To my mother Maxine, who gave me a love of learning; to Susan, who is as happy and amazed as I am that The Book is finally completed; to Josh, Tim, and Teddy, who are impressed that their father is an Author; and to my late father George, who would have been proud.

No abstract

We argue that the right way to understand distributed protocols is by considering how messages change the state of knowledge of a system. We present a hierarchy of knowledge states that a system may be in, and discuss how communication can move the system's state of knowledge of a fact up the hierarchy. Of special interest is the notion of common knowledge. Common knowledge is an essential state of knowledge for reaching agreements and coordinating action. We show that in practical distributed systems, common knowledge is not attainable. We introduce various relaxations of common knowledge that are attainable in many cases of interest. We describe in what sense these notions are appropriate, and discuss their relationship to each other. We conclude with a discussion of the role of knowledge in distributed systems.

Abstract-A face recognition system must recognize a face from a novel image despite the variations between images of the same face. A common approach to overcoming image variations because of changes in the illumination conditions is to use image representations that are relatively insensitive to these variations. Examples of such representations are edge maps, image intensity derivatives, and images convolved with 2D Gabor-like filters. Here we present an empirical study that evaluates the sensitivity of these representations to changes in illumination, as well as viewpoint and facial expression. Our findings indicated that none of the representations considered is sufficient by itself to overcome image variations because of a change in the direction of illumination. Similar results were obtained for changes due to viewpoint and expression. Image representations that emphasized the horizontal features were found to be less sensitive to changes in the direction of illumination. However, systems based only on such representations failed to recognize up to 20 percent of the faces in our database. Humans performed considerably better under the same conditions. We discuss possible reasons for this superioriority and alternative methods for overcoming illumination effects in recognition.

By anaizng the stares of novledge Li 'he processors attain 19 an urre as.e SYS Tern of a sImp.e tYpe. we capture some o: :rs;c uxder!ying struc t ire of such svsams. particu ar. we study s hat facts become 'on mu kit 0 uiedgc at var:ous ponts in the execu ion of protoco!s n an un reliable system. This characterizes the sirnu Itaneous ac ions hat can he carHed 0:': in s:ch systems. For et ne. we obtain a ccco!e:e charac erizat 0:1 of the nurnSer of ro-iLds rccuired o reach ." ''uUu,eou.s By:anne ,Lq€rr:ent. giver. the pa: tert in hich fa:ares 0CC U. From ri we derive a new protcico: for th: p rbern t ha is optimal in a(l ru7. rarher than just alavs matching the worst-case lower bound. Er some cases this pro? oco a, ains S rnulta ii togs Rvza nt me Agreenient In as few as 2 rounds. e also present a non-trivial simu [tarieou agreement problem called b ualerfl agreement for which there is a protocol that aiways halts in two rounds. Our analysis applies to simultaneous actions in general, and not just to Byzantine agreement. The lower bound proofs presented here generalize and siniplifv ihe previously known proofs.

Abstract. This work applies the theory of knowledge in distributed systems to the design of efficient fault-tolerant protocols. We define a large class of problems requiring coordinated, simultaneous action in synchronous systems, and give a method of transforming specifications of such problems into protocols that are optimal in all runs: these protocols are guaranteed to perform the simultaneous actions as soon as any other protocol could possibly perform them, given the input to the system and faulty processor behavior. This transformation is performed in two steps. In the first step we extract, directly from the problem specification, a high-level protocol programmed using explicit tests for common knowledge. In the second step we carefully analyze when facts become common knowledge, thereby providing a method of efficiently implementing these protocols in many variants of the omissions failure model. In the generalized omissions model, however, our analysis shows that testing for common knowledge is NP-hard. Given the close correspondence between common knowledge and simultaneous actions, we are able to show that no optimal protocol for any such problem can be computationally efficient in this model. The analysis in this paper exposes many subtle differences between the failure models, including the precise point at which this gap in complexity occurs.

We present a novel method for recovering the 3D structure and scene flow from calibrated multi-view sequences. We propose a 3D point cloud parametrization of the 3D structure and scene flow that allows us to directly estimate the desired unknowns. A unified global energy functional is proposed to incorporate the information from the available sequences and simultaneously recover both depth and scene flow. The functional enforces multi-view geometric consistency and imposes brightness constancy and piecewise smoothness assumptions directly on the 3D unknowns. It inherently handles the challenges of discontinuities, occlusions, and large displacements. The main contribution of this work is the fusion of a 3D representation and an advanced variational framework that directly uses the available multi-view information. The minimization of the functional is successfully obtained despite the non-convex optimization problem. The proposed method was tested on real and synthetic data.

Reasoning about knowledge seems to play a fundamental role in distributed systems. Indeed, such reasoning is a central part of the informal intuitive arguments used in the design of distributed protocols. Communication in a distributed system can be viewed as the act of transforming the system's state of knowledge. This paper presents a general framework for formalizing and reasoning about knowledge in distributed systems. It is shown that states of knowledge of groups of processors are useful concepts for the design and analysis of distributed protocols. In particular, distributed knowledge corresponds to knowledge that is "distributed" among the members of the group, while common knowledge corresponds to a fact being "publicly known." The relationship between common knowledge and a variety of desirable actions in a distributed system is illustrated. Furthermore, it is shown that, formally speaking, in practical systems common knowledge cannot be attained. A number of weaker variants of common knowledge that are attainable in many cases of interest are introduced and investigated.