CSE 525: Randomized Algorithms and Probability Spring 2026 HW3 Submit to gradescope

Instructions

  • You should think about each problem by yourself for at least an hour before choosing to collaborate with others.

  • You are allowed to collaborate with fellow students taking the class in solving the problems. But you must write your solution on your own.

  • You are not allowed to search for answers or hints on the web. You are encouraged to contact the instructor or the TAs for a possible hint.

  • You cannot collaborate on Extra credit problems

  • Solutions typeset in LATEX are preferred.

  • Feel free to use the Discussion Board or email the instructor or the TAs if you have any questions or would like any clarifications about the problems.

  • Please upload your solutions to Gradescope.

 

In solving these assignments, feel free to use these approximations:

1xex,1x1x/2,n!(n/e)n,(nk)k(nk)(enk)k

The problems of this HW are hard. To get full points it is enough to solve one of the following problems, although you are encouraged to try both.

  1. 1)

    Let G=(V,E) be an undirected graph and suppose each vV is associated with a set S(v) of 8r colors, where r1. Suppose further that for each vV and cS(v) there are at most r neighbors u of v such that c lies in S(u).

    • Prove that there exists a proper coloring of G assigning to each vertex v a color from its class S(v) such that, for any edge uv, the colors assigned to u and v are different.

    • Design a randomized polynomial time algorithms. Feel free to assume |S(v)|cr (for all v) for a constant c8.

  2. 2)

    Let 𝒜1,,𝒜𝓃 be a set of bad events defined on independent variables Z1,,Zm with corresponding dependency graph G (ignore conditions of the LLL for this problem). Note that [¬𝒜𝒾] is uniquely defined as a function of the probabilities of Zi’s. Prove or disprove: In the Moser-Tardos algorithm, for a fixed t>1 and i[n], the probability that the t-th resampled event is 𝒜i is at most [𝒜i]. Prove this or find a counterexample.