Two module 3 lectures: Flow matching, normalising flows
Week 4
Module 3 help session
Assignment 3 due
Oral exam modules 1–3
Two module 4 lectures: VAEs
Week 5
Module 4 help session
Assignment 4 due
Two module 5 lectures: Diffusion models, score matching, guidance
Week 6
Module 5 help session
Assignment 5 due
Two module 6 lectures: GANs and adversarial learning
Week 7
Two module 7 lectures: Evaluation of generative models
Module 6 help session
Assignment 6 due
Module 8 lecture: Ethical and societal aspects of generative AI
Module 7 help session
Exam week 1
Assignment 7 due
Module 8 Canvas quiz due
Oral exam modules 4–7
Exam week 2
Presentation video for grades A and B due
Opportunity to re-take one oral exam
Seminars for grades A and B
Preparations before course start
Recommended prerequisites
Good programming skills (equiv. to DD1310–1319/DD1331/DD1332/DD1337/ID1018) including Python, PyTorch, Jupyter Notebooks.
Algebra and geometry (equiv. to SF1624) including vectors, matrices, systems of linear equations, inner and outer products, norms, triangle inequality, metric spaces, determinants, eigenvalues, linear dependence, subspaces, trace of a matrix.
Single-variable calculus (equiv. to SF1625) including functions, domains, ranges, monotonicity, exponentials and logarithms, limits, sequences, change of variables, convex functions, ordinary differential equations, Euler’s method.
Multivariate calculus (equiv. to SF1626/SF1674) including partial derivatives, multivariate chain rule, change of variables, gradients, Hessian and Jacobian matrices, Fourier series.
Probability theory (equiv. to SF1900–SF1935) including probability, conditional probability, Bayes’ law, independence, random variables, probability mass and density functions, samples, random sampling, expectation/mean, variance, standard deviation, median, correlation, covariance, uniform distributions, multivariate Gaussian distributions and their properties, conditional expectation, parameter estimation, maximum-likelihood estimation, consistency, change of variables, Jensen’s inequality, Markov chains, least-squares regression.
Machine learning (equiv. to DD1420/DD2421) including optimisation, loss functions, train/val/test sets, mean squared error, classification, accuracy, overfitting, Gaussian mixture models, high-dimensional geometry/curse of dimensionality, baselines, ablation studies. Information theory for machine learning including entropy, bits, cross-entropy.
Deep learning (equiv. to DD2424/DD2437) including feed-forward networks, activation functions, ReLU, softmax, CNNs, RNNs, residual networks, skip connections, U-Nets, self-attention, position embeddings, transformers, mean and variance normalisation, initialisation, hyperparameters, stochastic gradient descent, updates, epochs, dropout.
Literature
Lecture slides (and lecture recordings for some modules) on Canvas. There is no course book.
Support for students with disabilities
Students at KTH with a permanent disability can get support during studies from Funka:
LAB1 - Digital Assignment with Oral Comprehension Questions, 7.5 credits, grading scale: A, B, C, D, E, FX, F
Based on recommendation from KTH’s coordinator for disabilities, the examiner will decide how to adapt an examination for students with documented disability.
The examiner may apply another examination format when re-examining individual students.
If the course is discontinued, students may request to be examined during the following two academic years.
The section below is not retrieved from the course syllabus:
LAB1 - Digital Assignment with Oral Comprehension Questions, 7.5 credits
The following course activities contribute towards the grade on the course:
The seven notebook assignments (for grades A, B, D, and E)
The two oral exams (for grades C, D, and E)
The Canvas quiz (for grade E)
The presentation-video assignment (for grades A and B)
The in-person seminar (for grades A and B)
Grading criteria/assessment criteria
Intended learning outcomes for higher grades
For grades higher than E, in addition to the learning outcomes required for passing the course, students should furthermore be able to:
Grades D and above: Implement generative models and evaluation metrics from specifications
Grades C and above: Explain key theoretical results relevant to generative modelling
Grades A and B: Develop deep generative models for applications and motivate the design choices made
Grading criteria
For grades E and above, students must
Notebook assignments: Solve the tasks and questions marked “E”, accurately disclose any use of generative AI, and submit their entire assignments on time
Canvas quiz: Pass the mandatory Canvas quiz on time
Oral assessments: Demonstrate comprehension of the notebook solutions to mandatory tasks and questions and their implications
For grades D and above, students must additionally
Notebook assignments: Implement generative model loss functions and evaluation metrics based on mathematical specifications (marked “D” on the notebooks)
Oral assessments: Demonstrate comprehension of their own loss-function and evaluation-metric code
For grades C and above, students must additionally
Oral assessments: Re-state key theoretical results and where requested demonstrate familiarity with their significance and/or the steps used in their derivation
For grades B and above, students must additionally
Notebook assignments: Solve 1 notebook problem marked grade “A+B”
Video: Submit a short video presenting their solution
Seminar: Discuss their solution with other students at a designated seminar
For grade A, students must additionally
Notebook assignments: Solve 2 additional notebook problems marked grade “A+B”, for a total of 3 such problems solved
Opportunity to complete the requirements via supplementary examination
Grade Fx is assigned on LAB1 to students that do not meet the criteria for grade E, but meet the below criteria:
Notebook assignments: Solved the mandatory problems on all but one notebook, and accurately disclosed any use of generative AI on those notebooks and submitted them on time
This is rectified by submitting an updated version of the remaining notebook
Oral assessments: The student demonstrated comprehension of the notebook solutions to mandatory problems and their implications on all but one oral exam
This is rectified by re-taking that oral exam
Canvas quiz: Did not achieve a passing grade on the Canvas quiz before the deadline
This is rectified by re-taking the quiz
The course examiners can also, at their own discretion, assign an Fx on a case-by-case basis in other situations where the student does not meet the criteria for grade E.
Requests to raise an Fx grade may be denied unless communicated to course examiners well in advance of re-exam period 1, with all activities to raise one’s grade to be completed before the end of that re-exam period. After rectifying the issues that caused an Fx on LAB1 and notifying the course examiners of this, the student is upgraded to the highest grade for which they satisfy the grading criteria at that point, based on their best performance on any quizzes, assignments, and exams that the student has been permitted to take more than once.
Opportunity to raise an approved grade via renewed examination
All students who achieve grades E through B (inclusive) receive the same options to raise their grade after the end of the course as students who receive an Fx do. These are:
The opportunity to submit an updated version of at most one notebook assignment of their choice
The opportunity to re-take at most one oral exam of their choice
The opportunity to submit a video and subsequently attend a seminar on Zoom, if required for their new target grade
Requests to raise one’s grade may be denied unless communicated to course examiners well in advance of re-exam period 1, with all activities to raise one’s grade to be completed before the end of that re-exam period. After notifying the course examiners of having completed their renewed examination, the student is then upgraded to the highest grade for which they satisfy the grading criteria at that point, based on their best performance on any assignments and exams that the student has been permitted to take more than once.
Ethical approach
All members of a group are responsible for the group's work.
In any assessment, every student shall honestly disclose any help received and sources used.
In an oral assessment, every student shall be able to present and answer questions about the entire assignment and solution.
The section below is not retrieved from the course syllabus:
