Students’ Theory Seminar - 2005


Time & Place

 

Every Wednesday, 11:00, at the Pekeris room.


 

March-April 2005

 

2/3/2005

Amir Yehudayoff

Derandomizing Polynomial Identity Tests Means Proving Circuit Lower Bounds

9/3/2005

Gidi Amir

One-Dimensional Diffusion Limited Aggregation

16/3/2005

Ronen Gradwohl

Extracting Randomness Using Few Independent Sources

23/3/2005

 

(Papadimitriou’s lecture)

30/3/2005

Iftach Haitner

Construction of a Pseudo-Random Generator From Any One-Way Function

6/4/2005

Tal Moran

Martingales and Collective Coin Flipping

 

May 2005

 

4/5/2005

Ariel Gabizon

Pseudorandom generators for low degree polynomials

11/5/2005

 

(Yom Ha’zikaron)

18/5/2005

Danny Hefetz

On Avoider-Enforcer Games

25/5/2005

Dana Moshkovitz

Lattices in Computer Science and

A Sieve Algorithm for the Shortest Lattice Vector Problem

 

June 2005

 

1/6/2005

Ori Gurel-Gurevich

Traffic Jams on the BML highway

8/6/2005

 

(OBERWOLFACH)

16/6/2005

Eyal Rozenman

Expander Flows, Geometric Embeddings, and Graph Partitionings

22/6/2005

 

(cont’d)

29/6/2005

Danny Harnik

Pseudorandomness for Network Algorithms

 

July 2005

 

6/7/2005

Zeev Dvir

Bourgain's Two-Source Extractor(s)

13/7/2005

 

 

20/7/2005

Omer Reingold

SL=L and Beyond

27/7/2005

 

(cont’d)

 

August 2005

 

3/8/2005

Guy Rothblum

One-Way Functions are Essential for Complexity Based Cryptography

10/8/2005

Elchanan Mossel

On Sensitivity and Chaos

 


 

Abstracts

 

Title:

Derandomizing Polynomial Identity Tests Means Proving Circuit Lower Bounds

Speaker:

Amir Yehudayoff

 

(Based on a paper by Valentine Kabanets and Russell Impagliazzo, 2003).

 

Arithmetic Circuit Identity Testing is a version of the Polynomial Identity Testing Problem: Given an arithmetic circuit C computing a polynomial p, decide if pº0.

In their paper they show that "derandomization of Polynomial Identity Testing is, essentially, equivalent to proving circuit lower bounds for NEXP." We will not go over all the paper, but we will, essentially, go over the quote above, i.e.,

1. Derandomization of Polynomial Identity Testing proves some sort of a lower bound.

2. Proving some lower bound is (actually) a derandomization of Polynomial Identity Testing.

 


 

Title:

One-Dimensional Diffusion Limited Aggregation

Speaker:

Gidi Amir

 

 

In a paper from 1981 Sanders and Whiten introduced the (2-Dim) DLA process --- a stochastic growth model, in which a "Discrete fractal" subset of Z2 is grown from the origin in an inductive process, where in each step a particle, starting at "infinity" makes a simple random walk on Z2 until it becomes a neighbor of the aggregate, then glues to it. The process turned out to be extremely difficult to analyze, and in spite of attracting a lot of attention, very little is known on the structure of the n-th step set or the limit set.

 

In this lecture we will define a family of 1-dimensional analogs of the standard DLA process - the 1-DLA process, which are more prone to analysis. The process is parameterized by an integer random variable X, which we will call the step distribution. The aggregates are subsets of Z built inductively by starting with the set A0 = {0} and then at each step releasing a particle "near infinity" , letting it make a random walk with jump distribution X until it hits An. If the particle first hits the set An by making a jump from xÏAn to aÎAn then we set An+1 = AnÈ{x}. (Precise definitions will be given). We will analyze this process for various choices of the step distribution, getting bounds for the growth of the diameter, and show a transient walk for whom the limit aggregate is the whole of Z despite the diameter growing super exponentially.

 

No knowledge of random walks is assumed for the lecture, I will explain the facts that I use for the proofs.

 

Note that DLA creates great pictures to hang on the wall, and is even considered a practical approximation for various areas in physics.


 

Title:

Extracting Randomness Using Few Independent Sources

Speaker:

Ronen Gradwohl

 

 

Multiple source extractors with entropy requirement d are deterministic functions that take several inputs of n bits each, such that if each input is drawn from a distribution with min-entropy dn, the output is close to uniform. We will discuss the construction of Boaz Barak, Russell Impagliazzo, and Avi Wigderson, which is the first such extractor with entropy requirement less than 1/2 and a constant number of inputs. The results are based on several results from additive number theory, and their use may be of separate interest.


 

Title:

Construction of a Pseudo-Random Generator From Any One-Way Function

Speaker:

Iftach Haitner

 

 

Pseudorandom generator, a notion first introduced by \cite{FOCS82*112} and stated in its current, equivalent, form by \cite{MR780384}, is one of the basic primitives in the study of the interaction between random and deterministic algorithms. Pseudorandom generators also have many cryptographic applications (for example \cite{Naor89, FOCS84*464, Luby:1988:HCP}). Informally, a pseudorandom generator is a polynomial-time computable function g that stretches a short random string x into a long string g(x) that “looks" random to any efficient (i.e., polynomial-time) algorithm. Hence, there is no efficient algorithm that can distinguish between g(x) and a true random string of length |g(x)| with more than negligible probability. Thus we can use g to convert a small amount of randomness into larger number of effectively random bits.

 

The first construction of pseudorandom generator was given by \cite{FOCS82*112}, their construction based on
a particular one-way function was later generalized by \cite{MR780384} into a construction of pseudorandom generator based on any one-way permutation. The question of constructing a pseudorandom generator using any one-way function remained open until Håstad et al.\ \cite{Hill} who, following \cite{lncs403*146, STOC::ImpagliazzoLL1989}, presented such a construction.

 

In the talk I give a high level overview of the Håstad et al construction.


 

Title:

Martingales and Collective Coin Flipping

Speaker:

Tal Moran

 

 

Based on an unpublished manuscript by Richard Cleve and Russel Impagliazzo (1993)

The main result  concerns martingale sequences that correspond to a situation where information about an event is gradually released; We will show that  for any martingale X0,...,Xn where X0=1/2 and Xn is either 0 or 1, with constant probability there exists i such that |Xi-Xi-1|=Ω(1/√n). This is related to (but not a direct consequence of) the "martingale tail inequality" of Azuma (1967) and Hoeffding (1963).

This result has some interesting applications. In particular, we will show that for any r-round two-party collective coin flipping protocol where bit-commitment is provided as a primitive, there must be one player that can bias the outcome of the other player by Ω(1/√r). This improves a previous lower bound (by Cleve) of Ω(1/r), and is tight for this model.

The talk will not assume knowledge of either martingales or coin flipping.

 


 

Title:

Pseudorandom generators for low degree polynomials

Speaker:

Ariel Gabizon

 

(Based on a paper by Andrej Bogdanov)

 

Let d and m be some integers, and F be some finite field.

In this work, Bogdanov shows how to construct a small subset of Fm (that can be efficiently sampled from)

Such that, for any polynomial p:Fm-> F of total degree d, when an input is chosen uniformly from this small set, the output distribution of p is close to the output distribution of p on a uniform sample in all of Fm.

 

I will sketch the construction and its proof and present the mathematical tools used.

 

 


 

Title:

On Avoider-Enforcer Games

Speaker:

Dan Hefetz (School of Computer Science, Tel Aviv University)



Let p and q be positive integers and let H be a hypergraph. In a
(p,q,H) Avoider-Enforcer game two players, called Avoider and Enforcer,
take turns selecting previously unclaimed vertices of H. Avoider
selects p vertices per move and Enforcer selects q vertices per move.
Avoider loses if he claims all the vertices of some hyperedge of H,
otherwise Enforcer loses. We prove a sufficient condition for Avoider
to win the (p,q,H) game, and use it to analyze several classic games -
connectivity, hamiltonicity and perfect matching.
Some of our results are quite surprising as they differ from those
obtained for the analogous Maker-Breaker games.

Joint work with Michael Krivelevich and Tibor Szabo

 


 

Title:

Lattices in Computer Science

and

A Sieve Algorithm for the Shortest Lattice Vector Problem

Speaker:

Dana Moshkovitz

 

 

The talk has two goals:

 

1. Introduce an area

Lattices: definitions, algorithmic problems, known results and connections to complexity and cryptography.

 

2. Present a nice algorithm

The randomized sieve algorithm of Ajtai, Kumar and Sivakumar (STOC01),

finding a shortest vector in a lattice in time 2O(n).

[this problem is NP-hard and prior to this paper there were algorithms with larger exponents in the running-time].

 

Some of the talk will be based on an excellent course given by Oded Regev at Tel-Aviv University last fall.

 


 

Title:

Traffic Jams on the BML highway

Speaker:

Ori Gurel-Gurevich

 

Initially a car is placed with probability p at each site of the two-dimensional integer lattice.

Each car is equally likely to be East-facing or North-facing, and different sites receive independent assignments.

At odd time steps, each North-facing car moves one unit North if there is a vacant site for it to move into.

At even time steps, East-facing cars move East in the same way.

This is called the BML (Biham-Middleton-Levine) Traffic model.

It is long conjectured (and confirmed by simulations) that for large enough p,

all cars are eventually jammed, while for p below some critical pc, traffic flows freely.

The first part was recently proved by Angel, Holroyd and Martin.

I am going to present their paper and some background.

 

No prior knowledge is necessary.

 

 

Title:

Expander Flows, Geometric Embeddings, and Graph Partitionings

Speaker:

Eyal Rozenman

 

We present Arora, Rao and Vazirani's paper with the same title.

One main result is a √logn approximation for the edge-expansion of a graph (exact computation is NP-hard).

The techniques used are geometric embeddings in the spirit of Goemans-Williamson, and

concentration of measure.

 


 

Title:

Pseudorandomness for Network Algorithms

Speaker:

Danny Harnik

 

 

We define pseudorandom generators for Yao's two-party communication complexity model and exhibit a simple construction, based on expanders, for it. We then use a recursive composition of such generators to obtain pseudorandom generators that fool distributed network algorithms. While the construction and the proofs are simple, we demonstrate the generality of such generators by giving several applications.

 

Based on a paper by Impagliazzo, Nisan and Wigderson from 1994.

 


 

Title:

Bourgain's Two-Source Extractor(s)

Speaker:

Zeev Dvir

 

 

A two-source extractor is a function E(x,y) that takes as input two samples from two independent weak sources, and outputs an almost uniform bit (or bits).

Very recently Bourgain gave a new construction of a two-source extractor that works when the min-entropy rate of its two inputs is larger than ~0.49.

All previous constructions require at least one of the inputs to have min-entropy rate larger than some constant > 0.5.

 


 

Title:

SL=L and Beyond

Speaker:

Omer Reingold

 

I will give some additional details on undirected connectivity in deterministic log-space (e.g., universal traversal and exploration sequences). I will also discuss some attempts "towards" proving RL=L (or even BPL=L). These attempts are joint with Luca Trevisan and Salil Vadhan.

 


 

Title:

One-Way Functions are Essential for Complexity Based Cryptography

Speaker:

Guy Rothblum

 

 

Under the assumption that (a certain kind of) one-way function exists, it is known that a wide variety of cryptographic objects and protocols can be constructed: from pseudo-random generators to zero-knowledge proofs for any efficiently provable statement (i.e. for all of IP).

 

A natural question is whether, in fact, the assumption that one-way functions exist is simply too strong. In their paper from 1989 Impagliazzo and Luby tackled this question and formalized the intuition that one-way functions are not only a “sufficient'' tool for building good cryptographic primitives, but are also a “necessary'' tool for many central cryptographic tasks.

 

In this talk I hope to survey some results from Impaggliazo and Luby's original paper and a more complex result of Ostrovsky and Wigderson (from 1993), who showed that one-way functions are essential for non-trivial zero-knowledge.

 


 

Title:

On Sensitivity and Chaos

Speaker:

Elchanan Mossel (UC Berkeley)

 

I will discuss some (very) recent results showing how techniques from the theory of Gaussian Hilbert spaces can be used to solve a number of open problems in discrete Fourier analysis.

 

Joint work Ryan O'Donnell and Krzysztof Oleszkiewicz.