Streaming algorithms in adversarial settings have attracted considerable attention recently. We show that, in the white-box adversarial streaming model [Miklós Ajtai et al., 2022], the fundamental problem of estimating the F_p moment to within any constant factor requires Ω(n) memory. In this model, the internal state of the (randomized) streaming algorithm is visible to an adversary, who can exploit this information when constructing subsequent stream updates. As a corollary, we also obtain a white-box lower bound for the well-studied problem of estimating the maximum matching size in graphs. [Miklós Ajtai et al., 2022] proved that two-party white-box communication protocols can be derandomized. This allows them to prove deterministic communication lower bounds and automatically derive white-box (communication and streaming) lower bounds. However, such two-party lower bounds can only rule out approximation of the F_p moment within a specific constant factor. Ruling out approximation within any constant factor typically requires proving a lower bound for a multi-party communication problem. We show that white-box communication protocols involving any number of parties can be derandomized, provided they compute a total function. However, this derandomization fails entirely when extended to partial functions and, consequently, to approximation problems. We are therefore compelled to prove our moment estimation lower bound for the white-box model directly. Our proof introduces a novel hybrid technique that, instead of taking hybrids over input distributions, constructs hybrids over white-box adversaries.

Sampling a random walk is a fundamental primitive in many graph applications. In the streaming model, it is known that sampling an L-step random walk on an n-vertex directed graph requires Ω(n L) space, implying that no sublinear-space streaming algorithm exists for general graphs. We show that sublinear algorithms are possible for the case of dense graphs, where every vertex has out-degree at least Ω(n). In particular, we give a one-pass turnstile streaming algorithm that uses only 𝒪̃(L) memory for such graphs. More broadly, for graphs with minimum out-degree at least d, our streaming algorithm samples a random walk using 𝒪̃(n/d ⋅ L) memory. We show that our algorithm is optimal in a strong "beyond worst-case" sense. To formalize this, we introduce the notion of universal optimality for graph streaming algorithms. Informally, a streaming algorithm is universally optimal if it performs (almost) as well as possible on every graph, assuming a worst-case choice of the streaming order. This notion of universal optimality is a key conceptual contribution of our work.

The space complexity of deterministic streaming algorithms for approximating the length of the longest increasing subsequence (LIS) in a string of length n has been known to be Θ̃(√n) for almost two decades. In contrast, the space complexity of this problem for randomized streaming algorithms remains one of the few longstanding open problems in one-pass streaming. In fact, no better than Ω(log n) lower bounds are known, and the best upper bounds are no better than their deterministic counterparts. In this paper, we push the limits of our understanding of the streaming space complexity of the approximate LIS length problem by studying it in the white-box adversarial streaming model. This model is an intermediate model between deterministic and randomized streaming algorithms that has recently attracted attention. In the white-box model, the streaming algorithm can draw fresh randomness when processing each incoming element, but an adversary generating the stream observes all previously used randomness and adaptively chooses the subsequent elements of the stream. We prove a tight (up to logarithmic factors) Ω(√n) space lower bound for any white-box streaming algorithm that approximates the length of the LIS of a stream of length n to within a factor better than 1.1. Thus, for this problem, white-box algorithms offer no improvement over deterministic ones.

Can the n-party broadcast channel, where any symbol sent by one party is received by all, be made resilient to noise with low overhead? Namely, is it possible to construct interactive error-correcting codes that convert any protocol designed for the noiseless broadcast channel into one that works over the noisy broadcast channel and is not much longer than the original protocol?[12, STOC 2018] showed that such interactive codes with constant multiplicative overhead are possible under the assumption that the noiseless protocol being simulated is non-adaptive, meaning that it is restricted to have a pre-determined order of turns. Their noise resilient simulating protocols, however, require adaptivity, where each party can decide whether or not to broadcast given all the information available to them, including their input and received transcript. The question of whether such a simulation is possible for general, potentially adaptive, noiseless protocols was left open.We resolve this question negatively, proving that any interactive code that converts adaptive noiseless broadcast protocols into adaptive broadcast protocols resilient to stochastic errors must incur a multiplicative overhead of Ω(log n/ log log n), which is nearly tight.

Abstract. We consider the problem of finding a maximal independent set (MIS) in the shared blackboard communication model with vertex-partitioned inputs. There are [Formula: see text] players corresponding to vertices of an undirected graph, and each player sees the edges incident on its vertex; this way, each edge is known by both its endpoints and is thus shared by two players. The players communicate in simultaneous rounds by posting their messages on a shared blackboard visible to all players, with the goal of computing an MIS of the graph. While the MIS problem is well studied in other distributed models and while shared blackboard is, perhaps, the simplest broadcast model, lower bounds for our problem were only known against one-round protocols. We present a lower bound on the round-communication tradeoff for computing an MIS in this model. Specifically, we show that, when [Formula: see text] rounds of interaction are allowed, at least one player needs to communicate [Formula: see text] bits. In particular, with logarithmic bandwidth, finding an MIS requires [Formula: see text] rounds. This lower bound can be compared with the algorithm of Ghaffari et al. [ Proceedings of the 2018 ACM Symposium on Principles of Distributed Computing, 2018, pp. 129–138] that solves the MIS in [Formula: see text] rounds but with a logarithmic bandwidth for an average player. Additionally, our lower bound further extends to the closely related problem of maximal bipartite matching. The presence of edge-sharing gives the algorithms in our model a surprising power, and numerous algorithmic results exploiting this power are known. For a similar reason, proving lower bounds in this model is much more challenging because this sharing in the players’ inputs prohibits the use of standard number-in-hand communication complexity arguments. Thus, to prove our results, we devise a new round elimination framework, which we call partial-input embedding, that may also be useful in future work for proving round-sensitive lower bounds in the presence of shared inputs. Finally, we discuss several implications of our results to multiround (adaptive) distributed sketching algorithms, broadcast congested clique, and the welfare maximization problem in two-sided matching markets.

A group of $n$ users want to run a distributed protocol $π$ over a network where communication occurs via private point-to-point channels. Unfortunately, an adversary, who knows $π$, is able to maliciously flip bits on the channels. Can we efficiently simulate $π$ in the presence of such an adversary? We show that this is possible, even when $L$, the number of bits sent in $π$, and $T$, the number of bits flipped by the adversary are not known in advance. In particular, we show how to create a robust version of $π$ that 1) fails with probability at most $δ$, for any $δ>0$; and 2) sends $\tilde{O}(L + T)$ bits, where the $\tilde{O}$ notation hides a $\log (nL/ δ)$ term multiplying $L$. Additionally, we show how to improve this result when the average message size $α$ is not constant. In particular, we give an algorithm that sends $O( L (1 + (1/α) \log (n L/δ) + T)$ bits. This algorithm is adaptive in that it does not require a priori knowledge of $α$. We note that if $α$ is $Ω\left( \log (n L/δ) \right)$, then this improved algorithm sends only $O(L+T)$ bits, and is therefore within a constant factor of optimal.

The seminal work of Ahn, Guha, and McGregor in 2012 introduced the graph sketching technique and used it to present the first streaming algorithms for various graph problems over dynamic streams with both insertions and deletions of edges. This includes algorithms for cut sparsification, spanners, matchings, and minimum spanning trees (MSTs). These results have since been improved or generalized in various directions, leading to a vastly rich host of efficient algorithms for processing dynamic graph streams.

We consider the Max-Cut problem, asking how much space is needed by a streaming algorithm in order to estimate the value of the maximum cut in a graph. This problem has been extensively studied over the last decade, and we now have a near-optimal lower bound for one-pass streaming algorithms, showing that they require linear space to guarantee a better-than-2 approximation [50, 52]. This result relies on a lower bound for the cycle-finding problem, showing that it is hard for a one-pass streaming algorithm to find a cycle in a union of matchings. The end-goal of our research is to prove a similar lower bound for multi-pass streaming algorithms that guarantee a better-than-2 approximation for Max-Cut, a highly challenging open problem. In this paper, we take a significant step in this direction, showing that even o(log n)-pass streaming algorithms need nΩ(1) space to solve the cycle-finding problem. Our proof is quite involved, dividing the cycles in the graph into “short” and “long” cycles, and using tailor-made lower bound techniques to handle each case.

In this work, we design an interactive coding scheme that converts any two party interactive protocol Π into another interactive protocol Π', such that even if errors are introduced during the execution of Π', the parties are able to determine what the outcome of running Π would be in an error-free setting. Importantly, our scheme preserves the space complexity of the protocol, in addition to the communication and computational complexities. Specifically, if the protocol Π has communication complexity T, computational complexity t, and space complexity s, the resulting protocol Π' is resilient to a constant ε > 0 fraction of adversarial errors, and has communication complexity approaching T as ε approaches 0, computational complexity poly(t), and space complexity

Kol and Raz [STOC 2013] showed how to simulate any alternating two-party communication protocol designed to work over the noiseless channel, by a protocol that works over a stochastic channel that corrupts each sent symbol with probability ‍є>0 independently, with only a 1+O(√(є)) blowup to the communication. In particular, this implies that the maximum rate of such interactive codes approaches 1 as є goes to ‍0, as is also the case for the maximum rate of classical error correcting codes. Over the past decade, followup works have strengthened and generalized this result to other noisy channels, stressing on how fast the rate approaches 1 as є goes to 0, but retaining the assumption that the noiseless protocol is alternating.

We consider the problem of finding a maximal independent set (MIS) in the shared blackboard communication model with vertex-partitioned inputs. There are n players corresponding to vertices of an undirected graph, and each player sees the edges incident on its vertex – this way, each edge is known by both its endpoints and is thus shared by two players. The players communicate in simultaneous rounds by posting their messages on a shared blackboard visible to all players, with the goal of computing an MIS of the graph. While the MIS problem is well studied in other distributed models, and while shared blackboard is, perhaps, the simplest broadcast model, lower bounds for our problem were only known against one-round protocols. We present a lower bound on the round-communication tradeoff for computing an MIS in this model. Specifically, we show that when r rounds of interaction are allowed, at least one player needs to communicate $\Omega(n^{1/20^{r+1}})$ bits. In particular, with logarithmic bandwidth, finding an MIS requires $\Omega(\log\log n)$ rounds. This lower bound can be compared with the algorithm of Ghaffari, Gouleakis, Konrad, Mitrović, and Rubinfeld [PODC 2018] that solves MIS in $O(\log\log n)$ rounds but with a logarithmic bandwidth for an average player. Additionally, our lower bound further extends to the closely related problem of maximal bipartite matching. The presence of edge-sharing gives the algorithms in our model a surprising power and numerous algorithmic results exploiting this power are known. For a similar reason, proving lower bounds in this model is much more challenging, as this sharing in the players’ inputs prohibits the use of standard number-in-hand communication complexity arguments. Thus, to prove our results, we devise a new round elimination framework, which we call partial-input embedding, that may also be useful in future work for proving round-sensitive lower bounds in the presence of shared inputs. Finally, we discuss several implications of our results to multi-round (adaptive) distributed sketching algorithms, broadcast congested clique, and to the welfare maximization problem in two-sided matching markets.

In the reliable transmission problem, a sender, Alice, wishes to transmit a bit-string x to a remote receiver, Bob, over a binary channel with adversarial noise. The solution to this problem is to encode x using an error correcting code. As it is long known that the distance of binary codes is at most 1/2, reliable transmission is possible only if the channel corrupts (flips) at most a 1/4-fraction of the communicated bits.We revisit the reliable transmission problem in the two-way setting, where both Alice and Bob can send bits to each other. Our main result is the construction of two-way error correcting codes that are resilient to a constant fraction of corruptions strictly larger than 1/4. Moreover, our code has constant rate and requires Bob to only send one short message. We mention that our result resolves an open problem by Haeupler, Kamath, and Velingker [APPROX-RANDOM, 2015] and by Gupta, Kalai, and Zhang [STOC, 2022].Curiously, our new two-way code requires a fresh perspective on classical error correcting codes: While classical codes have only one distance guarantee for all pairs of codewords (i.e., the minimum distance), we construct codes where the distance between a pair of codewords depends on the “compatibility” of the messages they encode. We also prove that such codes are necessary for our result.

Zero knowledge proofs are one of the most influential concepts in theoretical computer science. In the seminal definition due to Goldwasser, Micali and Rackoff dating back to the 1980s, a computationally-bounded verifier interacts with a powerful but untrusted prover, with the goal of becoming convinced that the input is in some language. In addition to the usual requirements of completeness and soundness, in a zero knowledge proof, we protect the prover's knowledge: assuming the prover is honest, anything that the verifier can deduce after interacting with the prover, it could have deduced by itself. Zero knowledge proofs have found many applications within theoretical computer science and beyond, e.g., in cryptography, client-cloud computing, blockchains and cryptocurrencies, electronic voting and auctions, and in the financial industry. We define and study the notion of distributed zero knowledge proofs, reconciling the computational notion of zero-knowledge with the communication-based paradigm of distributed graph algorithms. In our setting, a network of verifiers interacts with an untrusted prover to decide some distributed language. As is usually the case in distributed graph algorithms, we assume that the verifiers have local views of the network and each only knows its neighbors. The prover, on the other hand, is assumed to know the entire network graph, as well as any input that the verifier may possess. As in the computational centralized setting, the protocol we design should protect this knowledge. In particular, due to the dual role of the underlying graph in distributed graph algorithms, serving as both the communication topology and the input to the problem, our protocol must protect the graph itself. We construct communication-efficient distributed zero knowledge proofs for two central problems: the 3-coloring problem, one of the poster children of computational zero-knowledge, and for the spanning-tree verification problem, a fundamental building block for designing graph algorithms. We also give a general scheme for converting proof labeling-schemes to distributed zero-knowledge protocols with related parameters. Our protocols combine ideas from computational complexity, distributed computing, and cryptography.

Given a Boolean circuit C, we wish to convert it to a circuit C′ that computes the same function as C even if some of its gates suffer from adversarial short circuit errors, i.e., their output is replaced by the value of one of their inputs. Can we design such a resilient circuit C′ whose size is roughly comparable to that of C? Prior work gave a positive answer for the special case where C is a formula.

Distributed learning protocols are designed to train on distributed data without gathering it all on a single centralized machine, thus contributing to the efficiency of the system and enhancing its privacy. We study a central problem in distributed learning, called {\it distributed learning of halfspaces}: let $U \subseteq \mathbb{R}^d$ be a known domain of size $n$ and let $h:\mathbb{R}^d\to \mathbb{R}$ be an unknown target affine function.\footnote{In practice, the domain $U$ is defined implicitly by the representation of $d$-dimensional vectors which is used in the protocol.} A set of examples $\{(u,b)\}$ is distributed between several parties, where~$u \in U$ is a point and $b = \mathsf{sign}(h(u)) \in \{\pm 1\}$ is its label. The parties goal is to agree on a classifier~$f: U\to\{\pm 1\}$ such that~$f(u)=b$ for every input example~$(u,b)$. We design a protocol for the distributed halfspace learning problem in the two-party setting, communicating only $\tilde O(d\log n)$ bits. To this end, we introduce a new tool called halfspace containers, that is closely related to bracketing numbers in statistics and to hyperplane cuttings in discrete geometry, and allows for a compressed approximate representation of every halfspace. We complement our upper bound result by an almost matching $\tilde \Omega(d\log n)$ lower bound on the communication complexity of any such protocol Since the distributed halfspace learning problem is closely related to the convex set disjointness problem in communication complexity and the problem of distributed linear programming in distributed optimization, we also derive upper and lower bounds of $\tilde O(d^2\log n)$ and~$\tilde{\Omega}(d\log n)$ on the communication complexity of both of these basic problems.

Hypothesis Selection is a fundamental distribution learning problem where given a comparator-class$\mathcal{Q}=\{q_{1}, \ldots, q_{n}\}$of distributions, and a sampling access to an unknown target distribution$p$, the goal is to output a distribution$q$such that$\mathsf{TV}(p, q)$is close to opt, where$\mathsf{opt}=\min\nolimits_{i}\{\mathsf{TV}(p, q_{i})\}$and TV (.,.) denotes the total-variation distance. Despite the fact that this problem has been studied since the 19th century, its complexity in terms of basic resources, such as number of samples and approximation guarantees, remains unsettled (this is discussed, e.g., in the charming book by Devroye and Lugosi '00). This is in stark contrast with other (younger) learning settings, such as PAC learning, for which these complexities are well understood. We derive an optimal 2-approximation learning strategy for the Hypothesis Selection problem, outputting$q$such that,$\mathsf{TV}(p, q)\leq 2\cdot\text{opt}+\varepsilon$, with a (nearly) optimal sample complexity of$\tilde{O}(\log n/\varepsilon^{2})$. This is the first algorithm that simultaneously achieves the best approximation factor and sample complexity: previously, Bousquet, Kane, and Moran (COLT ‘19) gave a learner achieving the optimal 2-approximation, but with an exponentially worse sample complexity of$\tilde{O}(\sqrt{n}/\varepsilon^{2.5})$, and Yatracos (Annals of Statistics '85) gave a learner with optimal sample complexity of$O(\log n/\varepsilon^{2})$but with a sub-optimal approximation factor of 3. We mention that many works in the Density Estimation (a.k.a., Distribution Learning) literature use Hypothesis Selection as a black box subroutine. Our result therefore implies an improvement on the approximation factors obtained by these works, while keeping their sample complexity intact. For example, our result improves the approximation factor of the algorithm of Ashtiani, Ben-David, Harvey, Liaw, and Mehrabian (JACM '20) for agnostic learning of mixtures of gaussians from 9 to 6, while maintaining its nearly-tight sample complexity.

Let II be a protocol over the n-party broadcast channel, where in each round, a pre-specified party broadcasts a symbol to all other parties. We wish to design a scheme that takes such a protocol II as input and outputs a noise resilient protocol II’ that simulates II over the noisy broadcast channel, where each received symbol is flipped with a fixed constant probability, independently. What is the minimum overhead in the number of rounds that is incurred by any such simulation scheme? A classical result by Gallager from the 80's shows that non-interactive T-round protocols, where the bit communicated in every round is independent of the communication history, can be converted to noise resilient ones with only an$\mathrm{O}(\log\log T$) multiplicative overhead in the number of rounds. Can the same be proved for any protocol? Or, are there protocols whose simulation requires an$\Omega(\log T)$overhead (which always suffices)? We answer both the above questions in the negative: We give a simulation scheme with an$\tilde{O}(\sqrt{\log T})$overhead for every protocol and channel alphabet. We also prove an (almost) matching lower bound of$\Omega(\sqrt{\log T})$on the overhead required to simulate the pointer chasing protocol with T = n and polynomial alphabet.

For a directed graph G with n vertices and a start vertex u_start, we wish to (approximately) sample an L-step random walk over G starting from u_start with minimum space using an algorithm that only makes few passes over the edges of the graph. This problem found many applications, for instance, in approximating the PageRank of a webpage. If only a single pass is allowed, the space complexity of this problem was shown to be Θ̃(n ⋅ L). Prior to our work, a better space complexity was only known with Õ(√L) passes. We essentially settle the space complexity of this random walk simulation problem for two-pass streaming algorithms, showing that it is Θ̃(n ⋅ √L), by giving almost matching upper and lower bounds. Our lower bound argument extends to every constant number of passes p, and shows that any p-pass algorithm for this problem uses Ω̃(n ⋅ L^{1/p}) space. In addition, we show a similar Θ̃(n ⋅ √L) bound on the space complexity of any algorithm (with any number of passes) for the related problem of sampling an L-step random walk from every vertex in the graph.

We study the n-party noisy broadcast channel with a constant fraction of malicious parties. Specifically, we assume that each non-malicious party holds an input bit, and communicates with the others in order to learn the input bits of all non-malicious parties. In each communication round, one of the parties broadcasts a bit to all other parties, and the bit received by each party is flipped with a fixed constant probability (independently for each recipient). How many rounds are needed? Assuming there are no malicious parties, Gallager gave an 𝒪(n log log n)-round protocol for the above problem, which was later shown to be optimal. This protocol, however, inherently breaks down in the presence of malicious parties. We present a novel n ⋅ 𝒪̃(√{log n})-round protocol, that solves this problem even when almost half of the parties are malicious. Our protocol uses a new type of error correcting code, which we call a locality sensitive code and which may be of independent interest. Roughly speaking, these codes map "close" messages to "close" codewords, while messages that are not close are mapped to codewords that are very far apart. We view our result as a first step towards a theory of property preserving interactive coding, i.e., interactive codes that preserve useful properties of the protocol being encoded. In our case, the naive protocol over the noiseless broadcast channel, where all the parties broadcast their input bit and output all the bits received, works even in the presence of malicious parties. Our simulation of this protocol, unlike Gallager’s, preserves this property of the original protocol.

We give an almost quadratic n2−o(1) lower bound on the space consumption of any o(√logn)-pass streaming algorithm solving the (directed) s-t reachability problem. This means that any such algorithm must essentially store the entire graph. As corollaries, we obtain almost quadratic space lower bounds for additional fundamental problems, including maximum matching, shortest path, matrix rank, and linear programming.

We study the error resilience of the message exchange task: Two parties, each holding a private input, want to exchange their inputs. However, the channel connecting them is governed by an adversary that may corrupt a constant fraction of the transmissions. What is the maximum fraction of corruptions that still allows the parties to exchange their inputs?

We show an exponential gap between communication complexity and external information complexity by analyzing a communication task suggested as a candidate by Braverman [ A Hard-to-Compress Interactive Task?, in Proceedings of the 51th Annual Allerton Conference on Communication, Control, and Computing, IEEE, 2013]. Previously, only a separation of communication complexity and internal information complexity was known. More precisely, we obtain an explicit example of a search problem with external information complexity at most $O(k)$, with respect to any input distribution, and distributional communication complexity at least $2^k$, with respect to some input distribution. In particular, this shows that a communication protocol cannot always be compressed to its external information. By a result of Braverman [ SIAM J. Comput., 44 (2015), pp. 1698--1739], our gap is the largest possible. Moreover, since the upper bound of $O(k)$ on the external information complexity of the problem is obtained with respect to any input distribution, our result implies an exponential gap between communication complexity and information complexity (both internal and external) in the nondistributional setting of Braverman [ SIAM J. Comput., 44 (2015), pp. 1698--1739]. In this setting, no gap was previously known, even for internal information complexity.

Consider the following gap cycle counting problem in the streaming model: The edges of a 2-regular n-vertex graph G are arriving one-by-one in a stream and we are promised that G is a disjoint union of either k-cycles or 2k-cycles for some small k; the goal is to distinguish between these two cases using a limited memory. Verbin and Yu [SODA 2011] introduced this problem and showed that any single-pass streaming algorithm solving it requires n1-Ω(1/k)space. This result and the proof technique behind it-the Boolean Hidden Hypermatching communication problem-has since been used extensively for proving streaming lower bounds for various problems, including approximating MAX-CUT, matching size, property testing, matrix rank and Schatten norms, streaming unique games and CSPs, and many others. Despite its significance and broad range of applications, the lower bound technique of Verbin and Yu comes with a key weakness that is also inherited by all subsequent results: the Boolean Hidden Hypermatching problem is hard only if there is exactly one round of communication and, in fact, can be solved with logarithmic communication in two rounds. Therefore, all streaming lower bounds derived from this problem only hold for single-pass algorithms. Our goal in this paper is to remedy this state-of-affairs. We prove the first multi-pass lower bound for the gap cycle counting problem: Any p-pass streaming algorithm that can distinguish between disjoint union of k-cycles vs 2k-cycles-or even k-cycles vs one Hamiltonian cycle-requires n1-1/kΩ(1/p)space. This makes progress on multiple open questions in this line of research dating back to the work of Verbin and Yu. As a corollary of this result and by simple (or even no) modification of prior reductions, we can extend many of previous lower bounds to multi-pass algorithms. For instance, we can now prove that any streaming algorithm that ( 1+ε) -approximates the value of MAX-CUT, maximum matching size, or rank of an n-by- n matrix, requires either nΩ(1)space or Ω(log(1/ε)) passes. For all these problems, prior work left open the possibility of even an O(logn) space algorithm in only two passes.

Interactive error correcting codesInteractive error correcting codes are codes that encode a two party communication protocol to an error-resilient protocol that succeeds even if a constant fraction of the communicated symbols are adversarially corrupted, at the cost of increasing the communication by a constant factor. What is the largest fraction of corruptions that such codes can protect against? If the error-resilient protocol is allowed to communicate large (constant sized) symbols, Braverman and Rao (STOC, 2011) show that the maximum rate of corruptions that can be tolerated is1/4. They also give a binary interactive error correcting protocol that only communicates bits and is resilient to1/2 fraction of errors, but leave the optimality of this scheme as an open problem. We answer this question in the negative, breaking the1/8 barrier. Specifically, we give a binary interactive error correcting scheme that is resilient to5/39 >1/8 fraction of adversarial errors. Our scheme builds upon a novel construction of binary list-decodable interactive codes with small list size.

Consider the following distributed graph sketching model: There is a referee and n vertices in an undirected graph G sharing public randomness. Each vertex v only knows its neighborhood in G and the referee receives no input initially. The vertices simultaneously each sends a message, called a sketch, to the referee who then based on the received sketches outputs a solution to some combinatorial problem on G, say, the minimum spanning tree problem.

We study the effect of noise on the n-party beeping model. In this model, in every round, each party may decide to either 'beep' or not. All parties hear a beep if and only if at least one party beeps. The beeping model is becoming increasingly popular, as it offers a very simple abstraction of wireless networks and is very well suited for studying biological phenomena. Still, the noise resilience of the beeping model is yet to be understood.

Interactive error correcting codes can protect interactive communication protocols against a constant fraction of adversarial errors, while incurring only a constant multiplicative overhead in the total communication. What is the maximum fraction of errors that such codes can protect against?

We consider the celebrated radio network model for abstracting communication in wireless networks. In this model, in any round, each node in the network may broadcast a message to all its neighbors. However, a node is able to hear a message broadcast by a neighbor only if no collision occurred, meaning that it was the only neighbor broadcasting. While the (noiseless) radio network model received a lot of attention over the last few decades, the effect of noise on radio networks is still not well understood. In this paper, we take a step forward and show that making radio network protocols resilient to noise may require a substantial performance overhead. Specifically, we construct a multi-hop network and a communication protocol over this network that works in T rounds when there is no noise. We prove that any scheme that simulates our protocol and is resilient to stochastic noise, requires at least cT log(n) rounds, for some constant c. This stands in contrast to our previous result (STOC, 2018), showing that protocols over the single-hop (clique) network can be made noise resilient with only a constant overhead. Our result also settles a recent conjecture by Censor-Hillel, Haeupler, Hershkowitz, Zuzic (2018). We complement the above result by giving a scheme to simulate any protocol with a fixed order of transmissions with only an O(log (n)) overhead.

Radio networks can be a challenging platform for which to develop distributed algorithms, because the network nodes must contend for a shared channel. In some cases, though, the shared medium is an advantage rather than a disadvantage: for example, many radio network algorithms cleverly use the shared channel to approximate the degree of a node, or estimate the contention. In this paper we ask how far the inherent power of a shared radio channel goes, and whether it can efficiently compute "classicaly hard" functions such as Majority, Approximate Sum, and Parity. Using techniques from circuit complexity, we show that in many cases, the answer is "no". We show that simple radio channels, such as the beeping model or the channel with collision-detection, can be approximated by a low-degree polynomial, which makes them subject to known lower bounds on functions such as Parity and Majority; we obtain round lower bounds of the form Omega(n^{delta}) on these functions, for delta in (0,1). Next, we use the technique of random restrictions, used to prove AC^0 lower bounds, to prove a tight lower bound of Omega(1/epsilon^2) on computing a (1 +/- epsilon)-approximation to the sum of the nodes' inputs. Our techniques are general, and apply to many types of radio channels studied in the literature.

The weak interactive compression conjecture asserts that any two-party communication protocol with communication complexity C and information complexity I can be compressed to a protocol with communication complexity poly(I)polylog(C). We describe a communication problem that is a candidate for refuting that conjecture. Specifically, while we show that the problem can be solved by a protocol with communication complexity C and information complexity I=polylog(C), the problem seems to be hard for protocols with communication complexity poly(I)polylog(C)=polylog(C).

We describe a new way of compressing two-party communication protocols to get protocols with potentially smaller communication. We show that every communication protocol that communicates C bits and reveals I bits of information about the participants’ private inputs to an observer that watches the communication, can be simulated by a new protocol that communicates at most poly(I) · loglog(C) bits. Our result is tight up to polynomial factors, as it matches the recent work separating communication complexity from external information cost.

A set of n players, each holding a private input bit, communicate over a noisy broadcast channel. Their mutual goal is for all players to learn all inputs. At each round one of the players broadcasts a bit to all the other players, and the bit received by each player is flipped with a fixed constant probability (independently for each recipient). How many rounds are needed?

We show an explicit (that is, efficient and deterministic) capacity approaching interactive coding scheme that simulates any interactive protocol under random errors with nearly optimal communication rate. Specifically, over the binary symmetric channel with crossover probability ϵ, our coding scheme achieves a communication rate of 1- O(√/H(ϵ)), together with negligible exp(-Ω(ϵ4n/logn)) failure probability (over the randomness of the channel). A rate of 1 - Θ(√/H(ϵ)) is likely asymptotically optimal as a result of Kol and Raz (2013) suggests. Prior to this paper, such a communication rate was achievable only using randomized coding schemes [Kol and Raz (2013); Hauepler (2014)].

We define a concept class ℱ to be time-space hard (or memory-samples hard) if any learning algorithm for ℱ requires either a memory of size super-linear in n or a number of samples super-polynomial in n, where n is the length of one sample.

The field of compression studies the question of how many bits of communication are necessary to convey a given piece of data. For one-way communication between a sender and a receiver, the seminal work of Shannon and Huffman showed that the communication required is characterized by the entropy of the data; in recent years, there has been a great amount of interest in extending this line of research to interactive communication, where instead of a sender and a receiver we have two parties communication back-and-forth. In this paper we initiate the study of interactive compression for distributed multi-player protocols. We consider the classical shared blackboard model, where players take turns speaking, and each player's message is immediately seen by all the other players. We show that in the shared blackboard model with k players, one can compress protocols down to ~O(Ik), where I is the information content of the protocol and k is the number of players. We complement this result with an almost matching lower bound of ~Omega(Ik), which shows that a nearly-linear dependence on the number of players cannot be avoided.

We study efficient, deterministic interactive coding schemes that simulate any interactive protocol both under random and adversarial errors, and can achieve a constant communication rate independent of the protocol length. For channels that flip bits independently with probability ∊ < 1/2, our coding scheme achieves a communication rate of and a failure probability of exp(−n/log n) in length n protocols. Prior to our work, all nontrivial deterministic schemes (either efficient or not) had a rate bounded away from 1. Furthermore, the best failure probability achievable by an efficient deterministic coding scheme with constant rate was only quasi-polynomial, i.e., of the form exp(− logO(1) n) (Braverman, ITCS 2012). For channels in which an adversary controls the noise pattern our coding scheme can tolerate Ω(1/log n) fraction of errors with rate approaching 1. Once more, all previously known nontrivial deterministic schemes (either efficient or not) in the adversarial setting had a rate bounded away from 1, and no nontrivial efficient deterministic coding schemes were known with any constant rate. Essential to both results is an explicit, efficiently encodable and decodable systematic tree code of length n that has relative distance Ω(1/log n) and rate approaching 1, defined over an O(log n)-bit alphabet. No nontrivial tree code (either efficient or not) was known to approach rate 1, and no nontrivial distance bound was known for any efficient constant rate tree code. The fact that our tree code is systematic, turns out to play an important role in obtaining rate in the random error model, and approaching rate 1 in the adversarial error model.

We study the interactive compression problem: Given a two-party communication protocol with small information cost, can it be compressed so that the total number of bits communicated is also small? We consider the case where the parties have inputs that are independent of each other, and give a simulation protocol that communicates I^2 * polylog(I) bits, where I is the information cost of the original protocol. Our protocol is the first simulation protocol whose communication complexity is bounded by a polynomial in the information cost of the original protocol.

We show an exponential gap between communication complexity and information complexity by giving an explicit example of a partial boolean function with information complexity ≤ O ( k ), and distributional communication complexity ≥ 2 k . This shows that a communication protocol cannot always be compressed to its internal information. By a result of Braverman [2015], our gap is the largest possible. By a result of Braverman and Rao [2014], our example shows a gap between communication complexity and amortized communication complexity, implying that a tight direct sum result for distributional communication complexity cannot hold, answering a long-standing open problem. Another (conceptual) contribution of our work is the relative discrepancy method, a new rectangle-based method for proving communication complexity lower bounds for boolean functions, powerful enough to separate information complexity and communication complexity.

We show an exponential gap between communication complexity and information complexity for boolean functions, by giving an explicit example of a partial function with information complexity ≤ O(k), and distributional communication complexity ≥ 2k. This shows that a communication protocol for a partial boolean function cannot always be compressed to its internal information. By a result of Braverman [Bra12], our gap is the largest possible. By a result of Braverman and Rao [BR11], our example shows a gap between communication complexity and amortized communication complexity, implying that a tight direct sum result for distributional communication complexity of boolean functions cannot hold, answering a long standing open problem. Our techniques build on [GKR14], that proved a similar result for relations with very long outputs (double exponentially long in k). In addition to the stronger result, the current work gives a simpler proof, benefiting from the short output length of boolean functions.

We show an exponential gap between communication complexity and information complexity, by giving an explicit example for a communication task (relation), with information complexity ≤ O(k), and distributional communication complexity ≥2k. This shows that a communication protocol cannot always be compressed to its internal information. By a result of Braverman [1], our gap is the largest possible. By a result of Braverman and Rao [2], our example shows a gap between communication complexity and amortized communication complexity, implying that a tight direct sum result for distributional communication complexity cannot hold.

We show that in the model of zero error communication complexity, direct sum fails for average communication complexity as well as for external information cost. Our example also refutes a version of a conjecture by Braverman et al. that in the zero error case amortized communication complexity equals external information cost.

We study the covering complexity of constraint satisfaction problems (CSPs). The covering number of a CSP instance C, denoted v(C), is the smallest number of assignments to the variables, such that each constraint is satisfied by at least one of the assignments. This covering notion describes situations in which we must satisfy all the constraints, and are willing to use more than one assignment to do so. At the same time, we want to minimize the number of assignments. We study the covering problem for different constraint predicates. We first observe that if the predicate contains an odd predicate, then it is covered by any assignment and its negation. In particular, 3CNF and 3LIN, that are hard in the max-CSP sense, are easy to cover. However, the covering problem is hard for predicates that do not contain an odd predicate: 1. For the 4LIN predicate, it is NP-hard to decide if a given instance C has v(C) at most 2, or v(C) is super-constant. 2. (a) We propose a framework of covering dictatorship tests. We design and analyze such a dictatorship test for every predicate that supports a pair wise independent distribution. (b) We introduce a covering unique games conjecture, and use it to convert the covering dictatorship tests into conditional hardness results. 3. Finally, we study a hypothesis about the hardness of covering random instances that is similar to Feige's R3SAT hypothesis. We show the following somewhat surprising implication: If our hypothesis holds for dense enough instances, then it is hard to color an O(1)-colorable hyper graph with a polynomial number of colors.

Let C be a (fan-in 2) Boolean circuit of size s and depth d, and let x be an input for C. Assume that a verifier that knows C but doesn't know x can access the low degree extension of x at one random point. Two competing provers try to convince the verifier that C(x)=0 and C(x)=1, respectively, and assume that one of the provers is honest.

We study the interactive channel capacity of an ε-noisy channel. The interactive channel capacity C(ε) is defined as the minimal ratio between the communication complexity of a problem (over a non-noisy channel), and the communication complexity of the same problem over the binary symmetric channel with noise rate ε, where the communication complexity tends to infinity.

The Unique Games Conjecture (UGC) is an important open problem in the research of PCPs and hardness of approximation. The conjecture is a strengthening of the PCP Theorem, predicting the existence of a special type of PCP verifiers: 2-query verifiers that only make unique tests. Moreover, the UGC predicts that such PCP verifiers can have almost-perfect completeness and low-soundness.

We consider the rational versions of two of the classical problems in foundations of cryptography: secret sharing and multiparty computation, suggested by Halpern and Teague (STOC 2004). Our goal is to design games and fair strategies that encourage rational participants to exchange information about their inputs for their mutual benefit, when the only mean of communication is a broadcast channel.