Open philschulz opened 8 years ago
5) Better guidelines for late submissions. Concretely, students that submit late should notify the TA. This is particularly important for the programming peer-reviewing where other students depend on such submissions. (For the 'theory'-part the worst that can happen is a delayed correction of the homework.)
Tell us explicitly when quitting the course.
6) More distributions: in the discrete case we should add Poisson and Boltzmann. In the continuous case we need at least Gaussian and exponential.
make/gave an overview of all those distributions
7) Expand Information theory to also include maximum entropy. Maybe even make the link to statistical physics.
9) Add proof that EM maximizes likelihood and also information-theoretic formulation of EM, finish the explicit example. 13) Make description of EM formally correct: Either simplify the statistic to always be counts (but then interpret t(y) as vector of values). Also, carry out the example in the text until the very end. Otherwise, the algorithm remains unclear. 14) Introduce sufficient statistic (and then we can refer to it in the EM part)
10) Maybe start the first week with basic math (linear functions, limits, derivatives, integrals). Make an overview sheet of what is required (or find appropriate overview somewhere).
11) Also add Categorial distribution and maybe discuss types of data (nominal, ordinal, continuous).
12) Include Central Limit Theorem when discussing sample means (presupposes knowledge of Normal distribution).
1) Include continuous probability theory.
2) Offer a git tutorial at the beginning of the course that is open to all of ILLC. (watch out for technical overflow/flooding)
3) Be more explicit about the workload (tell students ahead of time how many assignments there will be). Workload will increase with time!
4) Provide solutions to the exercises in the script.