MIT 6.036: Introduction to Machine Learning
6.036, renumbered 6.3900 in the current catalog but still searched overwhelmingly by its old number, is MIT's introductory machine learning course: perceptrons, gradient descent, neural networks, and reinforcement learning basics, with the OCW version serving a huge self-study audience.
Fennie is independent and not affiliated with MIT. This is an unofficial study guide.
What makes it hard
The course compresses the mathematical core of ML (gradients, linear algebra in high dimensions, loss surfaces) into a fast weekly rhythm of exercises and labs, and students whose 18.02-level math has gone cold feel it immediately. The concepts arrive linearly but build multiplicatively: a shaky week on gradient descent makes the neural network weeks opaque.
What you'll cover
- • Linear classifiers and perceptrons
- • Gradient descent and optimization
- • Regression and regularization
- • Neural networks and backpropagation
- • Convolutional networks
- • Reinforcement learning basics
The 6.036 study guide
How to study for MIT 6.036, step by step.
- 1
Rewarm matrix calculus before week one
Gradients of vector functions, chain rule through compositions, dimensions of every product: 6.036 assumes this is fluent. A focused week of review beats a semester of low-grade confusion.
- 2
Compute one tiny example by hand per concept
Run a two-weight perceptron update or a two-layer backprop pass on paper. The hand computation is what turns the formulas from notation into machinery you trust.
- 3
Keep dimensions as your running sanity check
Annotate the shape of every vector and matrix in every derivation. Most confusion in 6.036 is dimensional confusion wearing a disguise.
- 4
Stay strictly current with the weekly rhythm
Each week's exercises assume last week's are internalized; the multiplicative structure punishes catch-up mode. The week due now always outranks perfecting the one before.
Today
Today's 6.036 plan
What a Fennie Daily Plan looks like for 6.036. Yours is built from your own syllabus and adapts every day to your deadlines and progress.
First plan free, no card required. Fennie is independent and unaffiliated with your school.
FAQ
Is 6.036 now 6.390?
Yes. The 2022 renumbering made it 6.3900 (often written 6.390). OCW materials and most online discussion still say 6.036; it's the same course lineage.
Is 6.036 hard?
It's fast and math-forward rather than deep; the difficulty is keeping 18.02-level calculus and linear algebra fluent under a weekly cadence. Students current on the math find it very learnable.
What should I know before 6.036?
Multivariable calculus, basic linear algebra, and solid Python. Of the three, comfort taking gradients matters most.
More MIT courses
6.100A: Introduction to Computer Science Programming in Python
6.100A, formerly 6.0001 (the number most search results still use), is MIT's half-semester introduction to programming in Python for students with little or no experience. The 6.0001 lectures on OpenCourseWare are among the most popular free programming courses anywhere.
6.006: Introduction to Algorithms
6.006 is MIT's core algorithms class, covering sorting, hashing, trees, graph algorithms, shortest paths, and dynamic programming, with emphasis on both rigorous analysis and Python implementation. Its OpenCourseWare lectures are a global standard for learning algorithms and prepping technical interviews.
6.046J: Design and Analysis of Algorithms
6.046J, renumbered 6.1220 in MIT's current catalog but still searched overwhelmingly by its old number, is the advanced algorithms course following 6.006: divide and conquer, randomized algorithms, amortization, network flow, approximation, and complexity. The OCW lectures are a staple for advanced self-study.
6.042J: Mathematics for Computer Science
6.042J, now numbered 6.1200J, is MIT's discrete math course for CS: proofs, induction, number theory, graph theory, counting, and discrete probability. Its OCW versions, with full lecture videos and the famous free textbook, make it one of the most-used discrete math resources in the world.