The course will focus on understanding important concepts in machine learning, and will introduce the main paradigms and methods underlying modern machine learning. This will be done by understanding the mathematical formulation of statistical learning and the basics of statistical and computational learning theory, by studying specific learning algorithms, and by empirical experimentation with them.
Course Text: "Understanding Machine Learning: From Theory to Algorithms", Shalev Shwartz and Ben David, Cambridge University Press. All required material will be from the course text, or from supplemental notes posted on the website.
Prerequisites: The formal prerequisite is "236501 Introduction to Artificial Intelligence". It will be helpful to take 236501 before this class as it gives an initial introduction to learning, and we will be referring as examples to some models presented in 236501. We will also use Python, introduced in 236501. However, it is possible to take 236756 without 236501 as we will not directly rely on any material from 236501. Classes in Probability (094412 or alternatives) and Linear Algebra (104167) are very strongly recommended, and classes in Algorithms (234247) and in Numerical Analysis (234107) are also helpful.
Requirements: self-serve demos, mandatory homeworks, final exam.
Grade components: Homeworks (40%), self-serve demos (10%), final exam (50%), or just the final exam (100%), whichever is higher. But to receive a passing grade, all homeworks must be completed, each with a grade above 50/100.
Self-serve demos: Each submission is worth one point if done on time and half a point if submitted late, to a maximum of ten points. The grade for each will be binary based on submission (we reserve the right to not give credit if the answer indicates the demo was not actually done).
Homeworks: There will be 3 mandatory homework assignments. At least 50/100 in each assignment is required to pass the course. Submitted individually.
Late submission of HW: Each student is allowed one late submission without penalty. A second late submission will be counted towards the final grade as if the grade was half of the true grade(the true grade will be used to decide on the above requirment). A third late submission will contribute 0.2 of the true grade. In any case, any late submission might be returned to the students late, and possibly only partially graded. No exceptions or further extensions will be granted, except according to biding Technion rules (miluim and hospitalization).
Course Text: "Understanding Machine Learning: From Theory to Algorithms", Shalev Shwartz and Ben David, Cambridge University Press. All required material will be from the course text, or from supplemental notes posted on the website.
Prerequisites: The formal prerequisite is "236501 Introduction to Artificial Intelligence". It will be helpful to take 236501 before this class as it gives an initial introduction to learning, and we will be referring as examples to some models presented in 236501. We will also use Python, introduced in 236501. However, it is possible to take 236756 without 236501 as we will not directly rely on any material from 236501. Classes in Probability (094412 or alternatives) and Linear Algebra (104167) are very strongly recommended, and classes in Algorithms (234247) and in Numerical Analysis (234107) are also helpful.
Requirements: self-serve demos, mandatory homeworks, final exam.
Grade components: Homeworks (40%), self-serve demos (10%), final exam (50%), or just the final exam (100%), whichever is higher. But to receive a passing grade, all homeworks must be completed, each with a grade above 50/100.
Self-serve demos: Each submission is worth one point if done on time and half a point if submitted late, to a maximum of ten points. The grade for each will be binary based on submission (we reserve the right to not give credit if the answer indicates the demo was not actually done).
Homeworks: There will be 3 mandatory homework assignments. At least 50/100 in each assignment is required to pass the course. Submitted individually.
Late submission of HW: Each student is allowed one late submission without penalty. A second late submission will be counted towards the final grade as if the grade was half of the true grade(the true grade will be used to decide on the above requirment). A third late submission will contribute 0.2 of the true grade. In any case, any late submission might be returned to the students late, and possibly only partially graded. No exceptions or further extensions will be granted, except according to biding Technion rules (miluim and hospitalization).