Topics include optimization algorithms, hyperparameter tuning, and regularization techniques to prevent overfitting and achieve good generalization. You will compare different optimizers like SGD, momentum, and Adam, understand how learning rate and batch size affect training dynamics, and apply weight decay, dropout, early stopping, and batch normalization.
Duration: 3.5Ìýhours
This module introduces you to convolutional neural networks (CNNs), the foundation of modern computer vision. Topics include how convolutional and pooling layers work, CNN architecture design, and practical techniques like data augmentation and transfer learning. The module covers classic architectures like VGG and ResNet and explains why CNNs outperform fully-connected networks on image data.
Duration: 3Ìýhours
This module covers sequence modeling, starting with recurrent neural networks (RNNs) and long short-term memory networks (LSTMs), then progressing to the attention mechanism—the key innovation that led to transformers. Topics include how RNNs maintain hidden states across time steps, why the vanishing gradient problem motivated LSTMs, and how attention allows models to focus on relevant parts of their input.
Duration: 3Ìýhours
This final module covers the transformer architecture, which has revolutionized deep learning across domains. Topics include BERT and GPT as encoder-only and decoder-only variants, Vision Transformers (ViT) that apply attention to images, and CLIP for multimodal learning connecting vision and language. The module emphasizes applying pre-trained models to real tasks.
Duration: 1.5 hours
Final Exam Format: In-course exam
This exam has 50 questions which cover the contents from the entire course. An 80% or higher is considered passing.
The time estimate for this exam is 90 minutes and there is a time limit of 90 minutes per attempt.
Please note that this exam allows only two attempts. You will only be able to submit once per timed attempt. Highest grade recorded.
Notes
- Cross-listed Courses: CoursesÌýthat are offered under two or more programs. Considered equivalent when evaluating progress toward degree requirements. You may not earn credit for more than one version of a cross-listed course.
- Page Updates: This page is periodically updated. Course information on the Coursera platform supersedes the information on this page. Click theÌýView on CourseraÌýbuttonÌýabove for the most up-to-date information.