Mittagsseminar (in cooperation with A. Steger, D. Steurer and B. Sudakov)

Date and Time: Thursday, November 14, 2013, 12:15 pm

Duration: 30 minutes

Location: OAT S15/S16/S17

Speaker: Nemanja Skoric

Hard problems imply efficient derandomization

"Can randomness help computation?" and "Do natural hard problems exist?" are some of the most important questions in theoretical computer science. It turns out there is a surprising connection between these two.

It is reasonable to think that simulating a probabilistic algorithm with a deterministic one, requires significant loss in efficiency. However, there is evidence that this is not true. Under widely believed assumptions, BPP = P, where BPP is the class of problems computable by randomized polynomial algorithms, with an error probability of at most 1/3. Following a series of papers by Blum. Micali, Yao, Nisan, Wigderson, Impagliazzo et al. it was shown in '97 that if there is a problem computable in deterministic time 2 ^{c n}, for some constant c, which can not be computed by a circuit of size 2 ^{d n}, for some constant d, then BPP = P.

In this talk, we will present a high-level picture of the work by Nisan and Wigderson '94, which shows a weaker statement than the one above. Namely, under the assumption that a problem computable in time 2^{c n} can not be approximated by a circuit of size 2^{d n}, BPP = P.

Upcoming talks     |     All previous talks     |     Talks by speaker     |     Upcoming talks in iCal format (beta version!)

Previous talks by year:   2024  2023  2022  2021  2020  2019  2018  2017  2016  2015  2014  2013  2012  2011  2010  2009  2008  2007  2006  2005  2004  2003  2002  2001  2000  1999  1998  1997  1996  

Information for students and suggested topics for student talks

Automatic MiSe System Software Version 1.4803M   |   admin login