【抄録】The network of bloggers interconnected by the comment exchange relationship obviously represents an aspect of human relationships. The paper reveals that interrelationships among items (such as products and works of art) can also be inferred based on structural characteristics of the network as follows. First, for each of two items in question, a set of bloggers writing about the item is respectively built. By "plotting" members of each blogger set on the network described above, distributions of the blogger sets are obtained. Then, selecting an appropreate index for measuring proximity of the distributions brings in a correlation between the proximity and the relevance of the items.
【抄録】In this paper, we show the possibility of predicting the anomalous packets' behaviors to the near active addresses from small observation address space (Darknet) in Internet. We have proposed the distributed cooperative monitoring architecture (DCMA) which probes the anomalous packets that arrive at the distributed unused address segments and detects and defenses anomalous packets' behaviors to the near active addresses. To realize DCMA, it is necessary to investigate the time-series correlation between anomalous packets arriving at small observation address segments and those of near addresses. Thus, we calculated the correlation strength of anomalous packets that scan address segments from the pairs of the sub-observation address segments divided from the Darknet addresses. Furthermore, we observed the correlation strength when changing the sub-observation's size and investigated the size dependency of the correlation strength. As a result, we could indicate the possibility of predicting the anomalous packets' behaviors to the near address segments from small sub-observation addresses. We could also find that the base observation fixed to the specific sub-observation space contributes to the strong correlation coefficient. Therefore, these results imply that DCMA can predict the anomalous packets' behaviors to the near addresses using small observation space.
【抄録】The Internet consists of several thousand interconnected autonomous systems (ASes). For enabling the autonomous network management that ensures stable access at the inter-AS level and flexible inter-domain routing control, it is important to know how the routing information originated from an AS spreads throughout the Internet and to control inter-AS routing information using feedback actions based on observed network status at multiple ASes. Because each AS is controlled by a single administrative authority based on that AS's own policy, a cooperative distributed problem solving is desirable. To cope with these requirements, we have proposed a multi-agent-based inter-AS diagnostic system called ENCORE and inter-AS routing adjustment system called AISLE / VR. They consist of a collection of intelligent agents that are located in multiple ASes and perform collective observation, analysis, and control. These systems are constructed on the agent platform that provides utility functions on distributed environments. For the purpose of flexibility for incremental design and modification through the development and application phases, we adopted Lisp as the base language and have constructed stable systems that can demonstrate the effectiveness of our basic design for autonomous inter-AS network management. Copyright 2009 ACM.
【抄録】This paper discusses an availability of statistical property of network topology for the server development and server selection algorithm in distributed systems. As a first step, we introduce new server development and server selection algorithms based on a degree of a server - the number of links for a server - and a scope - coarse grained topological information around a server. Our simulation reveals that the use of both information as a static QoS parameter improves the fairness of server load in a real Internet AS-level topology.
【抄録】Unexpected temporal and spatial changes of inter-AS routing behavior often lead to the necessity of on-demand inter-domain routingadjustment. For resolving this problem, we apply the AISLE framework, which is a multi-agent-based model, to a policy-based routingadjustment system for transit ISPs and their customer ASs. This paper describes the BGP-control architecture called VR (Virtual Router) that can dynamically change forwarding paths considering alternative paths, which are inferred from historical data and confirmed when they are actually applied. VR can control conventional multiple border routers in an AS without any protocol extensions. The policy description, which is interpreted by an agent, enables network operators to define autonomous actions for analyzing network status and adjusting inter-AS routing based on these observed results by issuing requests to VR. Some evaluation results indicate that VR can effectively change routing over BGP data on the actual Internet and some control scenarios based on policy descriptions demonstrate the validity of our basic design framework. Copyright 2006 ACM.
【抄録】In the multi-agent systems, even if each agent behaves selfish or cooperative, its behavior influences the other agents' behaviors and their behaviors influence its behavior cyclically. Thereore a certain information about the relative position of each agent among all the agents may be able to be extracted only from its behavior. If this information can be extracted, each agent has the possibility to improve its efficiency only by seeing its own behavior without knowing about the information of the other agents. In this paper, we verified this hypothesis by using the competitive multi-agent simulation environment named Minority Game, and confirmed that each agent's performance gain was actually possible. This analytical technique may be especially useful for massively multi-agent systems. Copyright 2006 ACM.
【抄録】In order to achieve efficient progress in activities such as e-commerce and e-transactions in an open environment like the Internet, an agent must choose appropriate partner agents for collaboration. However, agents have no global information about the whole multi-agent system (MAS) and the state of the Internet; therefore, they must select the appropriate partners based on local knowledge and local observations. In this paper, using a multi-agent simulation, we discuss how total MAS performances are affected by local decisions when agents select partners to collaborate with. We also investigate how MAS performances change and how network structures between agents shift according to the progress of agents' local learning and observations. We then discuss the relationship between task load and agent network structure. This relates to estabilishing the optimum time when agents should learn about appropriate partners in an actual environment. Copyright 2006 ACM.
【抄録】This paper proposes a learning method to select the most appropriate abstract plans during hierarchical planning in the context of multi-agent systems (MAS). In hierarchical planning, a plan is first created at the most abstract level, and is then refined to a more concrete plan, level by level. Thus, selecting an appropriate plan at the abstract level is very important because the selected plan restricts the scope of lower concrete-level plans. This restriction can enable agents to create plans efficiently, but if all the plans under the selected plan contain serious and difficult-to-resolve conflicts with other agents' plans, the resulting plan does not work well or is of low quality. We propose a method in which, from the conflict pattern among agents' plans, an agent learns which abstract plans will cause conflicts with less probability or which conflicts are easy to resolve, thus inducing probabilistically higher-utility concrete plans after conflict resolution. We also show some experimental results to evaluate our method, with the results suggesting structures of resources where tasks are executed. Copyright 2005 ACM.
An implementation of flexible-playtime video skimming
柔軟なプレイタイムの動画スキミングの実装
Shigemi Aoyagi, Ken'ich Kourai, Koji Sato, Toshihiro Takada, Toshiharu Sugawara, Rikio Onai Proceedings of SPIE - The International Society for Optical Engineering
【抄録】In this paper, we propose a new time-reduction method for video skimming in which the focus is on the overall playback time. While fast-forwarding is a natural way to check whether or not items are of interest, the sound is not synchronized with the images and the lack of comprehensible audio data means that we must work from the images alone. The focus in video summarization has been solely on video segmentation, i.e. building a structure that represents the parts and flow of meaning in the video. In our system, the user simply specifies the running time required for the summarized video. We describe the current state of our prototype system and its results in testing, which show how well it works.
Reusing coordination and negotiation strategies in multi-agent systems for ubiquitous network environment(First author)
ユビキタスネットワーク環境のためのマルチエージェントシステムにおける調整・交渉戦略の再利用
Toshiharu Sugawara, Satoshi Kurihara, Kensuke Fukuda, Toshio Hirotsu, Shigemi Aoyagi, Toshihiro Takada Proceedings of the Third International Joint Conference on Autonomous Agents and Multiagent Systems, AAMAS 2004
【抄録】Recently, we proposed an intelligent ubiquitous computing (ubicomp) environment where sensors and/or their stations/servers have CPUs to cooperatively learn generalized series of sensed events that are involved in human activities. This can be regarded as a multi-agent application. Because ubicomp applications target support for daily-life activities, one of their characteristics is that the same/similar series of events occurs frequently. Multi-agent plans in applications of this type are used to foresee human activities and generate programs to assist them. Therefore, the same planning processes for conflict detection and resolution recur. This paper proposes a learning method in which past plans are exploited for problem solving in an environment where the same/similar problems appear repeatedly. We discuss how the plan is stored and reused using as an example the exploration of conflict-free routes in a room and then describe experimental results.
【抄録】In this paper, we analyze a simple adaptive model of competition called the Minority Game, which is used in analyzing competitive phenomena in markets, and suggest that the core elements required for the formation of self-organization are: (i) rules that place a good constraint on each agent's behavior, and (ii) rules that lead to indirect coordination, which is called "stigmergy." Finally, we tested the points suggested by this research in solving the El Farol's bar problem, which is an extended version of the Minority Game.
【抄録】The storage and analysis of the hierarchical planning results in the past to identify implicit costs and resource relationships between activities in multi-agent systems are discussed. To realize effective activities of agents, coordination and cooperation is essential in multi-agent systems. The used plans are retained with costs, time, used resources, conflicts among agents and the environments. The observed data for each primitive task sequence which indicates performance of tasks such as execution time and the period and amount of resource usage is compared.
Article 2002 5 IEEE : Institute of Electrical and Electronics Engineers
Agent system for inter-AS routing error diagnosis
エージェントシステムによるAS間ルーティングエラー診断
Osamu Akashi, Toshiharu Sugawara, Ken Ichiro Murakami, Mitsuru Maruyama, Keiichi Koyanagi IEEE Internet Computing
【抄録】The agent system for inter-autonomous system (AS) routing error diagnosis is discussed. The multiagent-based model is based on a reflector model of autonomous agents that collect and analyze raw routing information from many border routers while the Encore system is a diagnostic ensemble system that locates cooperative reflector agents in multiple AS. The test results shows that Encore found inter-AS anomalies that originated in a discontinous upstream Internet service provider (ISP).
Clea: A Framework for the Coordination of Applications and Networks(Last author)
Clea: アプリケーションとネットワークの協調のためのフレームワーク
Koji Sato, Toshio Hirotsu, Kensuke Fukuda, Osamu Akashi, Toshiharu Sugawara Proceedings of the Joint Conference on Information Sciences
【抄録】This paper proposes Clea, a framework for the coordination of applications and networks. Clea conveys requests from applications to networks and enables information on network characteristics, status, and functions to be used by applications. This makes it possible for an application to adapt flexibly to dynamic changes in network status and to utilize network resources effectively. Clea also enables coordination of applications and networks to be described in a uniform and concise manner.
Conference Paper 2000 IEEE : Institute of Electrical and Electronics Engineers
Multiagent-based cooperative inter-AS diagnosis in ENCORE
ENCOREにおけるマルチエージェントベースの協調的AS間診断
O. Akashi, Toshiharu Sugawara, K. Murakami, M. Maruyama, N. Takahashi IEEE Symposium Record on Network Operations and Management Symposium
【抄録】To cope with the requirements of the Internet, a multiagent-based inter-autonomous system (AS) diagnostic system called ENCORE is proposed. In this system, a collection of intelligent agents are located in multiple ASs and perform collective observation and analysis. The agents autonomous observation actions and their ability to perform cooperative and efficient analysis of routing anomalies are detailed. Further, the effectiveness and limitations of ENCORE based on analysis of a real AS are discussed.
Article 1999 IEEE : Institute of Electrical and Electronics Engineers
Dynamic multimedia integration with the WWW(Last author)
WWWによる動的なマルチメディアの統合
Koji Sato, Toshihiro Takada, Shigemi Aoyagi, Toshio Hirotsu, Toshiharu Sugawara IEEE Pacific RIM Conference on Communications, Computers, and Signal Processing - Proceedings
【抄録】This paper presents a novel framework for seamlessly integrating continuous media, such as audio and video, with the World Wide Web (WWW). Continuous Media with the Web (Cmew) enhances the interactivity of continuous media by associating hyperlinks with spatial-temporal parts of the media. The scenario control architecture in Cmew provides flexible and dynamic control over continuous media in multimedia documents. The Cmew media player has been implemented as a Java applet, which enables its use in the current WWW environment.
【抄録日本語訳】本論文では、オーディオやビデオなどの連続メディアをワールドワイドウェブ(WWW)とシームレスに統合するための新しいフレームワークを紹介する。Continuous Media with the Web (Cmew) は、ハイパーリンクをメディアの空間的・時間的部分に関連付けることによって、連続的メディアのインタラクティビティを強化する。Cmewのシナリオ制御アーキテクチャは、マルチメディア文書中の連続メディアに対して柔軟かつ動的な制御を提供する。Cmewメディアプレーヤーは、Javaアプレットとして実装されており、現在のWWW環境での利用を可能にしている。
Conference Paper 1999 1 1 IEEE : Institute of Electrical and Electronics Engineers
Cmew/U-a multimedia Web annotation sharing system(Last author)
Cmew/U-a マルチメディアWebアノテーション共有システム
T. Hirotsu, T. Takada, S. Aoyagi, K. Sato, Toshiharu Sugawara IEEE Region 10 Annual International Conference, Proceedings/TENCON
Learning to Improve Coordinated Actions in Cooperative Distributed Problem-Solving Environments(First author)
協調型分散問題解決環境における協調行動の改善学習
Toshiharu Sugawara, Victor Lesser Machine Learning
【抄録】Coordination is an essential technique in cooperative, distributed multiagent systems. However, sophisticated coordination strategies are not always cost-effective in all problem-solving situations. This paper presents a learning method to identify what information will improve coordination in specific problem-solving situations. Learning is accomplished by recording and analyzing traces of inferences after problem solving. The analysis identifies situations where inappropriate coordination strategies caused redundant activities, or the lack of timely execution of important activities, thus degrading system performance. To remedy this problem, situation-specific control rules are created which acquire additional nonlocal information about activities in the agent networks and then select another plan or another scheduling strategy. Examples from a real distributed problem-solving application involving diagnosis of a local area network are described.
Construction of a knowledge base by means of inter-frame structures(Single author)
フレーム間構造による知識基盤の構築
Toshiharu Sugawara Systems and Computers in Japan
【抄録】When the scale of a knowledge base is enlarged, the time required to construct it increases, and its management is made difficult. In a frame-based knowledge base, a number of common structures exists in interframe relations. This paper proposes a system in which the knowledge base is regarded as a sample in the construction process, and the common structure is extracted as a rule by which the later construction of the knowledge base is supported. Since the knowledge base contains exceptions, the exceptions are eliminated statistically in the rule extraction. A maintenance mechanism also is introduced, where the rule is rewritten on the basis of the result of application of the rules. The proposed system can also be used for checking the knowledge base to detect whether there is a mismatch.
Conference Paper 1993 IEEE : Institute of Electrical and Electronics Engineers
Using action benefits and plan certainties in multiagent problem solving(Single author)
マルチエージェントの問題解決における行動の利点と計画の確実性の利用
Toshiharu Sugawara
【抄録】Choosing socially coherent and rational actions is essential in multiagent problem solving. In some domains, exchanging agents' plans are helpful for understanding what are rational actions. If they has little shared knowledge or environment, however, it is hard to understand other agents' plans. This paper discusses the utility-based cooperation for this situation. A utility matrix are created based on the local plans and through communications with other agents instead of exchanging plans. Utility numbers are calculated according to action benefits and plan certainties. Intuitively, an action benefit expresses the importance of performing or verifying the current plan, and a plan certainty expresses how strongly the agent making the plan believes that it is correct or effective for the current problem solving. Actions based on a plan supported by many proofs have high utility-numbers and so are priority over other actions. Finally, we will show how the performance can be improved by our method through experiments.
Conference Paper 1990 IEEE : Institute of Electrical and Electronics Engineers
A cooperative LAN diagnostic and observation expert system(Single author)
協調型 LAN 診断・観測エキスパートシステム
Toshiharu Sugawara
【抄録】The author describes a LAN diagnostic expert system called the Large Internetwork observation and Diagnostic Expert System (LODES), which has transport=control=protocol/internet=protocol-(TCP/IP-) related and troubleshooting knowledge, and can handle problems occurring in a complex and heterogeneous network environment. Each constituent network has its own expert system containing management information about the local network. TCP/IP-related problems are thus resolved singly or cooperatively, depending on their type. This system also includes a network observer, which monitors packets flowing in the network and can find a number of problems and problem indicators for automatic problem detection and troubleshooting. The configuration of LODES is described, as well as the knowledge and control methods for cooperative diagnosis, especially the method for cooperative diagnostic task scheduling.
KRINE: A KNOWLEDGE BASE SYSTEM WITH FRAME-BASED LOGIC PROGRAMMING MECHANISMS.(Last author)
クリン:フレームベースの論理プログラミング機構を持つ知識ベースシステム
Yutaka Ogawa, Kenichi Shima, Toshiharu Sugawara
【抄録】This paper describes design objectives and mechanisms for the knowledge representation and inference environment: KRINE. KRINE is made for building a computer hardware logic design expert system and has the following features: (1) frame-based knowledge representation of hierarchical design objects; (2) frame-based logic programming mechanisms for design knowledge representation and; (4) other knowledge representation paradigms, such as rules, procedures, manipulations or graphics, are integrated with the frame mechanisms.
【抄録】To develop an expert system for intelligent CAD, entails much difficulty in describing the design knowledge including pattern matching procedures, in applying the design knowledge by trial and error or in displaying design objects dynamically, if only conventional object-oriented or frame-based knowledge representation tools are used. To cope with these problems, the following mechanisms were integrated with the frame-based knowledge representation system: (a) PROLOG programming functions with direct frame data unifier, (b) frame data recovery for backtracking of inference, and (c) design object display functions by interpreting frame data directly. This paper describes the design philosophy, mechanisms and examples of an integrated knowledge representation and inference environment (KRINE).