CS-E4710 - Machine Learning: Supervised Methods D, 08.09.2020-18.12.2020
This course space end date is set to 18.12.2020 Search Courses: CS-E4710
Topic outline
-
Note: All sessions of the course are online, there are no physical lectures or exercise sessions!
Contents
Generalization error analysis and estimation; Model selection; Optimization and computational complexity; Linear models; Support vector machines and kernel methods; Boosting; Feature selection and sparsity; Multi-layer perceptrons; Multi-class classification; Ranking; Multi-output learning
Course position and Prerequisites
Course is MSc course in Machine learning, targeted to 1st year MSc students in CCIS and Life Science Technologies programmes.
The course assumes basic background in computer science and statistics, as follows:
- CS-C3190 Machine Learning, or MS-C1620 Statistical inference, or equivalent knowledge
- Basics of probability theory
- Basic linear algebra
- Programming skills (Python preferable )
Learning OutcomesAfter the course, the student knows how to recognize and formalize supervised machine learning problems, how to implement basic optimization algorithms for supervised learning problems, how to evaluate the performance supervised machine learning models, and has understanding of the statistical and computational limits of supervised machine learning, as well as the principles behind commonly used machine learning models.
Course schedules
Note: All sessions of the course are online, there are no physical lectures or exercise sessions.
- Lectures (online): Tuesdays 10:15-12:00, streamed online and recorded (See the tab Streaming/Recording). Attending the lectures is voluntary.
- Assignments : completed at home, and submitted online (See the tab Assignments).
- Tutorial sessions: Fridays 10:15-12:00. The sessions alternate between
- Q&A sessions (help for solving the exercises). We will organize the Question & Answer sessions as chat sessions, where the submitted questions (to the "General discussion" section) will be answered by text during the already specified schedule. Please remember to submit your questions 24 hours ahead of the session and avoid submitting repetitive questions. Attending the Q&A sessions is voluntary.
- Solution sessions (presenting the solutions for the exercise set). Attending the solution sessions is voluntary.
- Exam (online): 18.12.2020, 13-16. The exam will be open book.
Course personnel- Lecturer: Prof. Juho Rousu
- Course assistants: Dr Sandor Szedmak, Dr Maryam Sabzevari, Dr Riikka Huusari
Grading
The course can be completed by two alternative ways:
- Exercises (max 30 points) + Exam (max 70 points) , giving a grade 0..5. Lowest passing points total is 50. 85 points will give the grade of 5.
- Exam only (max. 100 points), giving a grade 0...5. 50 points will give the grade 1, 85 points will give the grade of 5.
The better of the resulting two grades will be taken into account.
Language of Instruction
English
Course Material
Lecture slides and exercises are the examined content
Additional reading
The lectures are mostly based on the books:
- Shalev-Shwartz, Ben-David: Understanding Machine Learning, Cambridge University Press. Downloadable for personal use from https://www.cs.huji.ac.il/~shais/UnderstandingMachineLearning/courses.html
- Mohri, Rostamizadeh, Talwakar: Foundations of Machine Learning. Downloadable from https://cs.nyu.edu/~mohri/mlbook/
-
Contents
There will be six sets of assignments, to be completed as home work, according to the schedule below. The exercises are mixtures of pen-and-paper and computational exercises. The answers to the assignments are submitted by answering to a my courses Quiz below by the indicated deadline.
There are some information on how to submit an assignment (quiz) in "https://docs.moodle.org/38/en/Using_Quiz", which can be helpful. Please remember to submit your answers once you have finished your attempts to have your quiz scored.
Maximum of 30% of total course points are available through completing the assignments.
Solution sessions
In a solution session, the solutions to the exercises are presented by the course assistant. Attending the solution sessions is voluntary.
Q & A sessions
If you have questions on how to complete the assignments we organize Question & Answer sessions (see schedule below) where you can ask the course assistant. Attending the Q & A sessions is voluntary.
Note that the questions should be submitted to the "general discussion" section 24 hours before the Q & A session. In the "general discussion" section for each question of an exercise, a topic is created, for example: "EX1-QS2" a topic created for question 2 of the first exercise. You can submit your doubts about a specific question by replying to that particular question's created topic.
Schedule
Assignments are published according to the schedule below. Submission deadlines are absolute and late submissions will not be taken into account.
The answers are submitted by filling in the the Quiz below.
Assignment Publication date Q&A session Submission deadline Solution session Solutions published 1 8.9.2020 10:00 11.9.2020 10:00 16.9.2020 23:59 18.9.2020 10:00 18.9.2020 10:00 2 18.9.2020 17:00 25.9.2020 10:00 30.9.2020 23:59 2.10.2020 10:00 2.10.2020 10:00 3 2.10.2020 17:00 9.10.2020 10:00 14.10.2020 23:59 16.10.2020 10:00 16.10.2020 10:00 4 26.10.2020 17:00 30.10.2020 10:00 4.11.2020 23:59 6.11.2020 10:00 6.11.2020 10:00 5 6.11.2020 17:00 13.11.2020 10:00 18.11.2020 23:59 20.11.2020 10:00 20.11.2020 10:00 6 20.11.2020 17:00 27.11.2020 10:00 2.12.2020 23:59 4.12.2020 10:00 4.12.2020 10:00
-
- Lecture 1 (8.9.2020): Introduction
- Lecture 2 (15.9.2020): Statistical learning theory
- Lecture 3 (22.9.2020): Learning with infinite hypothesis classes
- Lecture 4 (29.9.2020): Linear classification
- Lecture 5 (6.10.2020): Support vector machines
- Lecture 6 (13.10.2020): Kernel methods
- Lecture 7 (27.10.2020): Neural networks
- Lecture 8 (3.11.2020): Ensemble learning
- Lecture 9 (10.11.2020): Feature engineering
- Lecture 10 (17.11.2020): Multi-class classification
- Lecture 11 (24.11.2020): Preference learning
- Lecture 12 (1.12.2020): Predicting multiple and structured labels
-
The links to the live streaming will appear on this page shortly before the session starts.
Period I
Week 37 (7.9-11.9.2020)
- Lecture:
Week 38 (14.9-18.9.2020)
- Lecture 2
The lecture 2 will be retaken on Friday 18.9.2020 at 10:15 in Zoom (in place of the “Solution session which will be postponed to Friday 25.9.2020). Join the zoom meeting for the lecture at
https://aalto.zoom.us/j/62638834523
Recording: scroll down towards the bottom of the page
Week 39 (21.9-25.9.2020)
- Machine Learning: Supervised Methods - Zoom lecture 22.09.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
- Recording: scroll down towards the bottom of the page
- Solution session: Zoom 25.09.2020. 10:00-11:00
- https://aalto.zoom.us/j/69111424244
Week 40 (28.9-2.10.2020)
- Machine Learning: Supervised Methods - Zoom lecture 29.09.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
- Solution session: Zoom 2.10.2020 10:15->
- https://aalto.zoom.us/j/63747157605
Week 41 (5.10-9.10.2020)
- Machine Learning: Supervised Methods - Zoom lecture 6.10.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
Week 42 (12.10-16.10.2020)
- Machine Learning: Supervised Methods - Zoom lecture 13.10.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
- Solution session, Zoom, 16.10.2020 10:15->
- https://aalto.zoom.us/j/69429946124
Period II
Week 44 (26.10.-30.10.2020)
- Machine Learning: Supervised Methods - Zoom lecture 27.10.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
Week 45 (2.11.-6.11.2020)
- Machine Learning: Supervised Methods - Zoom lecture 3.11.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
- Solution session, Zoom, 6.11.2020 10:15->
- https://aalto.zoom.us/j/62286511390
Week 46 (9.11.-13.11.2020)
- Machine Learning: Supervised Methods - Zoom lecture 10.11.2020 10:15-12:00
Week 47 (16.11.-20.11.2020)
- Machine Learning: Supervised Methods - Zoom lecture 17.11.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
- Solution session - Zoom 20.11.2020 10:15
- https://aalto.zoom.us/j/62976982428
Week 46 (23.11.-27.11.2020)
- Machine Learning: Supervised Methods - Zoom lecture 24.11.2020 10:15-12:00
- https://aalto.zoom.us/j/61064776491
Week 47 (30.11.-4.12.2020)
- Lecture
- Solution session - Zoom 4.12.2020 10:15->
- https://aalto.zoom.us/j/66593403962
-
The exam will appear here when the exam begins.
About the exam
The exam is an open book exam, thus all course material and computational tools can be used. Downloading the lecture slides prior the exam is recommended to reduce load on the web server.
The exam consists of 9 questions in total worth 70 points:
- one essay style question that will give maximum of 30 points
- 8 multiple choice questions each giving maximum of 5 points, in total maximum of 40 points
For the final grade of the course the maximum of the following two cases is taken into account:
- total points = exam points + exercise points + one point from answering the course feedback questionnaire
- total points = exam points / 0.7, rounded up to the nearest integer + one point from answering the course feedback questionnaire
Grades will be published by 23.3.2021Answering the multiple choice questions
The questions are randomized by question categories so that every student will receive a randomly chosen question of the category. The randomly chosen question variants are equivalent in their difficulty.
You have one attempt of the quiz. The correct answer will not be shown.
It is not allowed to make the exam in groups, everyone should complete their exam independently. Sharing solutions between students is prohibited.
Answering the essay question
Please note that directly copying text from a source without citation constitutes plagiarism which is forbidden. Write in your own words, not cut-and-paste from any source.
Hint: do not simulteneously read a source and write the answer. Read sources first and then write in your own words without looking at the sources. This way your text will be less likely to be a copy of the source text.
The essays will be checked against plagiarism, both between students in the exam and internet sources.
The essay is implemented as a Turnitin Assignment. By clicking the assignment you will be shown the submission inbox for your assignment, where you will see the essay topic in the description. The essay can be submitted there by selecting "submit a paper". The essay can be written either by using a external text editor or directly to the available text box. You need to accept to the Turnitin agreement to submit the essay.
When using an external text editor, one of the templates provided below (.tex,.docx,.odt) should be used. There is a length limit of one page for the essays written by external editor. Any part exceeeding the page limit will not be graded. Do not change the font size, paragraph spacing or margins of the template. For the text box submissions, note that there is a hard limit of 3700 characters (with spaces) that you should keep track of yourself; the text box will not show the character count. Any text exceeding this limit will not be graded.
- Exam 23.2.2021, 17.00-20.00
-
-