Project Details
Projekt Print View

Derandomizing Polynomial Identity Testing and the Isolation Lemma

Subject Area Theoretical Computer Science
Term from 2010 to 2019
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 167224914
 
Final Report Year 2019

Final Report Abstract

The most spectacular result we got is certainly the parallel algorithm for bipartite perfect matching. This is considered a breakthrough result. It solves a problem that was open for more than 30 years. It prepared the ground for a lot of further results. Derandomizing the Isolation Lemma was the very ambitious title of the project. This is the technique behind our perfect matching algorithm. Maybe nobody really expected that we get so close to this goal. We considerably generalized the perfect matching result in two subsequent papers. In the first step, we showed a similar result for the matroid intersection problem, where bipartite perfect matching is a special case of. The second step was even larger: we extended the isolation approach to all problems that can be described by polytopes with totally unimodular faces. This includes matroid intersection and bipartite perfect matching as very special cases. Moreover, a lot of combinatorial problems fall into this category. The above results should not overshadow the other work that was done in the project. Another highlight was the equivalence test for read-once oblivious arithmetic branching programs. We derandomized the probabilistic algorithm for this problem. We constructed a polynomial-time algorithm in the white-box setting, and even more surprisingly, a quasipolynomial-time algorithm in the black-box setting. This was an open problem for several years. In summary, the cooperation with IIT Kanpur was very fruitful. Many exchanges of visits lead to strong collaborations.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung