A Case Study of De-randomization Methods for Combinatorial Approximation Algorithms

We study three different de-randomization methods that are often applied to approximate combinatorial optimization problems. We analyze the conditional probabilities method in connection with randomized rounding for routing, packing and covering integer linear programming problems. We show extensions of such methods for non-independent randomized rounding for the assignment problem. The second method, the so-called random walks is exemplified with algorithms for dense instances of some NP problems. Another often used method is the bounded independence technique; we explicit this method for the sparsest cut and maximum concurrent flow problems.