The goal of precipitation nowcasting is to predict the future rainfall intensity in a local region over a relatively short period of time. Very few previous studies have examined this crucial and challenging weather forecasting problem from the machine learning perspective. In this paper, we formulate precipitation nowcasting as a spatiotemporal sequence forecasting problem in which both the input and the prediction target are spatiotemporal sequences. By extending the fully connected LSTM (FC-LSTM) to have convolutional structures in both the input-to-state and state-to-state transitions, we propose the convolutional LSTM (ConvLSTM) and use it to build an end-to-end trainable model for the precipitation nowcasting problem. Experiments show that our ConvLSTM network captures spatiotemporal correlations better and consistently outperforms FC-LSTM and the state-of-the-art operational ROVER algorithm for precipitation nowcasting.

Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting

In Distributed Algorithms, Nancy Lynch provides a blueprint for designing, implementing, and analyzing distributed algorithms. She directs her book at a wide audience, including students, programmers, system designers, and researchers.



Distributed Algorithms contains the most significant algorithms and impossibility results in the area, all in a simple automata-theoretic setting. The algorithms are proved correct, and their complexity is analyzed according to precisely defined complexity measures. The problems covered include resource allocation, communication, consensus among distributed processes, data consistency, deadlock detection, leader election, global snapshots, and many others.



The material is organized according to the system model-first by the timing model and then by the interprocess communication mechanism. The material on system models is isolated in separate chapters for easy reference.



The presentation is completely rigorous, yet is intuitive enough for immediate comprehension. This book familiarizes readers with important problems, algorithms, and impossibility results in the area: readers can then recognize the problems when they arise in practice, apply the algorithms to solve them, and use the impossibility results to determine whether problems are unsolvable. The book also provides readers with the basic mathematical tools for designing new algorithms and proving new impossibility results. In addition, it teaches readers how to reason carefully about distributed algorithms-to model them formally, devise precise specifications for their required behavior, prove their correctness, and evaluate their performance with realistic measures.


Table of Contents

1 Introduction 
2 Modelling I; Synchronous Network Model 
3 Leader Election in a Synchronous Ring 
4 Algorithms in General Synchronous Networks 
5 Distributed Consensus with Link Failures 
6 Distributed Consensus with Process Failures 
7 More Consensus Problems 
8 Modelling II: Asynchronous System Model 
9 Modelling III: Asynchronous Shared Memory Model 
10 Mutual Exclusion 
11 Resource Allocation 
12 Consensus 
13 Atomic Objects 
14 Modelling IV: Asynchronous Network Model 
15 Basic Asynchronous Network Algorithms 
16 Synchronizers 
17 Shared Memory versus Networks 
18 Logical Time 
19 Global Snapshots and Stable Properties 
20 Network Resource Allocation 
21 Asynchronous Networks with Process Failures 
22 Data Link Protocols 
23 Partially Synchronous System Models 
24 Mutual Exclusion with Partial Synchrony 
25 Consensus with Partial Synchrony

Distributed algorithms

/pdf/convolutional-lstm-network-a-machine-learning-approach-for-3mrj9pedfq.pdf

Convolutional LSTM Network: a machine learning approach for precipitation nowcasting

Multiagent systems are made up of multiple interacting intelligent agents -- computational entities to some degree autonomous and able to cooperate, compete, communicate, act flexibly, and exercise control over their behavior within the frame of their objectives They are the enabling technology for a wide range of advanced applications relying on distributed and parallel processing of data, information, and knowledge relevant in domains ranging from industrial manufacturing to e-commerce to health care This book offers a state-of-the-art introduction to multiagent systems, covering the field in both breadth and depth, and treating both theory and practice It is suitable for classroom use or independent study This second edition has been completely revised, capturing the tremendous developments in multiagent systems since the first edition appeared in 1999 Sixteen of the book's seventeen chapters were written for this edition; all chapters are by leaders in the field, with each author contributing to the broad base of knowledge and experience on which the book rests The book covers basic concepts of computational agency from the perspective of both individual agents and agent organizations; communication among agents; coordination among agents; distributed cognition; development and engineering of multiagent systems; and background knowledge in logics and game theory Each chapter includes references, many illustrations and examples, and exercises of varying degrees of difficulty The chapters and the overall book are designed to be self-contained and understandable without additional material Supplemental resources are available on the book's Web site Contributors:Rafael Bordini, Felix Brandt, Amit Chopra, Vincent Conitzer, Virginia Dignum, Jurgen Dix, Ed Durfee, Edith Elkind, Ulle Endriss, Alessandro Farinelli, Shaheen Fatima, Michael Fisher, Nicholas R Jennings, Kevin Leyton-Brown, Evangelos Markakis, Lin Padgham, Julian Padget, Iyad Rahwan, Talal Rahwan, Alex Rogers, Jordi Sabater-Mir, Yoav Shoham, Munindar P Singh, Kagan Tumer, Karl Tuyls, Wiebe van der Hoek, Laurent Vercouter, Meritxell Vinyals, Michael Winikoff, Michael Wooldridge, Shlomo Zilberstein

Multiagent Systems

From the Publisher:
Among the many approaches to formal reasoning about programs, Dynamic Logic enjoys the singular advantage of being strongly related to classical logic. Its variants constitute natural generalizations and extensions of classical formalisms. For example, Propositional Dynamic Logic (PDL) can be described as a blend of three complementary classical ingredients: propositional calculus, modal logic, and the algebra of regular events. In First-Order Dynamic Logic (DL), the propositional calculus is replaced by classical first-order predicate calculus. Dynamic Logic is a system of remarkable unity that is theoretically rich as well as of practical value. It can be used for formalizing correctness specifications and proving rigorously that those specifications are met by a particular program. Other uses include determining the equivalence of programs, comparing the expressive power of various programming constructs, and synthesizing programs from specifications. 
This book provides the first comprehensive introduction to Dynamic Logic. It is divided into three parts. The first part reviews the appropriate fundamental concepts of logic and computability theory and can stand alone as an introduction to these topics. The second part discusses PDL and its variants, and the third part discusses DL and its variants. Examples are provided throughout, and exercises and a short historical section are included at the end of each chapter.

Dynamic Logic

With hardware transactional memory (HTM) becoming available in mainstream processors, lock-based critical sections may now initiate a hardware transaction instead of taking the lock, enabling their concurrent execution unless a real data conflict occurs. However, just a few transactional aborts can cause the lock to be acquired non-transactionally resulting in the serialization of all the threads, severely degrading the amount of speedup obtained. In this paper we provide two software extension mechanisms that considerably improve the concurrency and speedup levels attained by lock based programs using HTM-based lock elision. The first sacrifices opacity to achieve higher levels of concurrency, and the second retains opacity while reaching slightly lower levels of concurrency.Evaluation on STAMP and on data structure benchmarks on an Intel Haswell processor shows that these techniques improve the speedup by up to 3.5 times and $10$ times respectively, compared to using Haswell's hardware lock elision as is.

/pdf/software-improved-hardware-lock-elision-499o87uwxp.pdf

Software-improved hardware lock elision

Producer-consumer pools, that is, collections of unordered objects or tasks, are a fundamental element of modern multiprocessor software and a target of extensive research and development. For example, there are three common ways to implement such pools in the Java JDK6.0: the SynchronousQueue, the LinkedBlockingQueue, and the ConcurrentLinkedQueue. Unfortunately, most pool implementations, including the ones in the JDK, are based on centralized structures like a queue or a stack, and thus are limited in their scalability.

This paper presents the ED-Tree, a distributed pool structure based on a combination of the elimination-tree and diffracting-tree paradigms, allowing high degrees of parallelism with reduced contention. We use the ED-Tree to provide new pool implementations that compete with those of the JDK.

In experiments on a 128 way Sun Maramba multicore machine, we show that ED-Tree based pools scale well, outperforming the corresponding algorithms in the JDK6.0 by a factor of 10 or more at high concurrency levels, while providing similar performance at low levels.

/pdf/scalable-producer-consumer-pools-based-on-elimination-4kuxkxf1hc.pdf

Scalable producer-consumer pools based on elimination-diffraction trees

We introduce a new primitive, the Resource Controller, which abstracts the problem of controlling the total amount of resources consumed by a distributed algorithm. We present an efficient distributed algorithm to implement this abstraction. The message complexity of our algorithm per participating node is polylogarithmic in the size of the network, compared to the linear cost per node of the naive algorithm. The implementation of our algorithm is simple and practical and the techniques used are interesting because a global quantity is managed in a distributed way. The Resource Controller can be used to construct efficient algorithms for a number of important problems, such as the problem of bounding the worst-case message complexity of a protocol and the problem of dynamically assigning unique names to nodes participating in a protocol.

Local management of a global resource in a communication network

Sparser: a paradigm for running distributed algorithms

An algorithm and two lower bounds are presented for the problem of constructing and maintaining routing schemes in dynamic networks. The algorithm distributively assigns addresses to nodes and constructs routing tables in a dynamically growing tree. The resulting scheme routes data messages over the shortest path between any source and destination, assigns addresses of O(log/sup 2/n) bits to each node, and uses in its routing table O(log/sup 3/n) bits of memory per incident link, where n is the final number of nodes in the tree. The amortized communication cost of the algorithm is O(log n) messages per node. Also given are two lower bounds on the tradeoff between the quality of routing schemes (i.e. their stretch factor) and their amortized communication cost in general dynamic networks. >

Yehuda Afek

Papers

Software-improved hardware lock elision

Scalable producer-consumer pools based on elimination-diffraction trees

Local management of a global resource in a communication network

Sparser: a paradigm for running distributed algorithms

Upper and lower bounds for routing schemes in dynamic networks