Schedule
Week 1
| Mon Feb. 09 |
Welcome! | |||||
| Course introduction. What is machine learning? How will this class work? Introduction to data as a combination of signal and noise. | ||||||
| Objectives Welcome! Theory |
Notes Welcome slides |
|||||
| Wed Feb. 11 |
Data = Signal + Noise | |||||
| Gaussian distribution, linear data with Gaussian noise. Likelihoods and log-likelihoods. Introduction to maximum likelihood estimation (MLE). | ||||||
| Objectives Theory Experimentation |
Reading Calculus review |
Warmup Log and critical points |
Notes Class notes |
|||
Week 2
| Mon Feb. 16 |
Maximum likelihood, gradients | |||||
| Calculating gradients and checking with Torch. Gradient descent for maximum-likelihood estimation in one dimension. | ||||||
| Objectives Theory Experimentation |
Warmup Partial derivatives, gradient |
Notes Class notes |
Assignments HW 1 (due 2/23) |
|||
| Wed Feb. 18 |
Higher Dimensions | |||||
| Linear-Gaussian model with many features. Matrix-vector notation for linear models. Loss function and gradient in higher dimensions. | ||||||
| Objectives Theory Implementation |
Warmup No warmup problem -- 15 minute quiz at the beginning of class. |
Notes Class notes |
||||
Week 3
| Mon Feb. 23 |
Feature Maps and Regularization | |||||
| Feature maps, overfitting, and regularization. Ridge regression and Lasso regression. | ||||||
| Objectives Theory Experimentation |
Warmup Overfitting in linear regression |
Notes Class notes |
Assignments HW 2 (due 3/2) |
|||
| Wed Feb. 25 |
More on Overfitting and Model Complexity | |||||
| Bias-variance tradeoff, double-descent. | ||||||
| Objectives Theory Experimentation |
Warmup Practice with expectation and variance |
Notes Class notes |
||||
Week 4
| Mon Mar. 02 |
Introducing Classification | |||||
| Predicting binary labels with logistic regression | ||||||
| Objectives Theory Experimentation |
Warmup Gradient of signal function in logistic regression |
Notes Class notes |
Assignments Miniproject 1: Bikeshare Demand Prediction (due 3/9) |
|||
| Wed Mar. 04 |
Classification with Multiple Classes | |||||
| Extending logistic regression to multiple classes. | ||||||
| Objectives Theory Implementation |
Warmup No warmup problem -- 15 minute quiz at the beginning of class. |
Notes Class notes |
||||
Week 5
| Mon Mar. 09 |
Assessment of Classification Models | |||||
| How to tell if your classifier is working. | ||||||
| Objectives Experimentation |
Warmup Gut checks for classification models |
Notes Class notes More on Assessing Classifiers |
Assignments HW 3 (due 3/16) |
|||
| Wed Mar. 11 |
Decision Theory in Classification | |||||
| Using classifiers to make decisions, or not. | ||||||
| Objectives Experimentation |
Warmup Thresholds for decision-making |
Notes Class notes |
||||
Week 6
| Mon Mar. 16 |
Automatic Differentiation | |||||
| Teaching the computer to do our calculus for us. | ||||||
| Objectives Theory Implementation |
Warmup Automatic differentiation |
Notes Class notes |
Assignments Miniproject 2: Music Genre Prediction (due 3/30) |
|||
| Wed Mar. 18 |
Modern Optimization and Stochastic Gradient Descent | |||||
| Gradient descent and its limitations. Momentum. Stochastic gradient descent for large datasets. | ||||||
| Objectives Theory Experimentation |
Warmup Prep for talking with Dr. David Rolnick |
Notes Class notes |
||||
Break
| Mon Mar. 23 |
Spring break! No class. | |||||
| Wed Mar. 25 |
Spring break! No class. | |||||
Week 7
| Mon Mar. 30 |
Visit by David Rolnick (McGill) on machine learning and climate change. | |||||
| We have the opportunity to chat with David Rolnick, a professor at McGill who works on the intersection of machine learning and climate change. David is also giving a public lecture at 4:30pm today. | ||||||
| Objectives Social Impacts |
Warmup Please review your question that you proposed to discuss with Dr. Rolnick, and be prepared to ask it during class. |
Notes NA |
Assignments Please consider attending Dr. Rolnick's public lecture at 4:30pm today! |
|||
| Wed Apr. 01 |
Midterm Exam | |||||
| No class today -- we'll have an in-class cumulative midterm exam instead. | ||||||
Week 8
| Mon Apr. 06 |
Introducing Deep Learning | |||||
| We try learning the feature maps in a classification model, which leads us to deep learning. | ||||||
| Objectives Theory Implementation Navigation |
Warmup TBD |
Notes Class notes |
||||
| Wed Apr. 08 |
Convolutional Neural Networks and Image Classification | |||||
| We study convolutional neural networks as an architecture for feature extraction from image data. | ||||||
| Objectives Theory Implementation Experimentation Navigation |
Warmup TBD |
Notes Class notes |
||||
Week 9
| Mon Apr. 13 |
Some Practical Considerations in Deep Learning | |||||
| We consider some practical considerations that often arise in deep learning, such as data sourcing, data augmentation, organization of data sets on disk, and data transformations. We apply these techniques to a problem of audio classification. | ||||||
| Objectives Implementation Experimentation Navigation |
Warmup TBD |
Notes Class notes |
Assignments Miniproject 3: Identifying Bird Calls (due 3/9) |
|||
| Wed Apr. 15 |
No class (Phil is traveling for research, hello from Boulder, CO! ) | |||||
Week 10
| Mon Apr. 20 |
Tokenization and Next-Token Prediction | |||||
| Objectives Implementation Experimentation Navigation |
||||||
| Wed Apr. 22 |
Introducing Transformers | |||||
| Objectives Implementation Experimentation Navigation |
||||||
Week 11
| Mon Apr. 27 |
More Transformers | |||||
| Objectives Implementation Experimentation Navigation |
||||||
| Wed Apr. 29 |
Bias in Language Models | |||||
Week 12
| Mon May. 04 |
TBD | |||||
| Objectives Implementation Experimentation Navigation |
||||||
| Wed May. 06 |
TBD | |||||
No matching items