Students’ Theory Seminar - 2006

About ~ Previous Years

 


Time & Place

 

The seminar is on Sundays, 2pm, at Pekeris (unless otherwise is specified).

 


Previous Months

 

November

 

7/11/2005

Ariel Gabizon

Isomorphism of Graphs with Bounded Eigenvalue Multiplicity

14/11/2005

Dana Moshkovitz

More On Irit’s PCP Construction

21/11/2005

Zeev Dvir

Extractors for a Constant Number of Polynomial Min-Entropy Independent Sources

28/11/2005

Amir Yehudayoff

Estimation of a certain exponential sum arising in complexity theory

 

 

December

 

13/12/2005 [Tuesday!]

Adam Smith

Evolving Notions of Security for Quantum Protocols

19/12/2005

Elad Verbin

Markov-Chain Monte-Carlo (MCMC) Methods in Theoretical

Computer Science: An overview

26/12/2005

 

Hanuka

 

 

January

 

2/1/2006

 

Department Seminar: A talk by Julia Kempe

9/1/2006

 

Department Seminar: A talk by Eli Upfal

16/1/2006

Hovav Shacham

Waters Signing for Fun and Profit

23/1/2006

Guy Rothblum

Cryptography in NC0

30/1/2006

Ronen Gradwohl

The Influence of Variables on Boolean Functions

 

 

February

 

6/2/2006

Ariel Gabizon

Recent advances in List Decoding

13/2/2006

Gidi Amir

Biased Graph Processes and The Emergence of The Giant Component

20/2/2006

[at 1pm!]

Ariel Yadin

Everyday Inequalities

 

 

March

 

2/3/2006

[Thursday!]

Omer Reingold

Tutorial on Black-Box Separations

6/3/2006

Emmanuelle Lebhar

 

A doubling dimension threshold O(loglog n) for augmented graphs navigability

13/3/2006

 

 

Ta’anit Esther

19/3/2006

[Sunday!]

Moni Naor,

Danny Harnik

How to prove that Minicrypt=Cryptomania (in the future)

[Note: The talk will be at room 5 in Feinberg!]

26/3/2006

Hovav Shacham

 

Finding small roots of modular polynomials

 

 

April

 

2/4/2006

Zeev Dvir

Semi-direct product in groups and Zig-zag product in graphs: connections and applications

6/4/2006 [Thur, 4pm!]

Mira Meyerovich

Provably Secure Steganography with Imperfect Sampling

16/4/2006

 

Passover

23/4/2006

 

 

30/4/2006

Dana Moshkovitz

Reducing Randomness for Sampling – Part I

 

May

 

7/5/2006

Dana Moshkovitz

Reducing Randomness for Sampling – Part II

14/5/2006

 

Bourgain’s talk (Jerusalem)

21/5/2006

Noam Livne

Proving Lower Bounds for Secret Sharing Schemes Using the Entropy Function

28/5/2006

Dana Moshkovitz

STOC’06 Report – Part I

 

June

 

4/6/2006

Dana Moshkovitz

STOC’06 Report – Part II

11/6/2006

Amir Yehudayoff

On the Efficiency of Local Decoding Procedures for Error-Correcting Codes

18/6/2006

 

 

25/6/2006

 

 

 

July

 

2/7/2006

Hovav Shacham

Blind Signatures and Unique Zero Knowledge

 

 

 

16/7/2006

[room 261]

Anup Rao

Deterministic extractors for small space sources

24/7/2006 [Monday!]

Ariel Gabizon

Stepanov's elementary method for proving Weil’s bounds

30/7/2006

Or Meir

Graph Isomorphism for Planar Graphs

 

August

 

6/8/2006

Oded Goldreich

Pseudorandomness -- an overview

 

13/8/2006

Oded Goldreich

Pseudorandomness -- an overview (cont’d)

 

September

 

17/9/2006

Michael Schapira

Truthful Randomized Mechanisms for Combinatorial Auctions

 

24/9/2006

 

Rosh Ha’Shana

 


Abstracts

 

 

Title:

Isomorphism of Graphs with Bounded Eigenvalue Multiplicity

Speaker:

Ariel Gabizon

 

The graph isomorphism problem is one of the most interesting problems in NP whose complexity is

undetermined. It is believed not to be NP-Complete and it is not known whether it is in CO-NP or P.

However, some very nice algorithms have been discovered for restricted classes of graphs. Maybe the best

example is that of Luks (80) which shows how to test isomorphism in polynomial time when the degrees of the

vertices are bounded by a constant. In this talk we show how to solve the following problem:

Let A and A' be the adjacency matrices of graphs G and G' on n vertices. Both A and A' have n eigenvalues

which are not necessarily distinct. If no specific eigenvalue appears more than c times, for some constant c, then

we can check in polynomial time whether G and G' are isomorphic.

 

The paper is by Babai, Grigoryev and Mount from 1982.

 


 

Title:

More On Irit’s PCP Construction

Speaker:

Dana Moshkovitz

 

The PCP Theorem [AS92,ALMSS92] is one of the great discoveries of complexity theory.

It argues that all NP languages have Probabilistically Checkable Proofs,

that is, proofs that can be efficiently (probabilistically) verified by reading only a constant number of their symbols (picked in some random manner).

 

A few months ago, Irit Dinur from the Hebrew University published a new proof for the PCP Theorem [Dinur05].

This is a truly beautiful proof, considerably simpler than previous ones.

Also, it can be used to obtain the shortest PCPs known to date.

 

We’ll outline the entire construction, give the appropriate intuitions and background,

and focus on the composition step.

This step is not the novelty of the current proof, but:

1. It’s a key step, involving important (and relatively new) notions, such as that of an “assignment tester” or a “PCP of proximity” [DR04,BGHSV04].

2. It was not covered in detail in Irit’s talk (who, naturally, preferred to show the proof of the new amplification step).

 


 

Title:

Extractors for a Constant Number of Polynomial Min-Entropy Independent Sources

Speaker:

Zeev Dvir

 

A randomness extractor is a function transforming several independent sources of randomness into an almost uniform output. In order for the output of the extractor to be close to uniform the input sources must satisfy some lower bound on their min-entropy. Much effort has been made in order to construct extractors for as few sources as possible with as low min-entropy as possible.

 

Recently, Anup Rao gave a beautiful construction of an extractor for O(1) n-bit sources, each with min-entropy at least nr for any constant 0<r<1 (the number of sources depends on r). Rao's construction is interesting for two reasons. The first is that this is the only construction that works for a constant number of sources with sub-linear entropy and the second is that this construction does not rely on results from additive number theory and uses only "classical" extractor techniques.

 


 

Title:

Estimation of a certain exponential sum arising in complexity theory

Speaker:

Amir Yehudayoff

 

The paper is by J. Bourgain.

Bourgain bounds the correlation between two functions using exponential sums.

Then, using the epsilon-discriminator lemma of Hajnal, Maass, Pudlak, Szegedy and Turan,

he proves an exponential lower bound on the size of a certain depth 3 circuit

(a Maj o Mod_q o And_d circuit; that is, a majority of mod q of ands).

 


 

Title:

Evolving Notions of Security for Quantum Protocols

Speaker:

Adam Smith

 

The talk will discuss some of the difficulties of defining and proving security of cryptographic protocols in a context where the adversary (and occasionally the participants) have quantum computers.

 

First, we'll discuss existing classical security proofs -- not all of them go through when the adversary has a quantum computer. We'll explain the difficulties as well as some of the techniques that have been used to 'patch' these proofs. Second, we'll discuss new, truly 'quantum' protocols and reductions.

 

Familiarity with the basic notation of quantum computing will be helpful but not necessary.

 


 

Title:

Markov-Chain Monte-Carlo (MCMC) Methods in Theoretical

Computer Science: An overview

Speaker:

Elad Verbin (Tel-Aviv University)

 

 

In this lecture I will give an overview presentation of the topic of Markov Chain Monte Carlo (MCMC) Algorithms.

 

The field of MCMC Algorithms is about Markov chains for combinatorial objects and trying to upper bound their so-called mixing time (the time it takes them to "mix" to the uniform distribution, or close to it).

This as opposed to ergodicity theory, developed in the early half of the 20th century, which dealt only with researching the (existence of) a limit distribution, without researching the rate of convergence to it.

 

I will do my best to give enough background and intuition into MCMC so that the listeners will be able to identify "themes" that relate to MCMC in their problems, and to adopt the MCMC "way of thinking". The lecture will try to give intuition without being technical. Only elementary backround in probability theory will be assumed.

 

We will discuss the theory of MCMC, as opposed to practical (and often heuristic) applications of it. The theoretical study of MCMC started in the '80s, and rapid progress has been made since then. Some of the successes of this field are:

(i) an efficient algorithm for approximating (to arbitrary precision) the permanent of a positive matrix;

(ii) An efficient algorithm for approximating the volume of an n-dimensional convex body (represented by a membership oracle) that runs in time polynomial in n;

(iii) An efficient algorithm for approximating the number of proper k-colorings of a graph, for large enough k.

 

Here is an example: Let G be a graph with maximum degree D, let k be an integer such that k>D+1. You wish to uniformly sample a legal coloring of the graph G with k colors (this is called a k-coloring). It is easy to see that there is at least one such legal coloring, and it is easy to find it. (Note: the problem is not NP-Hard because we are coloring not by the minimal number of colors, but by a number of colors that is guaranteed to suffice). But how can we pick a random legal coloring? We take the (arbitrary) coloring that we found, and we make small local random changes to it repeatedly, while making sure it stays a legal coloring. A process of this form is called a Markov Chain (in this case, the chain is defined on the colorings of the graph). With an appropriate Markov Chain, one can prove that if we run for a sufficiently large number of steps then we will get a coloring from a distribution that is (almost) uniform. The problem is how many steps this actually requires: is it polynomial? The answer to this question depends, of course, on the Markov Chain Chosen, that is on the type of "local, random changes".

 


 

Title:

Waters Signing for Fun and Profit

Speaker:

Hovav Shacham

 

 

Moni Naor observed that every identity-based encryption (IBE) scheme gives rise to a signature scheme.  For example, BLS signatures are the signatures obtained from Boneh-Franklin IBE. For BLS signatures and Boneh-Franklin IBE, only a heuristic argument of security is known, in the so-called random oracle model.

 

At Eurocrypt 2005, Brent Waters gave the first practical IBE scheme secure without random oracles. In this talk, we consider the signature scheme obtained from Waters IBE.

 

Waters signatures are provably secure (without random oracles) under the computational Diffie-Hellman assumption.

Furthermore, the scheme yields short signatures -- 320 bits -- and, like BLS, has algebraic structure that is useful for constructing extensions and variants.

 


 

Title:

Cryptography in NC0

Speaker:

Guy Rothblum

 

In a recent work Appelbaum, Ishai, and Kushilevtiz studied the parallel time-complexity of basic cryptographic primitives such as one-way functions (OWFs) and pseudorandom generators (PRGs). They considered the possibility of computing instances of these primitives by NC0 circuits, in which each output bit depends on a constant number of input bits, and showed overwhelming evidence for the possibility of cryptography in NC0.

We will survey their surprising theorem, which states that every one-way function computable in NC1 can be compiled into a one-way function in which each output bit depends on a constant number of input bits.

 


 

Title:

The Influence of Variables on Boolean Functions

Speaker:

Ronen Gradwohl

 

 

In this talk we will consider the following question: given some balanced Boolean function f on n variables, how much influence do the individual variables have on the outcome? In particular, do there exist functions in which the influence of every variable is small?

 

Ben-Or and Linial defined the TRIBES function, in which every variable has influence $\Theta(\log n / n)$, and conjectured that this was the best possible. Kahn, Kalai and Linial proved this conjecture by showing that in every Boolean function there exists a variable with influence $\Omega(\log n / n)$.

 

In the talk we will prove both of these results (although most of the time will be spent on the latter). On the way, we will survey the main tools from Fourier analysis of Boolean functions used in the proof of KKL.

 

Bibliography:

·        Michael Ben-Or and Nati Linial. Collective Coin Flipping. In "Randomness and Computation," 1990.

·        Jeff Kahn, Gil Kalai and Nati Linial. The Influence of Variables on Boolean Functions. In FOCS, 1988.

 


 

Title:

Recent advances in List Decoding

Speaker:

Ariel Gabizon

 

The rate of a code C:Fkà Fn is k/n. Intuitively, this is the amount of information transferred per symbol.

The list decoding problem with (constant) parameter ε>0, is as follows:

Given a received word r we wish to find in poly(n) time, all the codewords that disagree with r in at most a (1-ε)-fraction of the coordinates.

In particular, this implies that there can be at most poly(n) codewords in any hamming sphere of radius (1-ε).

Given this requirement we would like to maximize the rate R of the code. It can be shown that we must have R <= ε.

Non-constructively, we can get rate R= ε–o(1). Guruswami and Sudan gave a list decoding algorithm for Reed-Solomon codes that can work with R=ε2<< ε. For a long time, attempts to go past the O(ε2) 'barrier' were unsuccessful. Recently, breakthrough works of Parvaresh and Vardi, shortly after improved by Guruswami and Rudra achieved R= ε - ε' for any ε'>0.

 

In the talk we will review the standard Berlkamp-Welch Reed Solomon decoding, and the Guruswami-Sudan list-decoding algorithm. Then we will go over the ideas in these two new works.

 


 

Title:

Biased Graph Processes and The Emergence of The Giant Component

Speaker:

Gidi Amir

 

A random graph process, G1(n), is a sequence of graphs on n vertices which begins with the edgeless graph, and where at each step a single edge is added according to a uniform distribution on the missing edges. It is well known that in such a process a giant component (of linear size) typically emerges after (1+o(1))n/2 edges (a phenomenon known as “the double jump”), i.e., at time t=1 when using a timescale of n/2 edges in each step.

We consider a generalization of this process, GK(n), which gives a weight of size 1 to missing edges between pairs of isolated vertices, and a weight of size KÎ[0,¥) otherwise. This corresponds to a case where links are added between n initially isolated settlements, where the probability of a new link in each step is biased according to whether or not its two endpoint settlements are still isolated.

Using methods of Spencer and Wormald, we describe the typical emerging time of a giant component in this process, tc(K), as the singularity point of a solution to a set of differential equations. We proceed to analyze these differential equations and obtain properties of GK, and in particular, we show that tc(K) decreases from 3/2 to 0 as K increases from 0 to ¥, and that tc(K) =  4/Ö(3K)(1 + o(1)).

Joint work with Ori Gurel-Gurevich, Eyal Lubetzky  and Amit Singer.

 


 

Title:

Everyday Inequalities

Speaker:

Ariel Yadin

 

We plan to review and prove very basic inequalities that are frequently used:

Jensen, Markov, Chebyshev, estimates for (1 + x)n, Chernoff, Azuma, Shearer, Hölder & Cauchy-Schwartz, Janson,...

 

A link to Ariel’s inequalities file: http://www.wisdom.weizmann.ac.il/~ayadin/Ineq.pdf.

 


 

Title:

Tutorial on Black-Box Separations

Speaker:

Omer Reingold

 

A practice talk for a tutorial on black-box separations to be given in TCC 06.

 


 

Title:

A doubling dimension threshold O(loglog n) for augmented graphs navigability

Speaker:

Emmanuelle Lebhar (LIP, ENS Lyon [France])

 

In his seminal work, Kleinberg showed how to augment meshes using random edges, so that they become navigable; that is, greedy routing computes paths of poly-logarithmic expected length between any pairs of nodes. This yields the crucial question of determining whether such an augmentation is possible for all graphs.

 

In this paper, we answer negatively to this question by exhibiting a threshold on the doubling dimension, above which an infinite family of Cayley graphs cannot be augmented to become navigable whatever the distribution of random edges is. Precisely, it was known that graphs of doubling dimension at most O(log log n) are navigable. We show that for doubling dimension >> loglog n, an infinite family of Cayley graphs cannot be augmented to become navigable.

 

Our proof relies on a result of independent interest: we show that the analysis of greedy routing worst performances on graphs augmented by arbitrary distributions of edges can be reduced to the analysis assuming a symmetric distribution. Finally, we complete our result by studying the special case of square meshes that we prove to always be augmentable to become navigable.

 

Joint work with Pierre Fraigniaud and Zvi Lotker.

 


 

Title:

How to prove that Minicrypt=Cryptomania (in the future)

Speaker:

Moni Naor and Danny Harnik

 

Arguably the most significant question in the intersection of cryptography and complexity is whether Minicrypt=Cryptomania. One way to phrase the question is whether public-key cryptography is a fluke and can be based only on very special types of functions, or that in principal it is possible to base it on any one-way function.

In this talk we will present two approaches for showing the latter. The first, due to Rudich is based on secret sharing over general access structures. The second, due to Harnik and Naor (and continues Harnik's presentation in the Sunday seminar) is based on compression of NP problems. That is given a long NP problem with a relatively short witness the goal is to come up with another instance having the same solution.

 


 

Title:

Finding small roots of modular polynomials

Speaker:

Hovav Shacham

 

Consider a modulus N of unknown factorization.

How do we go about finding roots of a polynomial f(x) modulo N?  (Or, more generally, modulo some factor of N?)

 

In this talk, we give Coppersmith's method for computing, in polynomial time, roots of the equation

 

         f(x) = 0 mod b    where b divides N, and b > Nb

 

that are small: less than (roughly) Nb^2 / deg(f).

 

We begin by presenting the Lenstra-Lenstra-Lovasz approximation algorithm for the shortest-vector problem on lattices. We then present Coppersmith's method. This method uses the LLL algorithm to obtain from the modular polynomial f(x) a polynomial g(x) that has the same small root x0 over the integers; we also discuss Howgrave-Graham's sufficient condition for a polynomial to have this property.

Finally, we consider some applications of Coppersmith's method to cryptanalyzing certain classes of RSA keys.

 

Our exposition is a simplified version of that given in Alexander May's thesis.

 


 

Title:

Semi-direct product in groups and Zig-zag product in graphs: connections and applications

Speaker:

Zeev Dvir

 

The talk will present a paper by Alon, Lubotzky and Wigderson from 2001 showing a connection between the Zig-Zag product of [RVW] and a well known group theoretic operation called the semi-direct product.

Using this connection and results on the Zig-Zag product, the paper answers an open question in group theory.

 

I will not assume any previous knowledge in group theory.


 

Title:

Provably Secure Steganography with Imperfect Sampling

Speaker:

Mira Meyerovich

 

The goal of steganography is to hide the existence of communication.

The sender must disguise his secret messages as an innocent-looking covertext. For example, the sender can insert a secret letter into a photo and post it on the Internet.  Real world stegosystems are often broken because they make invalid assumptions about the covertext distribution.  All provably secure stegosystems in the literature assume that, at a minimum, the sender is able to sample the covertext distribution. We examine whether it is possible to weaken this assumption. By modeling the covertext distribution as a stateful Markov process, we create a sliding scale between real world and provably secure stegosystems. We also show that insufficient knowledge of past states can have catastrophic results.

 

Joint work with Anna Lysyanskaya


 

Title:

Reducing Randomness for Sampling

Speaker:

Dana Moshkovitz

 

This talk will cover basic topics in derandomization.

We will consider the following sampling problem:

 

Fn is some finite vector space over a field F.

We have access to a membership oracle for some subset AÍFn.

We wish to estimate the size of A via an efficient probabilistic algorithm.

 

The algorithm is given as input an accuracy parameter e and a confidence parameter d,

and outputs an estimate m that approximates |A|/|Fn| well with high probability;

specifically, | m - |A|/|Fn| | £ e with probability at least 1-d.

 

We would like to minimize the amount of randomness the algorithm uses (which is a function of |F|, n, 1/d and 1/e).

 

We will discuss several commonly known sampling techniques, such as: standard independent sampling, pairwise independent sampling and almost pairwise independent sampling (via Cayley expanders). We’ll see the connection between samplers and extractors and analyze the randomness needed for each of the sampling techniques.

We will also apply Fourier analysis to analyze a very simple and relatively randomness-efficient sampler from a recent paper with Ran Raz.

 

As usual, notes for the talk will be available for anyone to read and/or photocopy.

 

References:

·        Survey on samplers: Oded Goldreich, A Sample of Samplers: A Computational Perspective on Sampling, 1997, ECCC TR97-020.

·        Almost k-wise independence: Noga Alon, Oded Goldreich, Johan Håstad and René Peralta, Simple constructions of almost k-wise independent random variables, FOCS’90.

·        Connection to extractors: David Zuckerman, Randomness-Optimal Oblivious Sampling, 1997 (preliminary version in STOC’96).


 

Title:

Proving Lower Bounds for Secret Sharing Schemes Using the Entropy Function

Speaker:

Noam Livne

 

The entropy function, introduced by Shannon in 1948, is commonly used in the definition of cryptographic schemes. But using certain properties of this function and of related functions (such as the conditional entropy and the mutual information) enables us to use it as a tool for proving bounds on the supports of random variables. We demonstrate this by proving the only known lower bound for secret sharing to date.

 

The talk will consist of three parts:

1) basic notions in secret sharing;

2) entropy and related functions and their properties;

3) lower bound for secret sharing.


 

Title:

STOC’06 Report

Speaker:

Dana Moshkovitz

Talks/pictures and chocolates from Ran, straight from Seattle, where the 38th ACM Symposium on Theory of Computing was held on May 21-23, 2006.


 

Title:

On the Efficiency of Local Decoding Procedures for Error-Correcting Codes

Speaker:

Amir Yehudayoff

 

We present the paper On the Efficiency of Local Decoding Procedures for Error-Correcting Codes, by Jonathan Katz and Luca Trevisan.

In this paper Katz and Trevisan define locally decodable codes, and give lower bounds for the size of such codes. Their lower bounds are the best known.

 


 

Title:

Blind Signatures and Unique Zero Knowledge

Speaker:

Hovav Shacham

 

In a blind signature scheme, a user interacts with a signer to obtain a signature on message of his choice, in such a way that the signer does not learn anything about the message he is signing.  (Blind signatures have application, for example, in anonymous e-cash.) 

We describe Fischlin's recent construction of two-move concurrent blind signatures based on general assumptions.  The construction makes use, as a building block, of unique non-interactive zero-knowledge proofs -- i.e., NIZKs in which there is a bijection between witnesses and proofs. We describe Fischlin's construction, based on general assumptions, of a UNIZK for CircuitSAT.


 

Title:

Deterministic extractors for small space sources

Speaker:

Anup Rao

 

 

We give polynomial-time, deterministic randomness extractors for sources generated in small space, where we model space s sources on {0,1}n as sources generated by width 2s branching programs: For every constant d>0, we can extract .99dn bits that are exponentially close to uniform (in variation distance) from space s sources of min-entropy dn, where s=O(n). In addition, assuming an efficient deterministic algorithm for finding large primes, there is a constant h>0 such that for any b>n-h, we can extract m=(d-b)n bits that are exponentially close to uniform from space s sources with min-entropy dn, where s=O(b3n). Previously, nothing was known for d£½, even for space 0.

 

Joint work with Jesse Kamp, Salil Vadhan and David Zuckerman.


 

Title:

Stepanov's elementary method for proving Weil’s bounds

Speaker:

Ariel Gabizon

 

Let F be a field of size q, where q is odd. When choosing a random element t from F\{0} the probability of t being a quadratic residue, i.e., a square of another element in F, is exactly 1/2. Let f be a polynomial of degree d<<q that is not of the form f=c*g(t)2 for any constant c and polynomial g. Suppose we pick a random element t from F, and now apply f on it. What is the probability that f(t) will be a quadratic residue? It turns out that this probability is also very close to 1/2, specifically, ½ ± d/q1/2.

This result has many applications in theoratical computer science. It was originally derived from Andre Weil's important theorem known as the 'Riemann hypothesis for curves over finite fields'. Weil's proof, published in 1948, is considered very difficult and uses deep ideas from algebraic geometry. 20 years later, Sergei Stepanov gave elementary proofs of some of  Weil's most significant results including the one described above. We will show Stepanov's proof of this result.

 


 

Title:

Graph Isomorphism for Planar Graphs

Speaker:

Or Meir

 

The Graph Isomorphism problem is one of today's big question marks for algorithms and complexity.

On the one hand, it has resisted any attempts to solve it for quite a long time, and thus is conjectured not to be in P.

On the other hand, there are theoretical evidence pointing that the problem is not NP-complete, and actually, it is not even known to be P-complete.

In contrast to the lack of knowledge regarding the general problem, polynomial time algorithms were found for many special cases of Graph Isomorphism. In this talk we will present a O(nlogn) algorithm for the special case of planar graphs.


 

Title:

Pseudorandomness -- an overview

Speaker:

Oded Goldreich

 

 

A fresh view at the "question of randomness" was taken in the theory of computing: It has been postulated that a distribution is pseudorandom if it cannot be told apart from the uniform distribution by an efficient procedure.

 

The paradigm, originally associating efficient procedures with polynomial-time algorithms, has been applied also with respect to a variety of limited classes of such distinguishing procedures.

Starting with the general paradigm, I will present the case of general-purpose pseudorandom generators (running in polynomial-time and withstanding any polynomial-time distinguisher), as well as derandomization-aimed generators (used towards the derandomization of complexity classes such as BPP), generators withstanding space-bounded distinguishers, and some special-purpose generators.

 

The talk will have a significant intersection with my lectures on the subject during this year's complexity class.

But, the style and details will be different.


 

Title:

Truthful Randomized Mechanisms for Combinatorial Auctions

Speaker:

Michael Schapira

 

The field of Algorithmic Mechanism Design attempts to design efficient mechanisms for decentralized computerized settings. The combinatorial auctions problem has gained the status of the paradigmatic problem of this field. We design two computationally-efficient incentive-compatible mechanisms for combinatorial auctions with general bidder preferences. Both mechanisms are randomized, and are incentive-compatible in the universal sense. This is in contrast to recent previous work that only addresses the weaker notion of incentive compatibility in expectation. The first mechanism obtains an O(Öm)-approximation of the optimal social welfare for arbitrary bidder valuations -- this is the best approximation possible in polynomial time.  The second one obtains an O(log2 m)-approximation for a subclass of bidder valuations that includes all submodular bidders. This improves over the best previously obtained incentive-compatible mechanism for this class which only provides a O(Öm)-approximation.

Based on a recent joint work with Shahar Dobzinski and Noam Nisan.