There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

다중혈관 관상동맥 환자에서 y-문합을 이용하여 양쪽 내흉동맥만을 사용한 우회술의 조기 성적

Imagined communities: Reflections on the origin and spread of nationalism

The modern science of networks has brought significant advances to our understanding of complex systems. One of the most relevant features of graphs representing real systems is community structure, or clustering, i. e. the organization of vertices in clusters, with many edges joining vertices of the same cluster and comparatively few edges joining vertices of different clusters. Such clusters, or communities, can be considered as fairly independent compartments of a graph, playing a similar role like, e. g., the tissues or the organs in the human body. Detecting communities is of great importance in sociology, biology and computer science, disciplines where systems are often represented as graphs. This problem is very hard and not yet satisfactorily solved, despite the huge effort of a large interdisciplinary community of scientists working on it over the past few years. We will attempt a thorough exposition of the topic, from the definition of the main elements of the problem, to the presentation of most methods developed, with a special focus on techniques designed by statistical physicists, from the discussion of crucial issues like the significance of clustering and how methods should be tested and compared against each other, to the description of applications to real networks.

/pdf/community-detection-in-graphs-34rogm3r6p.pdf

Community detection in graphs

In many online machine learning and data science tasks such as data summarisation and feature compression, d-dimensional vectors are usually distributed across a large number of clients in a decentralised network and collected in a streaming fashion. This is increasingly common in modern applications due to the sheer volume of data generated and the clients’ constrained resources. In this setting, some clients are required to compute an update to a centralised target model independently using local data while other clients aggregate these updates with a low-complexity merging algorithm. However, some clients with limited storage might not be able to store all of the data samples if d is large, nor compute procedures requiring at least Ω(d) storage-complexity such as Principal Component Analysis, Subspace Tracking, or general Feature Correlation. In this work, we present a novel federated algorithm for PCA that is able to adaptively estimate the rank r of the dataset and compute its r leading principal components when only O(dr) memory is available. This inherent adaptability implies that r does not have to be supplied as a fixed hyper-parameter which is beneficial when the underlying data distribution is not known in advance, such as in a streaming setting. Numerical simulations show that, while using limited-memory, our algorithm exhibits state-of-the-art performance that closely matches or outperforms traditional non-federated algorithms, and in the absence of communication latency, it exhibits attractive horizontal scalability.

Federated PCA with Adaptive Rank Estimation.

The explosion of bots on the Web brings an unprecedented increase in traffic from non-human sources. This work studies bot traffic on Twitter, finding that almost 50% of traffic is generated and propagated by a rapidly growing bot population -- a major concern for networked systems in the future.

https://dl.acm.org/doi/pdf/10.1145/3123878.3131983

The Implications of Twitterbot Generated Data Traffic on Networked Systems

RFC3453: The Use of Forward Error Correction (FEC) in Reliable Multicast

Discusses experiences of the authors in transmitting the proceedings of some events at IEEE GLOBECOM '96 in London, England, during November 17-22, 1996. Live video and audio of all the events in the Churchill Auditorium of the Queen Elizabeth II Conference Center were captured and transmitted, in real time, as well as stored and transmitted later, for remote participants in three continents, over the Internet. Two independent systems were used simultaneously, one supplied by researchers from NTT Laboratories in Japan and the other by researchers from University College London. The former system is based on a server model of distribution, while the latter is based on the use of network-level packet multicast. Both systems employ compression algorithms, so the network capacity requirement in each case was on the order of 100 kb/s to 200 kb/s total, thus enabling remote participants without very high-end network connectivity to take part. Receivers only need software for a PC running most popular versions of Windows or a UNIX workstation to be able to receive either type of transmission, or to retrieve the recorded sessions from NTT Laboratories servers. The multimedia transmission was carried over carefully engineered links that traversed many different subnet technologies, including point-to-point circuits, SMDS networks, ATM networks, and fast Ethernet switches. This was both to give a high level of assurance that the traffic would not experience too much interference from other traffic, and to ensure very low packet store and forward delays. The system ran for four days continuously, and was generally very successful.

/pdf/real-time-audio-and-video-transmission-of-ieee-globecom-96-4wmyh6jnix.pdf

Real-time audio and video transmission of IEEE GLOBECOM '96 over the Internet

Wireless sensor networks are emerging technologies that offer low cost, distributed monitoring solutions for a wide variety of applications and systems. One application driving the development of sensor networks is the reporting of conditions within a region of interest where the environment can abruptly change due to an sudden event, such as enemy and target movements on the battlefield, or biochemical attacks, fires, etc. Our work focuses on sensor systems that need to efficiently deliver information during and immediately following an event that triggers such an abrupt change. Congestion control and load balancing are critical issues in such sensor networks where the sensor field can move instantaneously from almost zero load to overload conditions. It is during these impulse or overload periods that the events in transit are of most importance and most likely to be lost due to congestion. Existing congestion control algorithms [1] [2] [3] are limited under these conditions because they rely on rate control or packet drop mechanisms at source or intermediate sensor nodes that can significantly impact the application’s fidelity, as measured at one or more physical sinks. We propose randomly distributing a small number of allwireless dual radio virtual sinks throughout the sensor field that are capable of offering the existing low-power sensor network enhanced congestion avoidance support when persistent congestion is detected. In essence virtual sinks operate as safety valves in the sensor field to selectively siphon off high load traffic in order to maintain the fidelity of the application signal at the physical sink and to alleviate the funneling effect, as discussed in Section 2.1. We call these specialized nodes virtual sinks to distinguish them from physical sinks, which typically have a wireline

Jon Crowcroft

Papers

Federated PCA with Adaptive Rank Estimation.

The Implications of Twitterbot Generated Data Traffic on Networked Systems

RFC3453: The Use of Forward Error Correction (FEC) in Reliable Multicast

Real-time audio and video transmission of IEEE GLOBECOM '96 over the Internet

A case for all-wireless, dual-radio virtual sinks