Machine Learning II
This course develops the foundations needed to understand and use modern deep learning. We begin with machine-learning basics and linear models, then study multilayer perceptrons, backpropagation, optimization, convolutional and recurrent neural networks, attention and transformers, language models, and deep generative models. Throughout the course, we will connect the mathematics to implementation and examine why these methods work, where they fail, and how to evaluate them carefully.
Course information
Prerequisites. Students should be comfortable with basic linear algebra, probability, and core machine-learning concepts. You should also be able to implement and debug machine-learning experiments in Python using tools such as NumPy, scikit-learn, and PyTorch.
Assessment. In-class quizzes account for 20% of the grade; the semester-long research project accounts for 65%; and engagement and attendance account for 15%. There will be three quizzes, and the lowest quiz score will be dropped. Full policies and project requirements are in the syllabus.
Course communication. Canvas will be used for announcements, course materials, assignments, and class questions. Please use email for official or sensitive communication.
Full syllabus (PDF)Tentative schedule
Topics and ordering may be adjusted during the semester based on pacing. Assignment releases and project deadlines will be announced separately on Canvas.
| Week | Date | Topic |
|---|---|---|
| Machine-learning foundations | ||
| 1 | Thu, September 3 | Course introduction: what does it mean to learn? |
| 2 | Mon, September 7 | No class — Labor Day |
| 2 | Tue, September 8 | Supervised learning and empirical risk minimization (Monday class schedule) |
| 2 | Thu, September 10 | Generalization, validation, and evaluation |
| 3 | Mon, September 14 | Linear models for regression and classification |
| 3 | Thu, September 17 | Numerical computation, gradients, and automatic differentiation |
| Neural networks and optimization | ||
| 4 | Mon, September 21 | Multilayer perceptrons |
| 4 | Thu, September 24 | Computational graphs and backpropagation |
| 5 | Mon, September 28 | Activation functions |
| 5 | Thu, October 1 | Initialization and vanishing or exploding gradients |
| 6 | Mon, October 5 | Optimization for deep learning |
| 6 | Thu, October 8 | Regularization and generalization in deep networks |
| Vision and sequential data | ||
| 7 | Mon, October 12 | Convolutional neural networks |
| 7 | Thu, October 15 | Modern convolutional architectures |
| 8 | Mon, October 19 | Representation learning and transfer learning |
| 8 | Thu, October 22 | Recurrent neural networks |
| 9 | Mon, October 26 | LSTMs, GRUs, and sequence modeling |
| Attention and language models | ||
| 9 | Thu, October 29 | Attention mechanisms |
| 10 | Mon, November 2 | Transformers |
| 10 | Thu, November 5 | Bidirectional transformers and BERT |
| 11 | Mon, November 9 | Autoregressive language models and GPT |
| 11 | Thu, November 12 | Pretraining, fine-tuning, and adapting foundation models |
| Deep generative models | ||
| 12 | Mon, November 16 | Latent-variable models and variational autoencoders I |
| 12 | Thu, November 19 | Variational autoencoders II |
| 13 | Mon, November 23 | Generative adversarial networks |
| 13 | Thu, November 26 | No class — Thanksgiving recess |
| 14 | Mon, November 30 | Diffusion models |
| Course projects | ||
| 14 | Thu, December 3 | Final project presentations I |
| 15 | Mon, December 7 | Final project presentations II |
| 15 | Thu, December 10 | Final project presentations III |
The September 8 Monday-class designation and the November 26 recess follow the Rutgers 2026–2027 academic calendar.