文档库 最新最全的文档下载
当前位置:文档库 › EMADS an extendible multi-agent data miner

EMADS an extendible multi-agent data miner

EMADS an extendible multi-agent data miner
EMADS an extendible multi-agent data miner

EMADS:AN EXTENDIBLE MULTI-AGENT DATA MINER

Kamal Ali Albashiri,Frans Coenen,and Paul Leng

Abstract In this paper we describe EMADS,an Extendible Multi-Agent Data min-ing System.The EMADS vision is that of a community of data mining agents,con-tributed by many individuals,interacting under decentralised control to address data mining requests.EMADS is seen both as an end user application and a research tool. This paper details the EMADS vision,the associated conceptual framework and the current implementation.Although EMADS may be applied to many data mining tasks;the study described here,for the sake of brevity,concentrates on agent based data classi?cation.A full description of EMADS is presented.

Keywords Multi-Agent Data Mining(MADM),Classi?er Generation.

1Introduction

Multi-Agent Systems(MAS)offer a number of general advantages with respect to Computer Supported Cooperative Working,distributed computation and resource sharing.Well documented advantages[1]include:

1.Decentralised control.

2.Robustness.

3.Simple extendability.

4.Sharing of expertise.

5.Sharing of resources.

Decentralised control is,arguably,the most signi?cant feature of MAS that serves to distinguish such systems from distributed or parallel approaches to computation.

Kamal Ali Albashiri,Frans Coenen,and Paul Leng Decentralised control implies that individual agents,within a MAS,operate in an autonomous manner and are(in some sense)self deterministic.Robustness,in turn is a feature of the decentralised control,where the overall system continues to oper-ate even though a number of individual agents have“crashed”.Decentralised control also supports extendability in that additional functionality can be added simply by including further agents.The advantages of sharing expertise and resources are self evident.The advantages offered by MAS are particularly applicable to Knowledge Discovery in Data(KDD)where a considerable collection of tools and techniques are current.MAS also has some particular advantages to offer with respect to KDD, and particularly data mining,in the context of sharing resources and expertise.KDD is concerned with the extraction of hidden knowledge from data.Very often data rel-evant to one search is not located at a single site,it maybe widely-distributed and in many different forms.There is a clear advantage to be gained from an organisation that can locate,evaluate and consolidate data from these diverse sources.KDD has evolved to become a well established technology that has many commercial appli-cations.It encompasses sub-?elds such as classi?cation,clustering,and rule mining. Research work in these?elds continues to develop ideas,generate new algorithms and modify/extend existing algorithms.A diverse body of work therefore exists. KDD research groups and commercial enterprises,are prepared(at least to some extent)to share their expertise.In addition,many KDD research groups have made software freely available for download1.This all serves to promote and enhance the current“state of the art”in KDD.However,although the free availability of data mining software is of a considerable bene?t to the KDD community,it still require users to have some programming knowledge—this means that for many potential end users the use of such free software is not a viable option.One of the additional advantages offered by a MAS approach is that it would support greater end user access to data mining techniques.

A second advantages offered by MAS,in the context of data mining,is that of privacy and(to an extent)security.By its nature data mining is often applied to sensitive data.MAS allows data to be mined remotely.Similarly,with respect to data mining algorithms,MAS can make use of algorithms without necessitating their transfer to users,thus contributing to the preservation of intellectual property rights.

In this paper the authors propose the Extendible Multi-Agent Data mining Sys-tem(EMADS).The EMADS vision is that of an anarchic collection of persistent, autonomous(but cooperating)KDD agents operating across the Internet.Individual agents have different functionality;the system currently comprises data agents,user agents,task agents,mining agents and a number of“house-keeping”https://www.wendangku.net/doc/c03531045.html,ers of EMADS may be data providers,data mining algorithm contributors or miners of data.The provision of data and mining software is facilitated by a system of https://www.wendangku.net/doc/c03531045.html,ers wishing to obtain(say)classi?ers or collections of patterns,need have no knowledge of how any particular piece of data mining software works or the lo-cation of the data to be used.The operation of EMADS is illustrated in this paper

EMADS:AN EXTENDIBLE MULTI-AGENT DATA MINER

through the application of a collection of classi?er data mining agents to a number of standard“benchmark”data sets held by data agents.

The paper is organised as follows.A brief review of some related work on Multi-Agent Data Mining(MADM)is presented in Section2.The conceptual framework for EMADS is presented in Section3.The current implementation of EMADS, together with an overview of the wrapper principle is given in4.The operation of EMADS is illustrated in Section5with a classi?cation scenario.Some conclusions are presented in Section6.

2Previous Work

There are a number of reports in the literature of the application of Agent techniques to data mining.Some example systems are brie?y presented here.One of the earli-est references to MADM is Kargupta et al.[2]who describe a parallel data mining system(PADMA)that uses software agents for local data accessing and analysis, and a Web based interface for interactive data visualisation.PADMA has been used in medical applications.Gorodetsky et al.[3]correctly consider that the core prob-lem in MADM is not the data mining algorithms themselves(in many case these are well understood),but the most appropriate mechanisms to allow agents to collabo-rate.Gorodetsky et al.present a MADM system to achieve distributed data mining and,speci?cally,classi?cation.They describe a distributed data mining architecture and a set of protocols for a multi-agent software tool.Peng et al.[4]present an interesting comparison between single-agent and multi-agent text classi?cation in terms of a number of criteria including response time,quality of classi?cation,and economic/privacy considerations.Their results indicate,not unexpectedly,in favour of a multi-agent approach.

Agent technology has also been employed in meta-data mining,the combina-tion of results of individual mining agents.One example is meta classi?cation,also sometimes referred to as meta-learning,this is a technique for generating a global classi?er from N distributed data sources by?rst computing N base classi?ers which are then collated to build a single meta classi?er(see for example[14]).The meta-learning strategy offers a way to mine classi?ers from homogeneously distributed data.

Perhaps the most mature agent-based meta-learning systems are:JAM[5],BODHI [6],and Papyrus[7].In contrast to JAM and BODHI,Papyrus can not only move models from site to site,but can also move data when that strategy is desired. Papyrus is a specialised system which is designed for clustering while JAM and BODHI are designed for data classi?cation.Basically,these systems try to combine local knowledge to optimise a global objective.The major criticism of such systems is that it is not always possible to obtain an exact?nal result,i.e.the global knowl-edge model obtained may be different from the one that might have been obtained by applying the one model approach to the same data.

Kamal Ali Albashiri,Frans Coenen,and Paul Leng It should be noted that the domains of distributed and multi-agent data mining tend to overlap,with much discussion amongst authors as to what a MADM system is.In this paper the authors concur with Wooldridge’s[1]de?nition of what an agent is as itemised in Section1.

3The EMADS Conceptual Framework

Conceptually EMADS is a hybrid peer to peer agent based system comprising a collection of collaborating agents that exist in a set of containers.Agents may be created and contributed to EMADS by any EMADS user/contributor.One of these containers,the main container,holds a number of house keeping agents that have no direct connection with MADM,but provide various facilities to maintain the operation of EMADS.In particular the main container holds an Agent Manage-ment System(AMS)agent and a Directory Facilitator(DF)agent.The terminology used is taken from the JADE(Java Agent Development)[9]framework in which EMADS is implemented(JADE implementation details are discussed further in Sec-tion4).Brie?y the AMS agent is used to control the life cycles of other agents in the platform,and the DF agent provides an agent lookup service.Both the main con-tainer and the remaining containers can hold various MADM agents.Note that the EMADS main container is located on the EMADS host organisation site(currently The University of Liverpool in the UK),while the other containers may be held at any other sites world wide.

Other than the house keeping agents held in the main container EMADS cur-rently supports four categories of MADM agents:

https://www.wendangku.net/doc/c03531045.html,er Agents:User agents are the interface agents that connect users to EMADS.

User agents allow users to pose requests and receive responses to such requests.

Individual users create and launch their own EMADS user agents,which reside in the users’EMADS containers and are hosted at the users’https://www.wendangku.net/doc/c03531045.html,er agents interact with task agents(see below)in order to process data mining requests. 2.Task Agents:Task agents are speci?c temporary agents that are automatically

created by user agents to address speci?c data mining requests.Task agents are located at the user’s site and persist till the response to the associated requests is complete.A user can cause any number of task agents to be created.The nature of individual task agents depends on the nature of the requests,for example a clas-si?cation task agent will be launched to respond to a classi?cation request while (say)a meta Association Rule Mining task agent will be launched to respond to

a meta-ARM request.Individual task agents posses meta-knowledge about data

mining processes,which in turn de?ne the methodology/approach best suited to respond to a particular data mining request;this includes input format require-ments for speci?c data mining agents(see below).This meta-knowledge is used

EMADS:AN EXTENDIBLE MULTI-AGENT DATA MINER

in initiate and execute a required data mining process.Task agents are also re-sponsible for communication to/from data mining agents,and(if appropriate)the activation and synchronisation of data mining agents.To execute a data mining process a task agent typically seeks the services of a group of data mining and data agents(see below)to obtain the desired result and return it to the user agent.

3.Mining Agents:Mining agents are an implementation of a speci?c data min-

ing technique or algorithm.Mining agents contain the methods for initiating and carrying out a data mining activity and communicating results back to the ap-propriate task agent.Note that to release the full potential of EMADS mining agents,in either the same or different containers,typically collaborate to resolve some data mining task;although they are not obliged to so.Data mining agents are contributed by any EMADS developer,and reside in their owner’s EMADS container hosted at the owner’s site.

4.Data Agents:An agent,located at a local site,that holds meta-data about spec-

i?ed data sources held at the same site.The data may be a single data set,part of a data set or a number of data sets.Data agents are provided by EMADS data contributors.One of the advantages offered by data agents is that of privacy preservation.

A high level view of the EMADS conceptualisation showing the various cate-gories of agents and their interaction is given in Figure1.The?gure shows a me-diator host(main container)and three local hosts(local containers).The mediator host holds a AMS and a DF agent.One of the local hosts has a user and a task agent, while the other two hosts hold data and mining agents.

It should be noted that EMADS containers may contained both mining and data agents simultaneously as well as user agents.It should also be noted that data min-ing and data agents are persistent,i.e.they continue to exist inde?nitely and are not created for a speci?c data mining https://www.wendangku.net/doc/c03531045.html,munication between agents is facilitated by the EMADS network.

3.1EMADS End User Categories

EMADS has several different modes of operation according to the nature of the participant.Each mode of operation(participant)has a corresponding category of user agent.Broadly,the supported categories are as follows:

?EMADS Users:These are participants,with restricted access to EMADS,who may pose data mining requests.

?EMADS Data Contributors:These are participants,again with restricted ac-cess,who are prepared to make data available to be used by EMADS mining agents.

?EMADS Developers:Developers are EMADS participants,who have full access and may contribute data mining algorithms.

Kamal Ali Albashiri,Frans Coenen,and Paul Leng

Fig.1High level view of EMADS conceptual framework.

Note that in each case,before interaction with EMADS can commence,appropriate software needs to be downloaded and launched by the participant.Note also that any individual participant may be a user as well as a contributor and/or developer.

Conceptually the nature of EMADS data mining requests,that may be posted by EMADS users,is extensive.In the current implementation,the following types of generic request are supported:

?Find the”best”classi?er(to be used by the requester at some later date in off line mode)for a data set provided by the user.

?Find the”best”classi?er for the indicated data set(i.e.provided by some other EMADS participant).

?Find a set of Association Rules(ARs)contained within the data set(s)provided by the user.

?Find a set of Association Rules(ARs)contained within the indicated type of data set(s)(i.e.provided by other EMADS participants).

A“best”classi?er is de?ned as a classi?er that will produce the highest accuracy on a given test set(identi?ed by the mining agent)according to the detail of the request. To obtain the“best”classi?er EMADS will attempt to access and communicate with as many classi?er generator data mining agents as possible and select the best result.The classi?cation style of user request will be discussed further in Section5 to illustrate the operation of EMADS in more detail.

EMADS:AN EXTENDIBLE MULTI-AGENT DATA MINER

The Association Rule Mining(ARM)style of request is not discussed further in this paper.However,the idea here was that an agent framework could be used to im-plement a form of Meta-ARM where the results of the parallel application of ARM to a collection of data sets,with not necessarily the same schema but conforming to a global schema,are combined.Details of this process can be found in Albashiri et al.[8].

4The EMADS Implementation

EMADS is implemented using the JADE framework.JADE is FIPA(Foundation for Intelligent Physical Agents)[10]compliant middleware that enables develop-ment of peer to peer applications based on the agent paradigm.JADE de?nes an agent platform that comprises a set of containers,which may be distributed across a network as in the case of EMADS.A JADE platform includes a main container in which is held a number of mandatory agent services.These include the AMS and DF agents whose functionality has already been described in Section3.Recall that the AMS agent is used to control the lifecycles of other agents in the platform, while the DF agent provides a lookup service by means of which agents can?nd other agents.When a data mining or data agent is created,upon entry into the sys-tem,it announces itself to the DF agent after which it can be recognised and found by other agents.

Within JADE agents are identi?ed by name and communicate using the FIPA Agent Communication Language(ACL).More speci?cally,agents communicate by formulating and sending individual messages to each other and can have“conversa-tions”using interaction protocols that range from query request protocols to negoti-ation protocols.ACL message communication between agents within the same con-tainer uses event dispatching.Message communication between agents in the same JADE platform,but in different containers,is founded on RMI.Message communi-cation between agents in different platforms uses the IIOP(Internet Inter-ORB Pro-tocol).The latter is facilitated by a special Agent Communication Channel(ACC) agent also located in the JADE platform main containers.

Figure2gives an overview of the implementation of EMADS using JADE.The ?gure is divided into three parts:at the top are listed N user sites.In the middle is the JADE platform holding the main container and N other containers.At the bottom a sample collection of agents is included.The solid arrows indicates a“belongs to”(or“is held by”)relationship while the dotted arrows indicate a“communicates with”relationship.So the data agent at the bottom left belongs to container1which in turn belongs to User Site1;and communicates with the AMS agent and(in this example)a single mining agent.

The principal advantage of this JADE architecture is that it does not overload a single host machine,but distributes the processing load among multiple machines. The results obtained can be correlated with one another in order to achieve compu-tationally ef?cient analysis at a distributed global level.

Kamal Ali Albashiri,Frans Coenen,and Paul Leng

Fig.2EMADS Architecture as Implemented in Jade

4.1EMADS Wrappers

One of the principal objectives of EMADS is to provide an easily extendible frame-work that could easily accept new data sources and new data mining techniques.In general,extendibility can be de?ned as the ease with which software can be modi-?ed to adapt to new requirements or changes in existing requirements.Adding a new data source or data mining techniques should be as easy as adding new agents to the system.The desired extendability is achieved by a system of wrappers.EMADS wrappers are used to“wrap”up data mining artifacts so that they become EMADS agents and can communicate with other EMADS agents.As such EMADS wrappers can be viewed as agents in their own right that are subsumed once that have been integrated with data or tools to become data mining agents.The wrappers essen-tially provide an application interface to EMADS that has to be implemented by the end user,although this has been designed to be a fairly trivial operation.Two broad categories of wrapper have been de?ned:(i)data wrappers and(ii)tool wrappers. Each is described in further detail in the following two sections.

4.1.1Data Wrappers

Data wrappers are used to“wrap”a data source and consequently create a data agent.Broadly a data wrapper holds the location(?le path)of a data source,so that it can be accessed by other agents;and meta information about the data.To assist

EMADS:AN EXTENDIBLE MULTI-AGENT DATA MINER

end users in the application of data wrappers a data wrapper GUI is available.Once created,the data agent announces itself to the DF agent as consequence of which it becomes available to all EMADS users.

4.1.2Tool Wrappers

Tool wrappers are used to“wrap”up data mining software systems and thus create a mining agent.Generally the software systems will be data mining tools of various kinds(classi?ers,clusters,association rule miners,etc.)although they could also be(say)data normalisation/discretisation or visualisation tools.It is intended that EMADS will incorporate a substantial number of different tool wrappers each de-?ned by the nature of the desired I/O which in turn will be informed by the nature of the generic data mining tasks that it us desirable for EMADS to be able to perform. Currently the research team have implemented two tool wrappers:

1.The binary valued data,single label,classi?er generator.

2.The meta AR generator.

Many more categories of tool wrapper can be envisaged.Mining tool wrappers are more complex than data wrappers because of the different kinds of information that needs to be exchanged.For example in the case of a“binary valued,single label,classi?er generator”wrapper the input is a binary valued data set together with meta information about the number of classes and a number slots to allow for the(optional)inclusion of threshold values.The output is then a classi?er expressed as a set of Classi?cation Rules(CRs).As with data agents,once created,the data mining agent announce themselves to the DF agent after which they will becomes available for use to EMADS users.

5EMADS Operation:Classi?er Generation

In this section the operation of EMADS is illustrated in the context of a classi-?er generation task;however much of the discussion is equally applicable to other generic data mining tasks such as clustering and ARM.The scenario is that of an end user who wishes to obtain a”best”classi?er founded on a given,pre-labelled, data set;which can then be applied to further unlabelled data.The assumption is that the given data set is binary valued and that the user requires a single-label,as opposed to a multi-labelled,classi?er.The request is made using the individual’s user agent which in turn will spawn an appropriate task agent.

For this scenario the task agent identi?es mining agents that hold single labelled classi?er generators that take binary valued data as input.Each of these mining agents is then accessed and a classi?er,together with an accuracy estimate,re-quested.The task agent then selects the classi?er with the best accuracy and returns this to the user agent.

Kamal Ali Albashiri,Frans Coenen,and Paul Leng The data mining agent wrapper in this case provides the interface that allows input for:(i)the data;and(ii)the number of class attributes(a value that the min-ing agent cannot currently deduce for itself)while the user agent interface allows input for threshold values(such as support and con?dence values).The output is a classi?er together with an accuracy measure.To obtain the accuracy measures the classi?er generator(data mining agent)builds the classi?er using the?rst half of the input data as the“training”set and the second half of the data as the“test”set.An alternative approach might have been to use Ten Cross Validation(TCV)to identify the best accuracy.

From the literature there are many reported techniques available for generating classi?ers.For the scenario the authors used implementations of eight different al-gorithms3:

Fig.3Classi?cation Task Sequence Diagram.

1.FOIL(First Order Inductive Learner)[11]the well established inductive learning

algorithm for generating Classi?cation Association Rules(CARs).

2.TFPC(Total From Partial Classi?cation)CAR generator[12]founded on the P-

and T-tree set enumeration tree data structures.

3.PRM(Predictive Rule Mining)[15]an extension of FOIL.

4.CPAR(Classi?cation based on Predictive Association Rules)[15]a further de-

velopment from FOIL and PRM.

EMADS:AN EXTENDIBLE MULTI-AGENT DATA MINER

5.IGDT(Information Gain Decision Tree)classi?er,an implementation of the

well established decision tree based classi?er using most information gain as the“splitting criteria”.

6.RDT(Random Decision Tree)classi?er,a decision tree based classi?er that uses

most frequent current attribute as the“splitting criteria”(so not really random).

7.CMAR(Classi?cation based on Multiple Association Rules)is a Classi?cation

Association Rule Mining(CARM)algorithm[16].

8.CBA(Classi?cation Based on Associations)is a CARM algorithm[17]. These were placed within an appropriately de?ned tool wrapper to produce eight (single label binary data classi?er generator)data mining agents.This was a trivial operation indicating the versatility of the wrapper concept.

Thus each mining agent’s basic function is to generate a classi?cation model us-ing its own classi?er and provide this to the task agent.The task agent then evaluates all the classi?er models and chooses the most accurate model to be returned to the user agent.The negotiation process amongst the agents is represented by the se-quence diagram given in Figure3(the?gure assumes that an appropriate data agent has ready been created).In the?gure includes N classi?cation agents.The sequence of events commences with a user agent which spawns a(classi?cation)task agent, which in turn announces itself to the DF agent.The DF agent returns a list of classi-?er data mining agents that can potentially be used to generate the desired classi?er. The task agent then contacts these data mining agents who each generate a classi?er and return statistical information regarding the accuracy of their classi?er.The task agent selects the data mining agent that has produced the best accuracy and requests the associated classi?er,this is then passed back to the user agent.

Table1Classi?cation Results

Data Set Accuracy

connect4.D129.N67557.C379.76

adult.D97.N48842.C286.05

letRecog.D106.N20000.C2691.79

anneal.D73.N898.C698.44

breast.D20.N699.C293.98

congres.D34.N435.C2100

cylBands.D124.N540.C297.78

dematology.D49.N366.C696.17

heart.D52.N303.C596.02

auto.D137.N205.C776.47

penDigits.D89.N10992.C1099.18

soybean-large.D118.N683.C1998.83

waveform.D101.N5000.C396.81

Kamal Ali Albashiri,Frans Coenen,and Paul Leng tion were taken from the UCI machine learning data repository[18].To simplify the scenario these data sets were preprocessed so that they were discretized/normalized into a binary form4.It should be noted here that the research team is currently implementing a normalisation/discretisation agent.

The results from a sequence of user requests,using different data sets,are pre-sented in Table1.Each row in the table represents a particular request and gives the name of the data set,the selected best algorithm,the best accuracy and the total EMADS execution time from creation of the initial task agent to the?nal classi?er being returned to the user agent.The naming convention used in the Table is that: D equals the number of attributes(after discretisation/normalisation),N the number of records and C the number of classes(although EMADS has no requirement for the adoption of this convention).

The results demonstrate?rstly that EMADS works(at least in the context of the current scenario).Secondly that operation of EMADS is not signi?cantly hindered by agent communication overheads,although this has some effect.The results also reinforce the often observed phenomena that there is no single best classi?er gener-ator suited to all kinds of data set.

6Conclusions and Future Work

This paper describes EMADS,a multi-agent framework for data mining.The prin-cipal advantages offered are that of experience and resource sharing,?exibility and extendibility,and(to an extent)protection of privacy and intellectual property rights. The paper presents the EMADS vision,the associated conceptualisation and the JADE implementation.Of note are the way that wrappers are used incorporate exist-ing software into EMADS.Experience indicates that,given an appropriate wrapper, existing data mining software can be very easily packaged to become an EMADS data mining agent.The EMADS operation is illustrated using a classi?cation sce-nario.

A good foundation has been established for both data mining research and gen-uine application based data mining.The current functionality of EMADS is limited to classi?cation and Meta-ARM.The research team is at present working towards increasing the diversity of mining tasks that EMADS can address.There are many directions in which the work can(and is being)taken forward.One interesting di-rection is to build on the wealth of distributed data mining research that is currently available and progress this in an MAS context.The research team are also enhanc-ing the system’s robustness so as to make it publicly available.It is hoped that once the system is live other interested data mining practitioners will be prepared to con-tribute algorithms and data.

EMADS:AN EXTENDIBLE MULTI-AGENT DATA MINER

References

1.Wooldridge,M.(2003).An Introduction to Multi-Agent Systems.John Wiley and Sons(Chich-ester,England).

2.Kargupta,H.,Hamzaoglu,I.and Stafford B.(1997).Scalable,Distributed Data Mining Using an Agent Based Architecture.Proceedings of Knowledge Discovery and Data Mining,AAAI Press,211-214.

3.Gorodetsky,V.,Karsaeyv,O.,Samoilov,V.(2003).Multi-agent technology for distributed data mining and classi?cation.Proc.Int.Conf.on Intelligent Agent Technology(IAT2003), IEEE/WIC,pp438-441.

4.Peng,S.,Mukhopadhyay,S.,Raje,R.,Palakal,M.and Mostafa,J.(2001).A Comparison Be-tween Single-agent and Multi-agent Classi?cation of Documents.Proc.15th International Paral-lel and Distributed Processing Symposium,pp935-944.

5.Stolfo,S.,Prodromidis,A.L.,Tselepis,S.and Lee,W.(1997).JAM:Java Agents for Meta-Learning over Distributed Databases.Proceedings of the International Conference on Knowl-edge Discovery and Data Mining,pp.74-81.

6.Kargupta,H.,Byung-Hoon,et al.(1999).Collective Data Mining:A New Perspective To-ward Distributed Data Mining.Advances in Distributed and Parallel Knowledge Discovery, MIT/AAAI Press.

7.Bailey,S.,Grossman,R.,Sivakumar,H.and Turinsky,A.(1999).Papyrus:a system for data mining over local and wide area clusters and super-clusters.In Proc.Conference on Supercom-puting,page63.ACM Press.

8.Albashiri,K.A.,Coenen,F.P.,Sanderson,R.and Leng.P.(2007).Frequent Set Meta Mining: Towards Multi-Agent Data Mining.In Bramer,M.,Coenen,F.P.and Petridis,M.(Eds.),Research and Development in Intelligent Systems XXIII.,Springer,London,(proc.AI’2007),pp139-151.

9.Bellifemine,F.Poggi,A.and Rimassi,G.(1999).JADE:A FIPA-Compliant agent frame-work.Proc.Practical Applications of Intelligent Agents and Multi-Agents,pg97-108(See http://sharon.cselt.it/projects/jade for latest information).

10.Foundation for Intelligent Physical Agents,FIPA2002Speci?cation.Geneva,Switzer-land.(See http://www.?https://www.wendangku.net/doc/c03531045.html,/speci?cations/index.html).

11.Quinlan,J.R.and Cameron-Jones,R.M.(1993).FOIL:A Midterm Report.Proc.ECML, Vienna,Austria,pp3-20.

12.Coenen,F.,Leng,P.and Zhang,L.(2005).Threshold Tuning for Improved Classi?cation Association Rule Mining.Proceeding PAKDD2005,LNAI3158,Springer,pp216-225.

13.Schollmeier,R.(2001).A De?nition of Peer-to-Peer Networking for the Classi?cation of Peer-to-Peer Architectures and Applications.First International Conference on Peer-to-Peer Comput-ing(P2P01)IEEE.

14.Prodromides,A.,Chan,P.and Stolfo,S.(2000).Meta-Learning in Distributed Data Mining Systems:Issues and Approaches.In Kargupta,H.and Chan,P.(Eds),Advances in Distributed and Parallel Knowledge Discovery.AAAI Press/The MIT Press,pp81-114.

15.Yin,X.and Han,J.(2003).CPAR:Classi?cation based on Predictive Association Rules.Proc. SIAM Int.Conf.on Data Mining(SDM’03),San Fransisco,CA,pp.331-335.

16.Li W.,Han,J.and Pei,J.(2001).CMAR:Accurate and Ef?cient Classi?cation Based on Multiple Class-Association Rules.Proc ICDM2001,pp369-376.

17.Liu,B.Hsu,W.and Ma,Y(1998).Integrating Classi?cation and Assocoiation Rule Mining. Proceedings KDD-98,New York,27-31August.AAAI.pp80-86.

18.Blake,C.L.and Merz,C.J.(1998).UCI Repository of machine learning databases htt p: //https://www.wendangku.net/doc/c03531045.html,/?mlearn/MLRepository.html,Irvine,CA:University of California,Depart-ment of Information and Computer Science.

制作vcd视频文件的方法

制作vcd视频文件的方法 描述: 如果想把avi,mepeg,wmv,rm等影音文件,做成普通的vcd碟片,就需要把这些影音文件转换成dat格式的文件,再用这个dat格式的文件替换vcd碟片中Mpegav文件夹(每一个VCD光盘里面都必定有的文件夹)中的文件。 下面介绍我做VCD的全过程: 第一步:先将你想制作的VCD原文件(它们是avi,wmv,rm等影音文件)转换成DA T 格式的文件,保存在你的电脑硬盘里面。我用的是软件是WinA VI Video Converter V7.7汉化版,感觉这个软件比较好用,你安装后照着提示完成文件的格式转换。注:这个软件是把avi,wmv,rm等文件转换成MPG格式,直接转换成DA T格式的软件不好找,再说这个软件转换成MPG格式的速度很快(600M的WMV格式转换成MPG的只用了30分钟)。当你的文件变成MPG格式后,直接把文件的后缀名改为DA T格式就完成了转换(注:其他格式不能直接该后缀名变成DA T格式,只能是MPG格式才可以直接改)。第二步:制作VCD,以下介绍两种方法。 方法一:制作视频光盘: 方法:直接将你已经制作好的DA T文件做成VCD。在你点击制作光盘软件时,选择制作VCD视频光盘,然后添加你制作好的DA T文件,点击下一步,直到光盘制作完,这样你就完成了VCD的制作。说明:此方法是最简单,它在制作视频数据时将自动升成很多文件(包括Cdda,Cdi,Karaoke,Mpegav,Vcd文件夹)。注意,上面的DA T 文件的文件名必须依次是Avseq01.dat,Avseq02.dat,Avseq03.dat……,不能用其他的文件名,否则将不能播放。 方法二:制作数据光盘: 方法:任意找一张VCD光盘,把它里面的所有文件都拷贝到你的电脑硬盘上(必须具备Cdda,Cdi,Karaoke,Mpegav,Vcd文件夹),放在一个文件夹里面(比如叫VCD文件夹里面),然后将Mpegav文件夹里面内容删除,将你制作好的DAT文件放在Mpegav文件夹里面,将这个DA T文件的文件名改为Avseq01.dat,如果有多个文件,将其他的分别命名为Avseq02.dat,Avseq03.dat……放在Mpegav文件夹,它们将依次播放。注意:这里的DA T文件必须依次为Avseq01.dat,Avseq02.dat,Avseq03.dat……,不能用其他的文件名,否则将不能播放。然后将Cdda,Cdi,Karaoke,Mpegav,Vcd 这些文件夹制作成数据光盘(注意:不能把这几个文件夹放在一个大的文件夹里面,否则将不能播放)。VCD制作完成。 第三步:测试阶段: 在刻录以前可以用虚拟光驱测试,把D:\vcd文件夹的所有文件和文件夹用虚拟光驱压缩成一个虚拟光驱文件,然后把这个文件装入虚拟光驱中,用超级解霸播放指定盘,看看效果如何。 另外也可以做一个dat格式的片头文件,这个片头文件EO Video软件就可以做成,具体的过程是:用Photoshop等图像处理软件把有关的片头信息做成Jpeg格式的图片,然后EO Video把这个图片按照上面的方法转换成mpeg的影音文件,把转换后的片头文件拷贝到d :\vcd\mpegav文件中,改名为Avseq01.dat,其他的文件依次命名为Avseq02,Avseq03.......这样属于个人的一张音像制品就做成了。

matlab建立dat文件简介

% Step 1: Read image Read in RGB = imread('pic4.jpg'); RGB=imresize(RGB,[1029,1029]); % Step 2: Convert image from rgb to gray GRAY = rgb2gray(RGB); % Step 3A:After eliminating the noise s=fftshift(fft2(GRAY)); [a,b]=size(GRAY); n=3;%Here take the order n comparison,n=1,2,3... d0=30; %Here d0 is the cutoff frequency,d0=10,20,30,... n1=fix(a/2); n2=fix(b/2); for i=1:a for j=1:b d=sqrt((i-n1)^2+(j-n2)^2); h=1/(1+0.414*(d/d0)^(2*n)); s(i,j)=h*s(i,j); end end s=uint8(real(ifft2(ifftshift(s)))); subplot(121),imshow(GRAY); title('Grayscale'); subplot(122),imshow(s); title('Image eliminated noise'); % Step 3B: Threshold the image Convert the image to black and white in order % to prepare for boundary tracing using bwboundaries. figure,imhist(s); title('Histogram'); [CStad,xs]=imhist(s); hold on; plot(xs,CStad,'r'); % save('Cstad.dat', 'CStad'); % close all; %Step3C: Two values, determined in accordance with 70/255 threshold, %dividing the target and background Inew=im2bw(s,70/255); figure;imshow(Inew); title('Image after thresholding'); fid = fopen('CStad.dat','wt'); fprintf(fid,'%g\n', CStad); %\n 换行 fclose(fid);

dat文件制作教程详解

Dat文件就是把DOS命令写在一个文本文件里面,然后保存的时候保存成"所有文件", 文件名是名字.bat 就可以了。 不过BAT文件有很多特殊命令... 批处理制作教程 批处理文件是无格式的文本文件,它包含一条或多条命令。它的文件扩展名 为 .bat 或 .cmd。在命令提示下键入批处理文件的名称,或者双击该批处理文件,系统就会调用Cmd.exe按照该文件中各个命令出现的顺序来逐个运行它们。 在无盘运用及Hack入侵过程中,经常都会用到。 一.简单批处理内部命令简介 1.Echo 命令 打开回显或关闭请求回显功能,或显示消息。如果没有任何参数,echo 命令将显示当前回显设置。 语法 echo [{on off}] [message] Sample:@echo off / echo hello world 在实际应用中我们会把这条命令和重定向符号(也称为管道符号,一般用> >> )结合来实现输入一些命令到特定格式的文件中.这将在以后的例子中体现出来。

2.@ 命令 表示不显示@后面的命令,在入侵过程中(例如使用批处理来格式化敌人的硬盘)自然不能让对方看到你使用的命令啦。 Sample:@echo off @echo Now initializing the program,please wait a minite... @format X: /q/u/autoset (format 这个命令是不可以使用/y这个参数的,可喜的是微软留了个autoset这个参数给我们,效果和/y是一样的。) 3.Goto 命令 指定跳转到标签,找到标签后,程序将处理从下一行开始的命令。 语法:goto label (label是参数,指定所要转向的批处理程序中的行。)Sample: if {%1}=={} goto noparms if {%2}=={} goto noparms(如果这里的if、%1、%2你不明白的话,先跳过去,后面会有详细的解释。) @Rem check parameters if null show usage :noparms echo Usage: monitor.bat ServerIP PortNumber goto end 标签的名字可以随便起,但是最好是有意义的字母啦,字母前加个:用来表示这个字母是标签,goto命令就是根据这个:来寻找下一步跳到到那里。最好有一些说明这样你别人看起来才会理解你的意图啊。

CCS中的.dat文件

详解CCS中的.dat文件 CCS支持的.dat文件的格式为: 定数数据格式起始地址页类型数据块大小 1651 其后是文件内容,每行表示一个数据。 定数固定为“1651”,数据格式可以选择“1”(十六进制整型)、“2”(十进制整型)、“3”(十进制长整型)、“4”(十进制浮点型)。起始地址为存储的地址,页类型和标示为程序或者数据。比如一个.dat文件: 1651 1 800 1 10 0x0000 0x0000 0x0000 0x0000 0x0000 0x0000 0x0000 0x0000 0x0000 0x0000 0x0000

0x0000 0x0000 制作.dat 文件的方法也很简单,可以用VC++或者MATLAB来实现。比如hellodsp的网友lwxing提供的使用matlab创建.dat文件的一个实例: matlab向dsp传递.dat文件 x=2*sin(2*pi*100*m*dt); for m=1:200; if x(m)>=0 y(m)=x(m); else y(m)=4+x(m); end; end; y=y*16384; fid=fopen('input.dat','w');%打开文件,'w'是将此文件定义为可 写的,fid是此文件的整数标示 fprintf(fid,'1651 1 0 1 0\n');%输出文件头,文件头必须是dsp 所能识别的,就如此句程序所设定的 fprintf(fid,'0x%x\n',round(y));%输出y数组,并写到与fid标示符相同的文件,即yinput.dat文件里。round是取y值的最近的数,即如果是1.2,就取1,如果1.6,就取2. fclose(fid); %关闭fid标示符的文件。 fid=fopen('input.dat','w');%打开文件,属性设置为写

maskout方法的使用以及maskout文件制作

maskout文件制作[复制链接] 电梯直达 本帖最后由传说中的谁于 2011-7-9 22:45 编辑 用GrADS作图时,常会遇到要把图形控制字在一定地理边界内的情况(例如,某个省或某个区域);用站点资料绘制等值线或填色图时,也通常会把地理边界以外的信息屏蔽掉。我在另一个帖子《GrADS中Basemap方法的应用》中提到了用basemap方法来解决这个问题,但basemap有其缺陷,就是你用basemap屏蔽掉边界以外的信息以后,再画下一个变量的时候边界以外的区域同样不能显示。现在介绍另一种方法:maskout。maskout是将一个变量在边界内的部分保留原来的值,而在边界外则一概赋给一概新的值:0,因此最后的结果就是只在边界内有值,从而达到边界控制的效果。下面详细介绍maskout方法的使用。 首先,要使用maskout就得有maskout文件,这里介绍的使用MeteoInfo 来制作maskout文件。下载MeteoInfo,安装后启动MeteoInfo。 第一步,添加图层,在弹出对话框中选择“b ou2_4p.shp”→“打地图文件”,将地图文件加载到程序界面中。可以通过缩放来调节地图到你想要的大小。 第二步,获取底图ID。在工具栏中选择“图元属性”,这时候鼠标指针会变

成一个“ i ”的形状。单击你要选择的区域,比如广东省,会弹出一个图元属性的窗口,index后面的数字就是你所选择区域的图元ID,如下图右中的“897”。 第三步,制作maskout文件。选择菜单栏中的“工具”→“输出地图数据”,弹出输出数据的对话框。选择你要的图元,设置输出格式为GrADS Maskout File,然后点“输出”。

DAT 31-.《纸质档案数字化规范》之欧阳家百创编

目录 欧阳家百(2021.03.07)前言4 引言6 纸质档案数字化规范7 1 范围7 2 规范性引用文件7 3 术语和定义7 4 总则8 5.组织与管理9 5.1 机构及人员9 5.2 基础设施10 5.3 工作方案10 5.4 管理制度11 5.5 工作流程控制12 5.6 工作文件管理12 5.7 档案数字化外包13 6 档案出库13 7 数字化前处理14 7.1 确定扫描页14 7.2 编制页号14

7.3 目录数据准备15 7.4 拆除装订15 7.5 技术修复15 8 目录数据库建立15 9 档案扫描16 9.1 基本要求16 9.2 扫描设备17 9.3 扫描色彩模式17 9.4 扫描分辨率18 9.5 存储格式18 9.6 图像命名19 10 图像处理19 10.1 图像拼接19 10.2 旋转及纠偏19 10.3 裁边20 10.4 去污20 10.5 图像质量检查20 11 数据挂接20 12 数字化成果验收与移交21 12.1 验收方式21 12.2 验收内容21 12.3 验收指标22 12.4 验收结论22

12.5 移交23 13 档案归还入库23 前言 本标准按照GB/T 1.1-2009给出的规则起草。 本标准替代DA/T 31-2005《纸质档案数字化技术规范》。 本标准与DA/T 31-2005相比,主要技术变化如下: ——标题进行了修改; ——增强组织与管理部分的内容,完善数字化工作中管理相关要求; ——增强数字化前处理部分的内容,包括对实体档案保护和档案规范化管理方面的要求; ——增加数字化过程中元数据采集的要求; ——修改了档案扫描部分参数要求; ——修改了图像处理部分内容,更加强调保持档案原貌的要求; ——细化了数字化成果验收的内容; ——删除原标准数据备份和数字化成果管理相关内容。 本标准由国家档案局提出并归口。 本标准起草单位:国家档案局档案科学技术研究所、国家档案局信息管理中心、国家档案局技术部。 本标准主要起草人:王良城、马淑桂、郝晨辉、程春雨、杜琳琳、蔡伟、宋涌、王大众、田军、曹燕、李华峰。

视频剪辑制作教程

电影魔方使用教程 电影魔方是一款品质优秀、功能强大、操作简单的多媒体数字视频编辑软件,非常的专业。 主要功能: 界面:自由组合的窗口模式,使用方便的项目及素材管理器,输入、输出双监视窗口;四个编辑轨道的时间轴。 预览:时间码准确定位;双监视窗口可同时预览或操作;在滑块拖动中实时预览;多级变速播放和逆向播放。 字幕:独立的字幕编辑器,快捷的字幕合成方式;丰富的图形绘制功能;16种字幕动态效果。 编辑:直观灵活的素材拖放操作;实用高效的编辑工具箱;支持音频视频同步调整;精确到每帧的编辑精度。 转场:多种精彩转场特效;轻松调整转场长度;任意设定转场参数;提供音频转场效果。 输出:可输出MPEG-1、MPEG-2、VCD、SVCD等视频文件。 支持格式视频:mpg、mpeg、mpv、dat、vob、ts、avi。 音频:mp1、mp2、mp3、AC3、wav。 图像:bmp、jpg、jpeg、gif、ico、wmf. 创作出各种不同用途的多媒体影片,并且可以刻录成VCD或DVD光盘。 下面一起来学习一下该软件。 一、软件下载和安装 通过https://www.wendangku.net/doc/c03531045.html,/download.asp 软件下载下来是一个压缩文件,解压后开始安装,软件的安装过程十分的简单,只要一路点击“下一步”即可安装完成。如图1所示。

图1 安装完成以后,首次运行时需要输入注册码,用户可到官方网站上免费申请。通过以下网址进行免费申请(https://www.wendangku.net/doc/c03531045.html,/download.asp)。如图2所示。 图2

只要填写了正确的e-mail稍后就能收到注册码。输入注册码,然后点击“确定”按钮就进入了软件主界面。如图3、4所示。 图3 图4 二、制作视频光盘

放样坐标(dat文件)快速导入GPS手簿(DC文件)的方法

放样坐标(dat文件)快速导入GPS手簿(DC文件)的方法 用GPS全球卫星定位系统进行工程坐标放样之前如何把dwg图纸里的dat坐标数据快速自动地导入至GPS手簿中先用相关软件打开后缀名为dwg的图形文件,用菜单里的:“计算与应用”→“指定点生成数据文件”,把待放样点坐标一个个用鼠标手工捕捉、导出、保存到桌面上。导出完毕,桌面上会自动生成一个dat文件,这个文件就是原始的待放样点规划坐标数据文件。 用手工捕捉导出待放样点规划坐标时要仔细小心,不得有误。如果这一步不出差错,以后出错的概率就几乎为零。 为防止GPS手簿不认识中文,把dat文件重命名,比如:“fy0103”,即1月3日放样之意,并同时把后缀名改为“txt”。 打开这个“fy0103.txt”的文件,把第一行表示点号总数的数字删除,使第二行的坐标数据从第一行开始排列起,保存,退出。 按照平时的使用习惯,将电脑和GPS手簿连接起来,开启数据传输软件,“发送”→“添加”。“桌面”→“文件类型:用逗号分界的坐标文件(*.txt)”→“fy0103.txt”→“选择”。“全部传输”,传输结束。关闭数据传输软件。 这样,这次待放样点的规划坐标数据统统传进了GPS手簿内。到这一步,已脱离电脑,下面可以在手簿上操作了。 不信,可以用Microsoft Activesync 软件查看手簿里的内容,确实多了一个新文件:“fy0103. txt”。 如果不习惯使用Activesync 软件,也没关系。因为万一把手簿内的文件误删,它是不进入回收站,直接删除的,很危险。所以,不使用Microsoft Activesync 软件也许并不是坏事。 下面,把手簿里那个“fy0103.txt”文件再导入到要放样的GPS网格系统内。第一次使用要注意格式“域”的设置。 设置好了,以后就不用再设置。这对初次尝试者显得尤为重要,北向是3开头,东坐标是5开头,不要搞反,dat文件里的逗号每行一共有4个,不要多了或少了。 1. 选好当前所使用的工地网格系统。 2. 文件,导入导出,导入固定格式文件, 3. 文件格式:逗号定界的文件(*.CSV,*.TXT) 4. 从名称:fy0103.txt 5. 点名:域1;点代码:域2;北向:域4;东向:域3;高程:域5; 6. 接受 这样,电脑里待放样点的规划坐标(*.dat数据文件)已经完全导入到GPS手簿里相应的*. dc文件内。 可以在手簿“文件”→“检查当前任务”中查看到坐标及高程。或者干脆把dc文件再按平常步骤导出至电脑上看看,是不是一样的? 这个办法一直以来被认为是有理论依据的,理论上完全可以实现,遗憾的是,在本文公布之前,并无可操作的详细方法步骤公布面世。诀窍一经点破就毫无悬念,至此,结束了人工输入大量坐标数据的原始作业方法,采用机器传输能极大提高效率,出错的概率几乎为零,心理压力减少,作业环节消减,对于大量放样任务的圆满完成和老图控制点坐标的检核(三维逆向建模),现实意义尤其重大,仿佛时光倒流,情形再现,也为工程建设的程序法制化、决策科学化,提供了必要的监测取证保全手段,有效遏制了测绘行业“周正龙现象”的出现,有据可查,有人可究,避免重大损失事故。

MATLAB中将数据保存为TXT或DAT格式四种方案模板

matlab中将数据保存为txt或dat格式四种方案 ——胡总结网上各种资源, 列出以下的四种方法( 以txt为例) 。 第一种方法: save( 最简单基本的) 具体的命令是: 用save*.txt-asciix x为变量 *.txt为文件名,该文件存储于当前工作目录下, 再打开就能够打开后,数据有可能是以指数形式保存的. 例子: a=[17241815;23571416;46132022;;11182529]; saveafile.txt-asciia; %保存文本文档的文件名 afile.txt打开之后, 是这样的: 1.7000000e+001 2.4000000e+0011.0000000e+0008.0000000e+0001.5 000000e+001 2.3000000e+0015.0000000e+0007.0000000e+0001.4000000e+0011.6 000000e+001 4.0000000e+0006.0000000e+0001.3000000e+0012.0000000e+0012. 000e+001 1.0000000e+0011. 000e+0011.9000000e+0012.1000000e+0013.0000000e+000

1.1000000e+0011.8000000e+001 2.5000000e+0012.0000000e+0009.0 000000e+000 第二种方法: dlmwrite dlmwrite('a.txt',a,'precision','%10.0f') 或者是dlmwrite('a.txt',a,'delimiter','\t') 对于只有一行或者一列的数据, 很适用, 可是多行的, 就乱了 网上有很多这一类似的问题, 可是都不是很理想 第三种方法: fopen+fprintf( 最常见) 下面主要介绍这种方法, 由以下的前两种情况最终导出第三种情况能够完美的解决以上问题。以上面的例子为例: 第一种情况: >>a=[17241815;23571416;46132022;;11182529]; >>fid=fopen('b.txt','wt'); fprintf(fid,'%g\n',a);\n换行 fclose(fid); 然后用写字板打开b.txt, 内容如下: 为列向量 17 23 4

用Helix Producer Plus制作RMVB文件教程

用Helix Producer Plus制作RMVB文件教程 前期工作: 思想准备:有看动漫/电影的朋友一般都知道,网络上有很多源片,但格式AVI格式的,因为AVI画面清晰,音效好,有DVDrip的效果,所有广受欢迎,适合收藏,但他有一个严重缺点,就是体积太大,普通的电影都有八九百兆以上,刻盘都不行,我们又知道RMVB体积小,质量好~所以,这软件,就有他的价值所在,他的功能就是把大体积的AVI 影片压缩成RMVB,但又不减质量,而且他说支持的格式非常广~ avi ,,,wma..mov..dat,,mpeg..mpg..等等,反正市面上所见到的,都是他的范围,,,还有,修改影片码率,去片头,去片尾,合并影片,修改影片的分辨率(画面大小),只要你的想法不过分,他一般都能完成~~ 使用软件:Helix Producer 9.0.1.250 软件简介:Elix是业界第一个跨平台,跨流媒体、高性能的流媒体服务器。配合Helix先进的功能,Realnetworks 还推出了第10代的流媒体压缩软件Helix Producer。Realnetworks全新改写代码的图形化专业流媒体文件制作工具。利用它,你可以轻松地实现RealAudio8、RealAudio9文件格式到实时文件的转换,转换后的文件更加适合实时观看、在线广播和下载。Helix Producer基于Realnetworks完全改写的核心代码,提供简单、高效的界面操作。 好了,现在进入正题~~ 下载,安装,结束后,桌面上会出现这么2个图标~~ 我们先来双击第一个图标,打开,这是打开的LOGO~

程序主界面如下,选择要压缩的源文件~~ 选择要压缩的文件

清华天河图库TH制作教程(第二版,详细)

清华天河图库TH制作教程(第二版,详细)清华天河图库TH制作教程 (第二版) 清华天河图库TH制作教程(第二版) 前段时间编写了《清华天河图库TH制作教程(第一版)》,受到了大家的欢迎,但是由于时间仓促,很多地方没有交代清楚,建立稍微复杂一些的图库可能还是 很吃力,为了使得越来越多的朋友会制作图库,制作出更好的图库,本人特编写 《清华天河图库TH制作教程(第二版)》,将各个环节细化充分,并将新摸索的经 验编入,希望朋友们学会后将图库共享,本人也能从中受益。 制作TH图库是个很繁琐的过程,很多功能还不太方便,如果您对制作过程一点不了解,请先参阅本教程附录部分《清华天河图库TH制作教程(第一版)》内容,里边的例题简单一些,如果您已经了解了初步的制作过程,请直接参阅《清 华天河图库TH制作教程(第二版)》内容。 -1- 清华天河图库TH制作教程(第二版) 目录 一、软件 二、热轧普通槽钢入库实例 1、绘制图形

2、绘制辅助线 3、尺寸标注 4、生成参数化图形 5.定义变量及表达式 6、参数图驱动 7、设置/删除其辅助对象 8、保存参数化结果 9、图形入库 10、尺寸系列增加 10.1、尺寸系列准备 10.2、尺寸系列入库 三、结束语 附录:清华天河图库TH制作教程(第一版) -2- 清华天河图库TH制作教程(第二版) 1、PCCAD任何版本 2、EXCEL 本人使用的是PCCAD2002版和EXCEL2000版,当然,用高版本制作也没有任何问题,过程是一样的。 下面以“热轧普通槽钢(GB/T707—1988)”为实例详细讲解制作TH图库的过程。 首先,您的手上要有GB/T707—1988热轧普通槽钢详细尺寸系列标准,查机械手册或者网上下载都可以。 按照“GB/T707—1988热轧普通槽钢”标准中其中一种槽钢的尺寸绘制图形,

用MATLAB生成DAT文件

用MATLAB 生成DAT 文件 f1=10;f2=100;fs=400;N=400; x = linspace(0,N/fs,N); y1 = sin(2*pi*f1*x); y2 = sin(2*pi*f2*x); y=y1+y2; figure plot(x,y) %创建文件sine.dat,可写入 fid=fopen('sine.dat','w'); %将文件头写入文件,将生成的y 信号写入到文件,格式四位小数 fprintf(fid,'1651 4 0 1 0\n'); fprintf(fid,'%.4f\n',y); %将文件头写入文件,将生成的y 信号写入到文件,格式十六进制,负数用补码 fprintf(fid,'1651 1 0 1 0\n'); fprintf(fid,'0x%x\n', round(y*100) + (y<0)*2^16); 设置两个叠加信号的频率分别为10hz,100hz ,采样频率400hz,采样点数400,采样时间1S 。 linspace(x1,x2,N)用法:linspace 是Matlab 中的一个指令,用于产生x1,x2之间的N 点行矢量。其中x1、x2、N 分别为起始值、终止值、元素个数。若缺省N ,默认点数为100。在matlab 的命令窗口下输入help linspace 或者doc linspace 可以获得该函数的帮助信息。 y1,y2为频率不同的两个正弦信号,y=y1+y2为叠加后信号。Plot 画图显示信号y 。 00.10.20.30.40.50.60.70.80.91 -2-1.5 -1 -0.5 0.5 1 1.5 2

相关文档