Introduction to Machine Learning

Fall 2026 • American University of Beirut

Course Schedule

Introduction to Machine Learning • Fall 2026

Introduction

▼
Sep 1
1
Introduction and Logistics
Why build intelligent machines, a brief history of AI, and how this course works
75 min Lecture
Course introduction covering learning objectives, expectations, and motivation. Exploration of AI's transformative impact on society and discussion of ethical implications in AI development and deployment.
Week note for this session
How to study this session
  • Explore the logistics and schedule pages on www.intro2ml.com
  • reread the Introduction and Logistics slides.
  • Set up Python per the logistics page for next Tuesday.
Preparation
Lecture
References
Homework
Sep 3
2
Data, Models, and Python
75 min Lecture
Week note for this session
How to study this session
  • Problem Set 0 is posted here: https://intro2ml.com/materials/assignments/hw0/
  • Think of Data and Models you encounter every day.
Lecture
References
References
  • Data Visualization Example (coming soon) Recommended
Sep 8
3
Math and Python Review - Your Computational Toolkit
Linear algebra and its computational implementation
75 min Lecture
Comprehensive review of mathematical foundations essential for machine learning, combined with hands-on Python implementation. You will bridge the gap between mathematical notation and computational implementation through interactive notebooks that combine linear algebra concepts with NumPy code examples.
How to study this session
  • Review the math and python review notebooks
  • Test your knowledge with the in-slide quizzes/questions
Preparation
Lecture
References

Function Approximation

▼
Sep 10
4
Supervised Learning and Linear Regression
First machine learning algorithm and foundational concepts
75 min Lecture
Introduction to supervised learning paradigm with linear regression as the first ML algorithm. Mathematical foundations, implementation, and practical applications with hands-on coding.
How to study this session
  • Read CS229 notes Part I, sections 1.1-1.2
  • Go through the interactive linear regression slides and explore each widget
  • Watch the videos in the references if you're struggling with some concepts
  • Start PS1 (posted today)
Preparation
Lecture
References
Sep 15
5
Feature Engineering and Generalization
From raw data to meaningful features, avoiding overfitting
75 min Lecture
Deep dive into feature engineering techniques and understanding generalization. Covers bias-variance tradeoff, cross-validation, and regularization methods.
How to study this session
  • Review generalization slides
  • Continue PS1
Lecture
References
Sep 17
6
Quiz 1, then Feature Engineering in Practice
The coding half of lecture 5: scikit-learn, train/test splits and choosing a model
75 min Lecture
Quiz 1 on the first four sessions, then the practical half of feature engineering and generalization: fitting, validating and comparing models in scikit-learn.
Before class
  • Quiz 1 covers sessions 1 to 4: supervised vs unsupervised learning, linear models in vector form, the train/validation/test split
  • Have Python and scikit-learn working on your machine; the video tutorial shows the setup
Takeaways
  • In scikit-learn every model is fit, then predict; the feature engineering happens before fit
  • The held-out set is the only honest estimate of how a model will do on new data
  • Model choice is an experiment: compare validation error across candidates, do not guess
How to study this session
  • Redo the coding walkthrough from the slides in your own notebook, with a dataset of your choice
  • Watch the scikit-learn video tutorial, then try Module 1 of the Inria MOOC
  • Continue PS1 (due Sep 23)
Lecture
References
Sep 22
7
Classification and Logistic Regression
From regression to classification with probabilistic models
75 min Lecture
Introduction to classification problems using logistic regression. Covers sigmoid function, maximum likelihood estimation, and decision boundaries.
Takeaways
  • Covered in class up to the two-feature decision boundary; softmax, metrics and the hands-on continue on Sep 24
How to study this session
  • Go through the interactive logistic regression slides
  • Compare the MSE vs cross-entropy loss surfaces in the slides
  • PS2 Problems 3 and 4 are on this material (due Oct 8)
Preparation
Lecture
Sep 24
8
Classification II: Softmax, Metrics and a Hands-on Dataset
The rest of logistic regression, many classes, and your first classification analysis
75 min Lecture
Second half of the logistic regression deck: decision boundary geometry and margin, precision and recall, softmax regression for many classes. Then a short hands-on exercise: a dataset, a question, and your own analysis.
Before class
  • Reread the first half of the logistic regression deck up to the two-feature boundary; the session starts there
  • Have Python and scikit-learn running: the hands-on part is you fitting a classifier, not watching one
Takeaways
  • The decision boundary is where the score is zero; the margin is how far the closest point sits from it
  • Accuracy is the wrong compass when classes are imbalanced; report precision and recall at the threshold you deploy
  • Softmax regression is logistic regression with one score per class, normalised by the exponentials
  • Analysing a dataset means: look at it, pick a question, fit, hold out, and read the confusion matrix
How to study this session
  • Finish the hands-on analysis at home if you did not in class; compare accuracy, precision and recall on a held-out split
  • Derive the gradient of the cross-entropy loss yourself, then check it against the slide
  • Try the same dataset with sklearn's LogisticRegression and compare with your own gradient descent
Preparation
Lecture
References

Probabilistic Reasoning

▼
Sep 29
9
Maximum Likelihood Estimation and Generalized Linear Models
Foundations of probabilistic reasoning and statistical learning
75 min Lecture
Deep dive into maximum likelihood estimation and generalized linear models, at the foundation of statistical learning theory.
How to study this session
  • Review the MLE slides and rederive least squares from the Gaussian likelihood, and cross-entropy from the Bernoulli one
  • Read CS229 Sections 1.3 and 2.1; Chapter 3 on GLMs if you want the general picture
  • PS2 is posted (due Oct 8)
Preparation
Lecture
References
Oct 1
10
Generative Learning and Language
Working with Language and the Naive Bayes Algorithm
75 min Lecture
Working with language, inferring algorithms from space v. non-spam, and the Naive Bayes Algorithm as an example for generative learning
How to study this session
  • Review the generative learning slides: Naive Bayes on the spam messages, GDA on the penguins, and the n-gram language model
  • Continue PS2 (due Oct 8)
  • Run the Naive Bayes from scratch notebook from the slides on your own text
Preparation
Lecture
References

Deep Learning

▼
Oct 6
11
Introduction to Neural Networks
From linear models to deep networks
75 min Lecture
Where the neuron came from (McCulloch and Pitts, Hodgkin and Huxley), the linear model drawn as a network, activations, XOR and universal approximation, what a network learns from 20,000 handwritten digits, a network in numpy and PyTorch, and a first look at training by gradient descent and backpropagation.
Takeaways
  • A layer is a matrix followed by a squash. Without the squash, a stack of layers collapses into a single matrix, so it is no better than linear regression.
  • A hidden layer is a learned feature map: the network chooses the features instead of you, and what it chooses rarely looks like a feature a person would name.
  • One hidden layer can approximate any continuous function; depth does it with fewer units, by building features of features.
  • On real handwritten digits a network beats a flat boundary only when it has enough data, and a fully connected layer has no idea that nearby pixels belong together: shift every digit by two pixels and both models fail.
  • Backpropagation is the chain rule run backwards: one forward pass stores every activation, one backward pass reuses them to get every gradient.
  • Covered in class up to slide 24 of the neural networks deck; the rest, with the handwritten digits example, continues on Oct 8
How to study this session
  • Go back through the slides and redo the interactive parts: build AND, OR, NAND and NOR with the McCulloch-Pitts neuron, watch the network bend its boundary around XOR, and draw a curve for the one-hidden-layer network to approximate
  • Run "A network in twelve lines" and "Run it on the digits" from the slides, then the Python Neural Network Example notebook
  • Read CS229 notes, Chapter 7, Sections 7.1 and 7.2
  • Finish PS2 (due Thursday 8 October, 23:59)
Preparation
Lecture
References
Oct 8
12
Backpropagation, Computation Graphs and Hands on
A deeper dive into the inner working of deep learning
75 min Lecture
The chain rule by hand and on a computation graph, backpropagation from scratch in numpy on real digits, the same network in PyTorch, the symmetry exercise, and what to check when training goes wrong.
Lecture
References
Oct 13
13
Advanced Neural Networks
Advanced Neural network architectures
75 min Lecture
Advanced deep learning architectures from fully connected networks, to autoencoders, recurrent neural networks, convolutional neural networks, etc. - snapshots from CS231n
Preparation
References
Oct 15
14
CNNs, RNNs and AI Assistants
How to use AI coding assistants in the context of deep learning
75 min Lecture
How to use AI coding assistants in the context of deep learning
Lecture

Midterm

▼
Oct 20
15
Midterm Review
A story-format review of the first half
75 min Review
A story-format review of the first half.
Oct 22
16
Midterm Exam
In class, 75 minutes
75 min Exam
In class, 75 minutes.

Unsupervised Learning

▼
Oct 27
17
Introduction to Unsupervised Learning & K-Means Clustering
Introduction to clustering and unsupervised pattern discovery
75 min Lecture
Introduction to clustering algorithms with focus on K-means. Covers cluster analysis, centroid-based clustering, and evaluation metrics for unsupervised learning.
Preparation
References
Oct 29
18
Principal Component Analysis
Principal component analysis and Dimensionality Reduction
75 min Lecture
Dimensionality reduction. Focus on Principal Component Analysis (PCA), SVD, and applications to data compression and visualization.
Lecture
References
Nov 3
19
EM Algorithm and Gaussian Mixture Models
Probabilistic clustering and the EM algorithm
75 min Lecture
Advanced clustering using Gaussian Mixture Models and the Expectation-Maximization algorithm. Covers probabilistic clustering, latent variable models, and maximum likelihood estimation.
Lecture
References
Nov 5
20
Other Unsupervised Learning Methods
Hierarchical clustering, Autoencoders, Kernel PCA and others
75 min Lecture
Survey of additional unsupervised learning techniques including hierarchical clustering, density-based clustering (DBSCAN), manifold learning, and dimensionality reduction methods.
Lecture
Nov 10
21
Story Session: The Rope as an Instrument
Unsupervised discovery on real physical data
75 min Story
Unsupervised discovery on real physical data.
Lecture

Reinforcement Learning

▼
Nov 12
22
Reinforcement Learning I
Introduction to RL, MDPs, and value iteration
75 min Lecture
Introduction to reinforcement learning paradigm. Covers Markov Decision Processes (MDPs), value functions, and dynamic programming solutions including value and policy iteration.
Preparation
References
Nov 17
23
Reinforcement Learning II
Introduction to RL, MDPs, and value iteration
75 min Lecture
Introduction to reinforcement learning paradigm. Covers Markov Decision Processes (MDPs), value functions, and dynamic programming solutions including value and policy iteration.
Preparation
Lecture
References

Advanced Topics

▼
Nov 19
24
Foundation Modeling
Large-scale models and transfer learning
75 min Lecture
Introduction to foundation models - large-scale pre-trained models that can be adapted to various downstream tasks. Covers transfer learning, fine-tuning, and model scaling.
References
Nov 24
25
Machine Learning for Modeling Physical Systems
How does ML apply to science and engineering where some physics-based models are already known?
75 min Lecture
Modeling scientific systems with machine learning' '
References

Projects and Closing

▼
Nov 26
26
Closing Story Session, Review and Exam Prep
The course in one arc; then review and exam prep
75 min Review
The course in one arc. Flipped Q and A session.
Dec 1
27
Project Presentations I
Final project presentations
75 min Project
Final project presentations.
Dec 3
28
Project Presentations II and Course Close
Final presentations and farewell
75 min Project
Final presentations and farewell.

This schedule is dynamically generated from lecture metadata. Materials and links are updated as they become available.