# Guidelines for this semester

- Attendance is mandatory for
__every__meeting; this is a requirement for receiving course credit. - During each meeting, a participant will present a paper from a specified list (incomplete sample below).
- You will be graded on
**active participation**:- no open smartphones or laptops in class
- before each lecture, you will be assigned a paper (or two) to read – the one(s) to be presented in the upcoming lecture
- following each lecture, homework questions on lecture material will be assigned
- these are
__all mandatory__– a requirement for receiving course credit - these will be turned in at the start of the
__next__lecture (no late submissions accepted) - these must be done
__individually__– not in pairs, etc - the usual academic integrity rules apply (see below)

- Grade breakdown: active participation-30%, presentation-70%

# Sample papers (incomplete list)

- Haussler, Littlestone, Warmuth. Predicting {0,1}-Functions on Randomly Drawn Points
- Haussler. Sphere packing numbers for subsets of the Boolean n-cube with bounded Vapnik-Chervonenkis dimension

[note that the lower bound in Thm. 2 has been improved in Thm. 4 of this paper ]. - Long. On the sample complexity of PAC learning half-spaces against the uniform distribution (Alaa Abd El)
- Floyd, Warmuth. Sample Compression, Learnability, and the Vapnik-Chervonenkis Dimension (Moshe Noivirt)
- Hanneke. The Optimal Sample Complexity of PAC Learning (Ahmad Droby)
- Kontorovich, Pinelis. Exact Lower Bounds for the Agnostic Probably-Approximately-Correct (PAC) Machine Learning Model
- Moran, Yehudayoff. Sample compression schemes for VC classes (Menachem Sadigurschi)
- Daniely, Shalev-Shwartz. Optimal Learners for Multiclass Problems
- Freund, Schapire. A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting;
- Freund, Schapire. Game theory, on-line prediction and boosting (Idan Attias)
- Littlestone. Learning Quickly When Irrelevant Attributes Abound: A New Linear-Threshold Algorithm
- Cover. Behavior of Sequential Predictors of Binary Sequences (Morad Muslimany)
- Mendelson, Vershynin. Entropy and the combinatorial dimension
- Alon, Ben-David, Cesa-Bianchi, Haussler. Scale-sensitive dimensions, uniform convergence, and learnability (Reem Al-asam)
- Cover, Hart. Nearest neighbor pattern classification (Borak Madi)
- Cover. On Determining the Irrationality of the Mean of a Random Variable (Yahav Azran)
- Orlitsky, Santhanam, Zhang. Always Good Turing; Asymptotically Optimal Probability Estimation (Yaara Shoval)
- Orlitsky, Suresh. Competitive Distribution Estimation (Itamar Peretz)
- Paninski. A coincidence-based test for uniformity given very sparsely-sampled discrete data (Dina Barak)
- Koltchinskii, Panchenko. Empirical Margin Distributions and Bounding the Generalization Error of Combined Classifiers
- Bartlett, Bousquet, Mendelson. Local Rademacher complexities
- Bartlett, Mendelson. Rademacher and Gaussian Complexities: Risk Bounds and Structural Results

# Academic Integrity

Cheating in university courses is regarded as a serious offense. To avoid any possible misunderstanding, please read the following carefully. Academic dishonesty includes any act of obtaining, soliciting or making available to others, material related to homework assignments. If you commit any of the above, then you are guilty of academic dishonesty. If your partner commits any of the above and you submit the assignment jointly, then you are just as guilty of academic dishonesty. If you choose to work with a partner, then you are both personally responsible for what you submit together. Claiming that you were not aware of the fact that your partner copied the assignment from somebody else will not absolve you of any responsibility.To eliminate any doubts, we make no distinction between the two (or more) sides of the cheating. If we suspect that Bob and Alice have copied an exercise one from the other, we see no way they could have done this without cooperation. It is your own responsibility to make sure that nobody can copy your assignment.

We will not tolerate academic dishonesty in this course. If you are suspected of academic dishonesty, then a complaint will be filed with the university disciplinary board (ועדת משמעת) and a detailed report placed in your academic records. The minimal penalty for this type of offense is a grade of zero in the course. You might also be expelled from the university.