34 Implicit Arguments RAJESH BHATT AND
datastage面试300题
1. What are the Environmental variables in Datastage?2. Check for Job Errors in datastage3. What are Stage V ariables, Derivations and Constants?4. What is Pipeline Parallelism?5. Debug stages in PX6. How do you remove duplicates in dataset7. What is the difference between Job Control and Job Sequence8. What is the max size of Data set stage?9. performance in sort stage10. How to develop the SCD using LOOKUP stage?12. What are the errors you expereiced with data stage13. what are the main diff between server job and parallel job in datastage14. Why you need Modify Stage?15. What is the difference between Squential Stage & Dataset Stage. When do u use them.16. memory allocation while using lookup stage17. What is Phantom error in the datastage. How to overcome this error.18. Parameter file usage in Datastage19. Explain the best approch to do a SCD type2 mapping in parallel job?20. how can we improve the performance of the job while handling huge amount of data21. HI How can we create read only jobs in Datastage.22. how to implement routines in data stage,have any one has any material for data stage23. How will you determine the sequence of jobs to load into data warehouse?24. How can we Test jobs in Datastage??25. DataStage - delete header and footer on the source sequential26. How can we implement Slowly Changing Dimensions in DataStage?.27. Differentiate Database data and Data warehouse data?28. How to run a Shell Script within the scope of a Data stage job?29. what is the difference between datastage and informatica30. Explain about job control language such as (DS_JOBS)32. What is Invocation ID?33. How to connect two stages which do not have any common columns between them?34. In SAP/R3, How do you declare and pass parameters in parallel job .35. Difference between Hashfile and Sequential File?36. How do you fix the error "OCI has fetched truncated data" in DataStage37. A batch is running and it is scheduled to run in 5 minutes. But after 10 days the time changes to 10 minutes. What type of error is this and how to fix it?38. Which partition we have to use for Aggregate Stage in parallel jobs ?39. What is the baseline to implement parition or parallel execution method in datastage job.e.g. more than 2 millions records only advised ?40. how do we create index in data satge?41. What is the flow of loading data into fact & dimensional tables?42. What is a sequential file that has single input link??43. Aggregators –What does the warning “Hash table has grown to …xyz‟ ….” mean?44. what is hashing algorithm?45. How do you load partial data after job failedsource has 10000 records, Job failed after 5000 records are loaded. This status of the job is abort , Instead of removing 5000 records from target , How can i resume the load46. What is Orchestrate options in generic stage, what are the option names. value ? Name of an Orchestrate operator to call. what are the orchestrate operators available in datastage for AIX environment.47. Type 30D hash file is GENERIC or SPECIFIC?48. Is Hashed file an Active or Passive Stage? When will be it useful?49. How do you extract job parameters from a file?50.1.What about System variables?2.How can we create Containers?3.How can we improve the performance of DataStage?4.what are the Job parameters?5.what is the difference between routine and transform and function?6.What are all the third party tools used in DataStage?7.How can we implement Lookup in DataStage Server jobs?8.How can we implement Slowly Changing Dimensions in DataStage?.9.How can we join one Oracle source and Sequential file?.10.What is iconv and oconv functions?51What are the difficulties faced in using DataStage ? or what are the constraints in using DataStage ?52. Have you ever involved in updating the DS versions like DS 5.X, if so tell us some the steps you have53. What r XML files and how do you read data from XML files and what stage to be used?54. How do you track performance statistics and enhance it?55. Types of vies in Datastage Director?There are 3 types of views in Datastage Director a) Job View - Dates of Jobs Compiled. b) Log View - Status of Job last run c) Status View - Warning Messages, Event Messages, Program Generated Messag56. What is the default cache size? How do you change the cache size if needed?Default cache size is 256 MB. We can incraese it by going into Datastage Administrator and selecting the Tunable Tab and specify the cache size over there.57. How do you pass the parameter to the job sequence if the job is running at night?58. How do you catch bad rows from OCI stage?59. what is quality stage and profile stage?60. what is the use and advantage of procedure in datastage?61. What are the important considerations while using join stage instead of lookups.62. how to implement type2 slowly changing dimenstion in datastage? give me with example?63. How to implement the type 2 Slowly Changing dimension in DataStage?64. What are Static Hash files and Dynamic Hash files?65. What is the difference between Datastage Server jobs and Datastage Parallel jobs?66. What is ' insert for update ' in datastage67. How did u connect to DB2 in your last project?Using DB2 ODBC drivers.68. How do you merge two files in DS?Either used Copy command as a Before-job subroutine if the metadata of the 2 files are same or created a job to concatenate the 2 files into one if the metadata is different.69. What is the order of execution done internally in the transformer with the stage editor having input links on the lft hand side and output links?70. How will you call external function or subroutine from datastage?71. What happens if the job fails at night?72. Types of Parallel Processing?Parallel Processing is broadly classified into 2 types. a) SMP - Symmetrical Multi Processing. b) MPP - Massive Parallel Processing.73. What is DS Administrator used for - did u use it?74. How do you do oracle 4 way inner join if there are 4 oracle input files?75. How do you pass filename as the parameter for a job?76. How do you populate source files?77. How to handle Date convertions in Datastage? Convert a mm/dd/yyyy format to yyyy-dd-mm? We use a) "Iconv" function - Internal Convertion. b) "Oconv" function - External Convertion. Function to convert mm/dd/yyyy format to yyyy-dd-mm is Oconv(Iconv(Filedname,"D/M78. How do you execute datastage job from command line prompt?Using "dsjob" command as follows. dsjob -run -jobstatus projectname jobname79. Differentiate Primary Key and Partition Key?Primary Key is a combination of unique and not null. It can be a collection of key values called as composite primary key. Partition Key is a just a part of Primary Key. There are several methods of80 How to install and configure DataStage EE on Sun Micro systems multi-processor hardware running the Solaris 9 operating system?Asked by: Kapil Jayne81. What are all the third party tools used in DataStage?82. How do you eliminate duplicate rows?83. what is the difference between routine and transform and function?84. Do you know about INTEGRITY/QUALITY stage?85. how to attach a mtr file (MapTrace) via email and the MapTrace is used to record all the execute map errors86. Is it possible to calculate a hash total for an EBCDIC file and have the hash total stored as EBCDIC using Datastage?Currently, the total is converted to ASCII, even tho the individual records are stored as EBCDIC.87. If your running 4 ways parallel and you have 10 stages on the canvas, how many processes does datastage create?88. Explain the differences between Oracle8i/9i?89. How will you pass the parameter to the job schedule if the job is running at night? What happens if one job fails in the night?90. what is an environment variable??91. how find duplicate records using transformer stage in server edition92. what is panthom error in data stage93. How can we increment the surrogate key value for every insert in to target database94. what is the use of environmental variables?95. how can we run the batch using command line?96. what is fact load?97. Explain a specific scenario where we would use range partitioning ?98. what is job commit in datastage?99. hi..Disadvantages of staging area Thanks,Jagan100. How do you configure api_dump102. Does type of partitioning change for SMP and MPP systems?103. what is the difference between RELEASE THE JOB and KILL THE JOB?104. Can you convert a snow flake schema into star schema?105. What is repository?106. What is Fact loading, how to do it?107. What is the alternative way where we can do job control??108.Where we can use these Stages Link Partetionar, Link Collector & Inter Process (OCI) Stage whether in Server Jobs or in Parallel Jobs ?And SMP is a Parallel or Server ?109. Where can you output data using the Peek Stage?110. Do u know about METASTAGE?111. In which situation,we are using RUN TIME COLUMN PROPAGA TION option?112. what is the difference between datasatge and datastage TX?113. 1 1. Difference between Hashfile and Sequential File?. What is modulus?2 2. What is iconv and oconv functions?.3 3. How can we join one Oracle source and Sequential file?.4 4. How can we implement Slowly Changing Dimensions in DataStage?.5 5. How can we implement Lookup in DataStage Server jobs?.6 6. What are all the third party tools used in DataStage?.7 7. what is the difference between routine and transform and function?.8 8. what are the Job parameters?.9 9. Plug-in?.10 10.How can we improv114. Is it possible to query a hash file? Justify your answer...115. How to enable the datastage engine?116. How I can convert Server Jobs into Parallel Jobs?117. Suppose you have table "sample" & three columns in that tablesample:Cola Colb Colc1 10 1002 20 2003 30 300Assume: cola is primary keyHow will you fetch the record with maximum cola value using data stage tool into the target system118. How to parametarise a field in a sequential file?I am using Datastage as ETL Tool,Sequential file as source.119. What is TX and what is the use of this in DataStage ? As I know TX stand for Transformer Extender, but I don't know how it will work and where we will used ?120. What is the difference betwen Merge Stage and Lookup Stage?121. Importance of Surrogate Key in Data warehousing?Surrogate Key is a Primary Key for a Dimension table. Most importance of using it is it is independent of underlying database. i.e Surrogate Key is not affected by the changes going on with a databas122. What is the difference between Symetrically parallel processing,Massively parallel processing?123.What is the diffrence between the Dynamic RDBMS Stage & Static RDBMS Stage ?124. How to run a job using command line?125. What is user activity in datastage?126. how can we improve the job performance?127. how we can create rank using datastge like in informatica128. What is the use of job controle??129. What does # indicate in environment variables?130. what are two types of hash files??131. What are different types of star schema??132. what are different types of file formats??133. What are different dimension table in your project??Plz explain me with an example?? 134. what is the difference between buildopts and subroutines ?135. how can we improve performance in aggregator stage??136. What is SQL tuning? how do you do it ?137. What is the use of tunnable??138. how to distinguish the surogate key in different dimensional tables?how can we give for different dimension tables?139. how can we load source into ODS?140. What is the difference between sequential file and a dataset? When to use the copy stage?141. how to eleminate duplicate rows in data stage?142. What is complex stage? In which situation we are using this one?143. What is the sequencer stage??144. where actually the flat files store?what is the path?145. what are the different types of lookups in datastage?146. What are the most important aspects that a beginner must consider doin his first DS project ?147. how to find errors in job sequence?148. it is possible to access the same job two users at a time in datastage?149. how to kill the job in data stage?150. how to find the process id?explain with steps?151. Why job sequence is use for? what is batches?what is the difference between job sequence and batches?152. What is Integrated & Unit testing in DataStage ?153. What is iconv and oconv functions?154. For what purpose is the Stage Variable is mainly used?155. purpose of using the key and difference between Surrogate keys and natural key156. how to read the data from XL FILES?my problem is my data file having some commas in data,but we are using delimitor is| ?how to read the data ,explain with steps?157. How can I schedule the cleaning of the file &PH& by dsjob?158. Hot Fix for ODBC Stage for AS400 V5R4 in Data Stage 7.1159. what is data stage engine?what is its purpose?160. What is the difference between Transform and Routine in DataStage?161. what is the meaning of the following..1)If an input file has an excessive number of rows and can be split-up then use standard 2)logic to run jobs in parallel3)Tuning should occur on a job-by-job basis. Use the power of DBMS.162. Why is hash file is faster than sequential file n odbc stage??163. Hello,Can both Source system(Oracle,SQLServer,...etc) and Target Data warehouse(may be oracle,SQLServer..etc) can be on windows environment or one of the system should be in UNIX/Linux environment.Thanks,Jagan164. How to write and execute routines for PX jobs in c++?165. what is a routine?166. how to distinguish the surrogate key in different dimentional tables?167. how can we generate a surrogate key in server/parallel jobs?168. what is NLS in datastage? how we use NLS in Datastage ? what advantages in that ? at thetime of installation i am not choosen that NLS option , now i want to use that options what can i do ? to reinstall that datastage or first uninstall and install once again ?169. how to read the data from XL FILES?explain with steps?170. whats the meaning of performance tunning techinque,Example??171. differentiate between pipeline and partion parallelism?172. What is the use of Hash file??insted of hash file why can we use sequential file itself?173. what is pivot stage?why are u using?what purpose that stage will be used?174. How did you handle reject data?175. Hiwhat is difference betweend ETL and ELT?176. how can we create environment variables in datasatage?177. what is the difference between static hash files n dynamic hash files?178. how can we test the jobs?179. What is the difference between reference link and straight link ?180. What are the command line functions that import and export the DS jobs?181. what is the size of the flat file?182. Whats difference betweeen operational data stage (ODS) & data warehouse?183. I have few questions1. What ar ethe various process which starts when the datastage engine starts?2. What are the changes need to be done on the database side, If I have to use dB2 stage?3. datastage engine is responsible for compilation or execution or both?184. Could anyone plz tell abt the full details of Datastage Certification.Title of Certification?Amount for Certification test?Where can v get the Tutorials available for certification?Who is Conducting the Certification Exam?Whether any training institute or person for guidens?I am very much pleased if anyone enlightwn me abt the above saidSuresh185. how to use rank&updatestratergy in datastage186. What is Ad-Hoc access? What is the difference between Managed Query and Ad-Hoc access?187. What is Runtime Column Propagation and how to use it?188. how we use the DataStage Director and its run-time engine to schedule running the solution, testing and debugging its components, and monitoring the resulting e/xecutable versions on ad hoc or scheduled basis?189. What is the difference bitween OCI stage and ODBC stage?190. Is there any difference b/n Ascential DataStage and DataStage.191. How do you remove duplicates without using remove duplicate stage?192. if we using two sources having same meta data and how to check the data in two sorces is same or nif we using two sources having same meta data and how to check the data in two sorces is same or not?and if the data is not same i want to abort the job ?how we can do this?193. If a DataStage job aborts after say 1000 records, how to continue the job from 1000th record after fixing the error?194. Can you tell me for what puorpse .dsx files are used in the datasatage195. how do u clean the datastage repository.196. give one real time situation where link partitioner stage used?197. What is environment variables?what is the use of this?198. How do you call procedures in datastage?199. How to remove duplicates in server job200. What is the exact difference betwwen Join,Merge and Lookup Stage??202. What are the new features of Datastage 7.1 from datastage 6.1203. How to run the job in command prompt in unix?204. How to know the no.of records in a sequential file before running a server job?205. Other than Round Robin, What is the algorithm used in link collecter? Also Explain How it will works?206. how to drop the index befor loading data in target and how to rebuild it in data stage?207. How can ETL excel file to Datamart?208. what is the transaction size and array size in OCI stage?how these can be used?209. what is job control?how it is developed?explain with steps?210. My requirement is like this :Here is the codification suggested: SALE_HEADER_XXXXX_YYYYMMDD.PSVSALEMy requirement is like this :Here is the codification suggested: SALE_HEADER_XXXXX_YYYYMMDD.PSVSALE_LINE_XXXXX_YYYYMMDD.PSVXXXXX = LVM sequence to ensure unicity and continuity of file exchangesCaution, there will an increment to implement.YYYYMMDD = LVM date of file creation COMPRESSION AND DELIVERY TO: SALE_HEADER_XXXXX_YYYYMMDD.ZIP AND SALE_LINE_XXXXX_YYYYMMDD.ZIPif we run that job the target file names are like this sale_header_1_20060206 & sale_line_1_20060206.If we run next time means the211. what is the purpose of exception activity in data stage 7.5?212. How to implement slowly changing dimentions in Datastage?213. What does separation option in static hash-file mean?214. how to improve the performance of hash file?215. Actually my requirement is like that :Here is the codification suggested: SALE_HEADER_XXXXX_YYYYMMActually my requirement is like that :Here is the codification suggested: SALE_HEADER_XXXXX_YYYYMMDD.PSVSALE_LINE_XXXXX_YYYYMMDD.PSVXXXXX = LVM sequence to ensure unicity and continuity of file exchangesCaution, there will an increment to implement.YYYYMMDD = LVM date of file creation COMPRESSION AND DELIVERY TO: SALE_HEADER_XXXXX_YYYYMMDD.ZIP AND SALE_LINE_XXXXX_YYYYMMDD.ZIPif we run that job the target file names are like this sale_header_1_20060206 & sale_line_1_20060206.if we run next216. How do u check for the consistency and integrity of model and repository?217. how we can call the routine in datastage job?explain with steps?218. what is job control?how can it used explain with steps?219. how to find the number of rows in a sequential file?220. If the size of the Hash file exceeds 2GB..What happens? Does it overwrite the current rows?221. where we use link partitioner in data stage job?explain with example?222 How i create datastage Engine stop start script.Actually my idea is as below.!#bin/bashdsadm - usersu - rootpassword (encript)DSHOMEBIN=/Ascential/DataStage/home/dsadm/Ascential/DataStage/DSEngine/binif check ps -ef | grep DataStage (client connection is there) { kill -9 PID (client connection) }uv -admin - stop > dev/nulluv -admin - start > dev/nullverify processcheck the connectionecho "Started properly"run it as dsadm223. can we use shared container as lookup in datastage server jobs?224. what is the meaning of instace in data stage?explain with examples?225. wht is the difference beteen validated ok and compiled in datastage.226. hi all what is auditstage,profilestage,qulaitystages in datastge please explain indetail227what is PROFILE STAGE , QUALITY STAGE,AUDIT STAGE in datastage..please expalin in detail.thanks in adv228. what are the environment variables in datastage?give some examples?229. What is difference between Merge stage and Join stage?230. Hican any one can explain what areDB2 UDB utilitiesub231. What is the difference between drs and odbc stage232. Will the data stage consider the second constraint in the transformer once the first condition is satisfied ( if the link odering is given)233. How do you do Usage analysis in datastage ?234. how can u implement slowly changed dimensions in datastage? explain?2) can u join flat file and database in datastage?how?235. How can you implement Complex Jobs in datastage236. DataStage from Staging to MDW is only running at 1 row per second! What do we do to remedy?237. what is the mean of Try to have the constraints in the 'Selection' criteria of the jobs iwhat is the mean of Try to have the constraints in the 'Selection' criteria of the jobs itself. This will eliminate the unnecessary records even getting in before joins are made?238. * What are constraints and derivation?* Explain the process of taking backup in DataStage?*What are the different types of lookups available in DataStage?239. # How does DataStage handle the user security?240. What are the Steps involved in development of a job in DataStage?241. What is a project? Specify its various components?242. What does a Config File in parallel extender consist of?Config file consists of the following. a) Number of Processes or Nodes. b) Actual Disk Storage Location.243. how to implement type2 slowly changing dimensions in data stage?explain with example?244. How much would be the size of the database in DataStage ?What is the difference between Inprocess and Interprocess ?245. Briefly describe the various client components?246. What are orabulk and bcp stages?247. What is DS Director used for - did u use it?248. what is meaning of file extender in data stage server jobs.can we run the data stage job from one job to another job that file data where it is stored and what is the file extender in ds jobs.249. What is the max capacity of Hash file in DataStage?250. what is merge and how it can be done plz explain with simple example taking 2 tables .......251. it is possible to run parallel jobs in server jobs?252. what are the enhancements made in datastage 7.5 compare with 7.0253. If I add a new environment variable in Windows, how can I access it in DataStage?254. what is OCI?255. Is it possible to move the data from oracle ware house to SAP Warehouse using withDA TASTAGE Tool.256. How can we create Containers?257. what is data set? and what is file set?258. How can I extract data from DB2 (on IBM iSeries) to the data warehouse via Datastage as the ETL tool. I mean do I first need to use ODBC to create connectivity and use an adapter for the extraction and transformation of data? Thanks so much if anybody could provide an answer.259. it is possible to call one job in another job in server jobs?260. how can we pass parameters to job by using file.261. How can we implement Lookup in DataStage Server jobs?262. what user varibale activity when it used how it used !where it is used with real example263. Did you Parameterize the job or hard-coded the values in the jobs?Always parameterized the job. Either the values are coming from Job Properties or from a …Parameter Manager‟ – a third part tool. There is no way you will hard–code some parameters in your jobs. The o264. what is hashing algorithm and explain breafly how it works?265. what happends out put of hash file is connected to transformer ..what error it throughs266. what is merge ?and how to use merge? merge is nothing but a filter conditions that have been used for filter condition267. What will you in a situation where somebody wants to send you a file and use that file as an input What will you in a situation where somebody wants to send you a file and use that file as an input or reference and then run job.268. What is the NLS equivalent to NLS oracle code American_7ASCII on Datastage NLS?269. Why do you use SQL LOADER or OCI STAGE?270. What about System variables?271. what are the differences between the data stage 7.0 and 7.5in server jobs?272. How the hash file is doing lookup in serverjobs?How is it comparing the key values?273. how to handle the rejected rows in datastage?274. how is datastage 4.0 functionally different from the enterprise edition now?? what are the exact changes?275. What is Hash file stage and what is it used for?Used for Look-ups. It is like a reference table. It is also used in-place of ODBC, OCI tables for better performance.276. What is the utility you use to schedule the jobs on a UNIX server other than using Ascential Director?Use crontab utility along with d***ecute() function along with proper parameters passed.277. How can I connect my DB2 database on AS400 to DataStage? Do I need to use ODBC 1st to open the database connectivity and then use an adapter for just connecting between the two? Thanks alot of any replies.278. what is the OCI? and how to use the ETL Tools?OCI means orabulk data which used client having bulk data its retrive time is much more ie., your used to orabulk data the divided and retrived Asked by: ramanamv279. what is difference between serverjobs & paraller jobs280. What is the difference between Datastage and Datastage TX?281. Hi!Can any one tell me how to extract data from more than 1 hetrogenious Sources.mean, example 1 sequenal file, Sybase , Oracle in a singale Job.282. How can we improve the performance of DataStage jobs?283. How good are you with your PL/SQL?On the scale of 1-10 say 8.5-9284. What are OConv () and Iconv () functions and where are they used?IConv() - Converts a string to an internal storage formatOConv() - Converts an expression to an output format.285. If data is partitioned in your job on key 1 and then you aggregate on key 2, what issues could arise?286. How can I specify a filter command for processing data while defining sequential file output data?287. There are three different types of user-created stages available for PX. What are they? Which would you use? What are the disadvantage for using each type?288. What is DS Manager used for - did u use it?289. What are Sequencers?Sequencers are job control programs that execute other jobs with preset Job parameters.290. Functionality of Link Partitioner and Link Collector?291. Containers : Usage and Types?Container is a collection of stages used for the purpose of Reusability. There are 2 types of Containers. a) Local Container: Job Specific b) Shared Container: Used in any job within a project.292. Does Enterprise Edition only add the parallel processing for better performance?Are any stages/transformations available in the enterprise edition only?293. what are validations you perform after creating jobs in designer.what r the different type of errors u faced during loading and how u solve them294. how can you do incremental load in datastage?295. how we use NLS function in Datastage? what are advantages of NLS function? where we can use that one? explain briefly?296. Dimension Modelling types along with their significanceData Modelling is Broadly classified into 2 types. a) E-R Diagrams (Entity - Relatioships). b) Dimensional Modelling.297. Did you work in UNIX environment?Yes. One of the most important requirements.298. What other ETL's you have worked with?Informatica and also DataJunction if it is present in your Resume.299. What is APT_CONFIG in datastage300. Does the BibhudataStage Oracle plug-in better than OCI plug-in coming from DataStage? What is theBibhudataStage extra functions?301. How do we do the automation of dsjobs?302. what is trouble shhoting in server jobs ? what are the diff kinds of errors encountered while。
JAVA考核题-面向对象选择题(答案)
北润JAVA考核-面向对象第一部分英语测试（每题分，共分）第二部分知识点测试（分）一、选择题（每题2分，共110分）1．下面关于变量及其作用范围的陈述哪个是不对的?（B ）A．实例变量是类的成员变量。
B．实例变量用关键字static声明。
C．在方法中定义的局部变量在该方法被执行时创建。
D．局部变量在使用前必须被初始化。
2．下面哪条语句把方法声明为抽象的公共方法?（B ）A．public abstract method(); B．public abstract void method();C．public abstract void method(){} D．public void method() extends abstract;3．若在某一个类定义中定义有如下的方法：final void aFinalFunction( ){}则该方法属于（ C ）。
A、本地方法B、静态方法C、最终方法D、抽象方法4．main方法是Java Application程序执行的入口点，关于main方法的方法头以下哪项是合法的（ B ）。
A、public static void main()B、public static void main(String[ ] args)C、public static int main(String[ ] args)D、public void main(String arg[ ])5．在Java中，一个类可同时定义许多同名的方法，这些方法的形式参数个数、类型或顺序各不相同，传回的值也可以不相同。
这种面向对象程序的特性称为（ C ）。
A、隐藏B、覆盖C、重载D、Java不支持此特性6．下列关于构造方法的叙述中，错误的是（C ）A．Java语言规定构造方法名与类名必须相同B．Java语言规定构造方法没有返回值，但不用void声明C．Java语言规定构造方法不可以重载D．Java语言规定构造方法只能通过new自动调用7．关于被私有访问控制符private修饰的成员变量，以下说法正确的是（C ）A．可以被三种类所引用：该类自身、与它在同一个包中的其他类、在其他包中的该类的子类B．可以被两种类访问和引用：该类本身、该类的所有子类C．只能被该类自身所访问和修改D．只能被同一个包中的类访问8．类Test1定义如下：1. public class Test1{2．public floataMethod(float a, float b){}3．4．}将以下哪种方法插入行3是不合法的。
FORTRAN运行错误消息列表中英对照
Fortran的运行时错误消息列表本节列出了英特尔Fortran运行时库（RTL）处理的错误。
对于每一个错误，该表提供了错误号，严重性代码，错误信息文本，条件符号名称，而错误的详细说明。
在程序中定义条件符号值（参数表），包括以下文件：for_iosdef.for如表中所述，消息的严重程度决定了发生下列情况：•与信息和警告，程序继续执行•与错误，结果可能会不正确•与严重的，程序执行停止（除非指定了恢复方法）在最后一种情况下，为防止程序终止，您必须包含一个合适的I / O错误处理说明符并重新编译，或者对于某些错误，改变信号的缺省操作您再次运行该程序之前。
在下面的表中，第一列列出的错误号返回检测到I / O错误时iostat的变量。
第二列的第一行提供的消息，因为它会显示（以下forrtl：?），包括严重级别，消息号，消息文本。
第二列下面的行包含状态条件符号（如$ IOS_INCRECTYP）和消息的解释。
脚注：1标识不IOSTAT返回的错误。
原网页：fortran_docs/compiler_f/main_for/mergedProjects/bldaps_for/common/bldaps_rterrs.htm 英文原版List of Run-Time Error MessagesThis section lists the errors processed by the Intel Fortran run-time library (RTL). For each error, the table provides the error number, the severity code, error message text, condition symbol name, and a detailed description of the error.To define the condition symbol values (PARAMETER statements) in your program, include the following file:for_iosdef.forAs described in the table, the severity of the message determines which of the following occurs:•with?info?and?warning, program execution continues•with?error, the results may be incorrect•with?severe, program execution stops (unless a recovery method is specified)In the last case, to prevent program termination, you must include either an appropriate I/O error-handling specifier and recompile or, for certain errors, change the default action of a signal before you run the program again.In the following table, the first column lists error numbers returned to IOSTAT variables when an I/O error is detected.The first line of the second column provides the message as it is displayed (following?forrtl:), including the severity level, message number, and the message text. The following lines of the second column contain the status condition symbol (such as FOR$IOS_INCRECTYP) and an explanation of the message.。
The Quadratic Assignment Problem
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
9 Available Computer Codes for the QAP 10 Polynomially Solvable Cases 11 QAP Instances with Known Optimal Solution 12 Asymptotic Behavior 13 Related Problems
11 11 12 13
5 QAP Polytopes
14
This research has been supported by the Spezialforschungsbereich F 003 \Optimierung und Kontrolle", Projektbereich Diskrete Optimierung. y Technische Universitat Graz, Institut fur Mathematik B, Steyrergasse 30, A-8010 Graz, Austria. z Center for Applied Optimization, Industrial and Systems Engineering Department, University of Florida, Gainesville, FL 32611
: : : : : : : : : : : : : : : : : : : : : : : : :
INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL
SUMMARY In this paper, a modular modification of the adaptive robust control (ARC) technique is presented. The modular design has all of the original ARC properties with an estimation-based update law instead of a Lyapunov-based update law. In this design, the controller is divided into two modules: a control module and an identification module. A key new idea is to set a priori bounds on the time derivatives of the estimates to be maintained by the update law. As a result, their effects on the system tracking accuracy can be dominated by the control law. A modification is proposed for the standard gradient and least-square update laws to guarantee the bounds. This modification also makes the controller robust against the generalized (unparameterized) uncertainties considered in the ARC formulation while allowing asymptotic output tracking without the generalized uncertainties. Both the ARC and the modular ARC techniques are applied to a force control problem for an active suspension system. Simulations and experimental results are provided to show that the update law of the modular design is less sensitive to measurement noise which results in smaller force tracking error and smaller control gain. Copyright # 2004 John Wiley & Sons, Ltd.
《神经网络与深度学习综述DeepLearning15May2014
Draft:Deep Learning in Neural Networks:An OverviewTechnical Report IDSIA-03-14/arXiv:1404.7828(v1.5)[cs.NE]J¨u rgen SchmidhuberThe Swiss AI Lab IDSIAIstituto Dalle Molle di Studi sull’Intelligenza ArtificialeUniversity of Lugano&SUPSIGalleria2,6928Manno-LuganoSwitzerland15May2014AbstractIn recent years,deep artificial neural networks(including recurrent ones)have won numerous con-tests in pattern recognition and machine learning.This historical survey compactly summarises relevantwork,much of it from the previous millennium.Shallow and deep learners are distinguished by thedepth of their credit assignment paths,which are chains of possibly learnable,causal links between ac-tions and effects.I review deep supervised learning(also recapitulating the history of backpropagation),unsupervised learning,reinforcement learning&evolutionary computation,and indirect search for shortprograms encoding deep and large networks.PDF of earlier draft(v1):http://www.idsia.ch/∼juergen/DeepLearning30April2014.pdfLATEX source:http://www.idsia.ch/∼juergen/DeepLearning30April2014.texComplete BIBTEXfile:http://www.idsia.ch/∼juergen/bib.bibPrefaceThis is the draft of an invited Deep Learning(DL)overview.One of its goals is to assign credit to those who contributed to the present state of the art.I acknowledge the limitations of attempting to achieve this goal.The DL research community itself may be viewed as a continually evolving,deep network of scientists who have influenced each other in complex ways.Starting from recent DL results,I tried to trace back the origins of relevant ideas through the past half century and beyond,sometimes using“local search”to follow citations of citations backwards in time.Since not all DL publications properly acknowledge earlier relevant work,additional global search strategies were employed,aided by consulting numerous neural network experts.As a result,the present draft mostly consists of references(about800entries so far).Nevertheless,through an expert selection bias I may have missed important work.A related bias was surely introduced by my special familiarity with the work of my own DL research group in the past quarter-century.For these reasons,the present draft should be viewed as merely a snapshot of an ongoing credit assignment process.To help improve it,please do not hesitate to send corrections and suggestions to juergen@idsia.ch.Contents1Introduction to Deep Learning(DL)in Neural Networks(NNs)3 2Event-Oriented Notation for Activation Spreading in FNNs/RNNs3 3Depth of Credit Assignment Paths(CAPs)and of Problems4 4Recurring Themes of Deep Learning54.1Dynamic Programming(DP)for DL (5)4.2Unsupervised Learning(UL)Facilitating Supervised Learning(SL)and RL (6)4.3Occam’s Razor:Compression and Minimum Description Length(MDL) (6)4.4Learning Hierarchical Representations Through Deep SL,UL,RL (6)4.5Fast Graphics Processing Units(GPUs)for DL in NNs (6)5Supervised NNs,Some Helped by Unsupervised NNs75.11940s and Earlier (7)5.2Around1960:More Neurobiological Inspiration for DL (7)5.31965:Deep Networks Based on the Group Method of Data Handling(GMDH) (8)5.41979:Convolution+Weight Replication+Winner-Take-All(WTA) (8)5.51960-1981and Beyond:Development of Backpropagation(BP)for NNs (8)5.5.1BP for Weight-Sharing Feedforward NNs(FNNs)and Recurrent NNs(RNNs)..95.6Late1980s-2000:Numerous Improvements of NNs (9)5.6.1Ideas for Dealing with Long Time Lags and Deep CAPs (10)5.6.2Better BP Through Advanced Gradient Descent (10)5.6.3Discovering Low-Complexity,Problem-Solving NNs (11)5.6.4Potential Benefits of UL for SL (11)5.71987:UL Through Autoencoder(AE)Hierarchies (12)5.81989:BP for Convolutional NNs(CNNs) (13)5.91991:Fundamental Deep Learning Problem of Gradient Descent (13)5.101991:UL-Based History Compression Through a Deep Hierarchy of RNNs (14)5.111992:Max-Pooling(MP):Towards MPCNNs (14)5.121994:Contest-Winning Not So Deep NNs (15)5.131995:Supervised Recurrent Very Deep Learner(LSTM RNN) (15)5.142003:More Contest-Winning/Record-Setting,Often Not So Deep NNs (16)5.152006/7:Deep Belief Networks(DBNs)&AE Stacks Fine-Tuned by BP (17)5.162006/7:Improved CNNs/GPU-CNNs/BP-Trained MPCNNs (17)5.172009:First Official Competitions Won by RNNs,and with MPCNNs (18)5.182010:Plain Backprop(+Distortions)on GPU Yields Excellent Results (18)5.192011:MPCNNs on GPU Achieve Superhuman Vision Performance (18)5.202011:Hessian-Free Optimization for RNNs (19)5.212012:First Contests Won on ImageNet&Object Detection&Segmentation (19)5.222013-:More Contests and Benchmark Records (20)5.22.1Currently Successful Supervised Techniques:LSTM RNNs/GPU-MPCNNs (21)5.23Recent Tricks for Improving SL Deep NNs(Compare Sec.5.6.2,5.6.3) (21)5.24Consequences for Neuroscience (22)5.25DL with Spiking Neurons? (22)6DL in FNNs and RNNs for Reinforcement Learning(RL)236.1RL Through NN World Models Yields RNNs With Deep CAPs (23)6.2Deep FNNs for Traditional RL and Markov Decision Processes(MDPs) (24)6.3Deep RL RNNs for Partially Observable MDPs(POMDPs) (24)6.4RL Facilitated by Deep UL in FNNs and RNNs (25)6.5Deep Hierarchical RL(HRL)and Subgoal Learning with FNNs and RNNs (25)6.6Deep RL by Direct NN Search/Policy Gradients/Evolution (25)6.7Deep RL by Indirect Policy Search/Compressed NN Search (26)6.8Universal RL (27)7Conclusion271Introduction to Deep Learning(DL)in Neural Networks(NNs) Which modifiable components of a learning system are responsible for its success or failure?What changes to them improve performance?This has been called the fundamental credit assignment problem(Minsky, 1963).There are general credit assignment methods for universal problem solvers that are time-optimal in various theoretical senses(Sec.6.8).The present survey,however,will focus on the narrower,but now commercially important,subfield of Deep Learning(DL)in Artificial Neural Networks(NNs).We are interested in accurate credit assignment across possibly many,often nonlinear,computational stages of NNs.Shallow NN-like models have been around for many decades if not centuries(Sec.5.1).Models with several successive nonlinear layers of neurons date back at least to the1960s(Sec.5.3)and1970s(Sec.5.5). An efficient gradient descent method for teacher-based Supervised Learning(SL)in discrete,differentiable networks of arbitrary depth called backpropagation(BP)was developed in the1960s and1970s,and ap-plied to NNs in1981(Sec.5.5).BP-based training of deep NNs with many layers,however,had been found to be difficult in practice by the late1980s(Sec.5.6),and had become an explicit research subject by the early1990s(Sec.5.9).DL became practically feasible to some extent through the help of Unsupervised Learning(UL)(e.g.,Sec.5.10,5.15).The1990s and2000s also saw many improvements of purely super-vised DL(Sec.5).In the new millennium,deep NNs havefinally attracted wide-spread attention,mainly by outperforming alternative machine learning methods such as kernel machines(Vapnik,1995;Sch¨o lkopf et al.,1998)in numerous important applications.In fact,supervised deep NNs have won numerous of-ficial international pattern recognition competitions(e.g.,Sec.5.17,5.19,5.21,5.22),achieving thefirst superhuman visual pattern recognition results in limited domains(Sec.5.19).Deep NNs also have become relevant for the more generalfield of Reinforcement Learning(RL)where there is no supervising teacher (Sec.6).Both feedforward(acyclic)NNs(FNNs)and recurrent(cyclic)NNs(RNNs)have won contests(Sec.5.12,5.14,5.17,5.19,5.21,5.22).In a sense,RNNs are the deepest of all NNs(Sec.3)—they are general computers more powerful than FNNs,and can in principle create and process memories of ar-bitrary sequences of input patterns(e.g.,Siegelmann and Sontag,1991;Schmidhuber,1990a).Unlike traditional methods for automatic sequential program synthesis(e.g.,Waldinger and Lee,1969;Balzer, 1985;Soloway,1986;Deville and Lau,1994),RNNs can learn programs that mix sequential and parallel information processing in a natural and efficient way,exploiting the massive parallelism viewed as crucial for sustaining the rapid decline of computation cost observed over the past75years.The rest of this paper is structured as follows.Sec.2introduces a compact,event-oriented notation that is simple yet general enough to accommodate both FNNs and RNNs.Sec.3introduces the concept of Credit Assignment Paths(CAPs)to measure whether learning in a given NN application is of the deep or shallow type.Sec.4lists recurring themes of DL in SL,UL,and RL.Sec.5focuses on SL and UL,and on how UL can facilitate SL,although pure SL has become dominant in recent competitions(Sec.5.17-5.22). Sec.5is arranged in a historical timeline format with subsections on important inspirations and technical contributions.Sec.6on deep RL discusses traditional Dynamic Programming(DP)-based RL combined with gradient-based search techniques for SL or UL in deep NNs,as well as general methods for direct and indirect search in the weight space of deep FNNs and RNNs,including successful policy gradient and evolutionary methods.2Event-Oriented Notation for Activation Spreading in FNNs/RNNs Throughout this paper,let i,j,k,t,p,q,r denote positive integer variables assuming ranges implicit in the given contexts.Let n,m,T denote positive integer constants.An NN’s topology may change over time(e.g.,Fahlman,1991;Ring,1991;Weng et al.,1992;Fritzke, 1994).At any given moment,it can be described as afinite subset of units(or nodes or neurons)N= {u1,u2,...,}and afinite set H⊆N×N of directed edges or connections between nodes.FNNs are acyclic graphs,RNNs cyclic.Thefirst(input)layer is the set of input units,a subset of N.In FNNs,the k-th layer(k>1)is the set of all nodes u∈N such that there is an edge path of length k−1(but no longer path)between some input unit and u.There may be shortcut connections between distant layers.The NN’s behavior or program is determined by a set of real-valued,possibly modifiable,parameters or weights w i(i=1,...,n).We now focus on a singlefinite episode or epoch of information processing and activation spreading,without learning through weight changes.The following slightly unconventional notation is designed to compactly describe what is happening during the runtime of the system.During an episode,there is a partially causal sequence x t(t=1,...,T)of real values that I call events.Each x t is either an input set by the environment,or the activation of a unit that may directly depend on other x k(k<t)through a current NN topology-dependent set in t of indices k representing incoming causal connections or links.Let the function v encode topology information and map such event index pairs(k,t)to weight indices.For example,in the non-input case we may have x t=f t(net t)with real-valued net t= k∈in t x k w v(k,t)(additive case)or net t= k∈in t x k w v(k,t)(multiplicative case), where f t is a typically nonlinear real-valued activation function such as tanh.In many recent competition-winning NNs(Sec.5.19,5.21,5.22)there also are events of the type x t=max k∈int (x k);some networktypes may also use complex polynomial activation functions(Sec.5.3).x t may directly affect certain x k(k>t)through outgoing connections or links represented through a current set out t of indices k with t∈in k.Some non-input events are called output events.Note that many of the x t may refer to different,time-varying activations of the same unit in sequence-processing RNNs(e.g.,Williams,1989,“unfolding in time”),or also in FNNs sequentially exposed to time-varying input patterns of a large training set encoded as input events.During an episode,the same weight may get reused over and over again in topology-dependent ways,e.g.,in RNNs,or in convolutional NNs(Sec.5.4,5.8).I call this weight sharing across space and/or time.Weight sharing may greatly reduce the NN’s descriptive complexity,which is the number of bits of information required to describe the NN (Sec.4.3).In Supervised Learning(SL),certain NN output events x t may be associated with teacher-given,real-valued labels or targets d t yielding errors e t,e.g.,e t=1/2(x t−d t)2.A typical goal of supervised NN training is tofind weights that yield episodes with small total error E,the sum of all such e t.The hope is that the NN will generalize well in later episodes,causing only small errors on previously unseen sequences of input events.Many alternative error functions for SL and UL are possible.SL assumes that input events are independent of earlier output events(which may affect the environ-ment through actions causing subsequent perceptions).This assumption does not hold in the broaderfields of Sequential Decision Making and Reinforcement Learning(RL)(Kaelbling et al.,1996;Sutton and Barto, 1998;Hutter,2005)(Sec.6).In RL,some of the input events may encode real-valued reward signals given by the environment,and a typical goal is tofind weights that yield episodes with a high sum of reward signals,through sequences of appropriate output actions.Sec.5.5will use the notation above to compactly describe a central algorithm of DL,namely,back-propagation(BP)for supervised weight-sharing FNNs and RNNs.(FNNs may be viewed as RNNs with certainfixed zero weights.)Sec.6will address the more general RL case.3Depth of Credit Assignment Paths(CAPs)and of ProblemsTo measure whether credit assignment in a given NN application is of the deep or shallow type,I introduce the concept of Credit Assignment Paths or CAPs,which are chains of possibly causal links between events.Let usfirst focus on SL.Consider two events x p and x q(1≤p<q≤T).Depending on the appli-cation,they may have a Potential Direct Causal Connection(PDCC)expressed by the Boolean predicate pdcc(p,q),which is true if and only if p∈in q.Then the2-element list(p,q)is defined to be a CAP from p to q(a minimal one).A learning algorithm may be allowed to change w v(p,q)to improve performance in future episodes.More general,possibly indirect,Potential Causal Connections(PCC)are expressed by the recursively defined Boolean predicate pcc(p,q),which in the SL case is true only if pdcc(p,q),or if pcc(p,k)for some k and pdcc(k,q).In the latter case,appending q to any CAP from p to k yields a CAP from p to q(this is a recursive definition,too).The set of such CAPs may be large but isfinite.Note that the same weight may affect many different PDCCs between successive events listed by a given CAP,e.g.,in the case of RNNs, or weight-sharing FNNs.Suppose a CAP has the form(...,k,t,...,q),where k and t(possibly t=q)are thefirst successive elements with modifiable w v(k,t).Then the length of the suffix list(t,...,q)is called the CAP’s depth (which is0if there are no modifiable links at all).This depth limits how far backwards credit assignment can move down the causal chain tofind a modifiable weight.1Suppose an episode and its event sequence x1,...,x T satisfy a computable criterion used to decide whether a given problem has been solved(e.g.,total error E below some threshold).Then the set of used weights is called a solution to the problem,and the depth of the deepest CAP within the sequence is called the solution’s depth.There may be other solutions(yielding different event sequences)with different depths.Given somefixed NN topology,the smallest depth of any solution is called the problem’s depth.Sometimes we also speak of the depth of an architecture:SL FNNs withfixed topology imply a problem-independent maximal problem depth bounded by the number of non-input layers.Certain SL RNNs withfixed weights for all connections except those to output units(Jaeger,2001;Maass et al.,2002; Jaeger,2004;Schrauwen et al.,2007)have a maximal problem depth of1,because only thefinal links in the corresponding CAPs are modifiable.In general,however,RNNs may learn to solve problems of potentially unlimited depth.Note that the definitions above are solely based on the depths of causal chains,and agnostic of the temporal distance between events.For example,shallow FNNs perceiving large“time windows”of in-put events may correctly classify long input sequences through appropriate output events,and thus solve shallow problems involving long time lags between relevant events.At which problem depth does Shallow Learning end,and Deep Learning begin?Discussions with DL experts have not yet yielded a conclusive response to this question.Instead of committing myself to a precise answer,let me just define for the purposes of this overview:problems of depth>10require Very Deep Learning.The difficulty of a problem may have little to do with its depth.Some NNs can quickly learn to solve certain deep problems,e.g.,through random weight guessing(Sec.5.9)or other types of direct search (Sec.6.6)or indirect search(Sec.6.7)in weight space,or through training an NNfirst on shallow problems whose solutions may then generalize to deep problems,or through collapsing sequences of(non)linear operations into a single(non)linear operation—but see an analysis of non-trivial aspects of deep linear networks(Baldi and Hornik,1994,Section B).In general,however,finding an NN that precisely models a given training set is an NP-complete problem(Judd,1990;Blum and Rivest,1992),also in the case of deep NNs(S´ıma,1994;de Souto et al.,1999;Windisch,2005);compare a survey of negative results(S´ıma, 2002,Section1).Above we have focused on SL.In the more general case of RL in unknown environments,pcc(p,q) is also true if x p is an output event and x q any later input event—any action may affect the environment and thus any later perception.(In the real world,the environment may even influence non-input events computed on a physical hardware entangled with the entire universe,but this is ignored here.)It is possible to model and replace such unmodifiable environmental PCCs through a part of the NN that has already learned to predict(through some of its units)input events(including reward signals)from former input events and actions(Sec.6.1).Its weights are frozen,but can help to assign credit to other,still modifiable weights used to compute actions(Sec.6.1).This approach may lead to very deep CAPs though.Some DL research is about automatically rephrasing problems such that their depth is reduced(Sec.4). In particular,sometimes UL is used to make SL problems less deep,e.g.,Sec.5.10.Often Dynamic Programming(Sec.4.1)is used to facilitate certain traditional RL problems,e.g.,Sec.6.2.Sec.5focuses on CAPs for SL,Sec.6on the more complex case of RL.4Recurring Themes of Deep Learning4.1Dynamic Programming(DP)for DLOne recurring theme of DL is Dynamic Programming(DP)(Bellman,1957),which can help to facili-tate credit assignment under certain assumptions.For example,in SL NNs,backpropagation itself can 1An alternative would be to count only modifiable links when measuring depth.In many typical NN applications this would not make a difference,but in some it would,e.g.,Sec.6.1.be viewed as a DP-derived method(Sec.5.5).In traditional RL based on strong Markovian assumptions, DP-derived methods can help to greatly reduce problem depth(Sec.6.2).DP algorithms are also essen-tial for systems that combine concepts of NNs and graphical models,such as Hidden Markov Models (HMMs)(Stratonovich,1960;Baum and Petrie,1966)and Expectation Maximization(EM)(Dempster et al.,1977),e.g.,(Bottou,1991;Bengio,1991;Bourlard and Morgan,1994;Baldi and Chauvin,1996; Jordan and Sejnowski,2001;Bishop,2006;Poon and Domingos,2011;Dahl et al.,2012;Hinton et al., 2012a).4.2Unsupervised Learning(UL)Facilitating Supervised Learning(SL)and RL Another recurring theme is how UL can facilitate both SL(Sec.5)and RL(Sec.6).UL(Sec.5.6.4) is normally used to encode raw incoming data such as video or speech streams in a form that is more convenient for subsequent goal-directed learning.In particular,codes that describe the original data in a less redundant or more compact way can be fed into SL(Sec.5.10,5.15)or RL machines(Sec.6.4),whose search spaces may thus become smaller(and whose CAPs shallower)than those necessary for dealing with the raw data.UL is closely connected to the topics of regularization and compression(Sec.4.3,5.6.3). 4.3Occam’s Razor:Compression and Minimum Description Length(MDL) Occam’s razor favors simple solutions over complex ones.Given some programming language,the prin-ciple of Minimum Description Length(MDL)can be used to measure the complexity of a solution candi-date by the length of the shortest program that computes it(e.g.,Solomonoff,1964;Kolmogorov,1965b; Chaitin,1966;Wallace and Boulton,1968;Levin,1973a;Rissanen,1986;Blumer et al.,1987;Li and Vit´a nyi,1997;Gr¨u nwald et al.,2005).Some methods explicitly take into account program runtime(Al-lender,1992;Watanabe,1992;Schmidhuber,2002,1995);many consider only programs with constant runtime,written in non-universal programming languages(e.g.,Rissanen,1986;Hinton and van Camp, 1993).In the NN case,the MDL principle suggests that low NN weight complexity corresponds to high NN probability in the Bayesian view(e.g.,MacKay,1992;Buntine and Weigend,1991;De Freitas,2003), and to high generalization performance(e.g.,Baum and Haussler,1989),without overfitting the training data.Many methods have been proposed for regularizing NNs,that is,searching for solution-computing, low-complexity SL NNs(Sec.5.6.3)and RL NNs(Sec.6.7).This is closely related to certain UL methods (Sec.4.2,5.6.4).4.4Learning Hierarchical Representations Through Deep SL,UL,RLMany methods of Good Old-Fashioned Artificial Intelligence(GOFAI)(Nilsson,1980)as well as more recent approaches to AI(Russell et al.,1995)and Machine Learning(Mitchell,1997)learn hierarchies of more and more abstract data representations.For example,certain methods of syntactic pattern recog-nition(Fu,1977)such as grammar induction discover hierarchies of formal rules to model observations. The partially(un)supervised Automated Mathematician/EURISKO(Lenat,1983;Lenat and Brown,1984) continually learns concepts by combining previously learnt concepts.Such hierarchical representation learning(Ring,1994;Bengio et al.,2013;Deng and Yu,2014)is also a recurring theme of DL NNs for SL (Sec.5),UL-aided SL(Sec.5.7,5.10,5.15),and hierarchical RL(Sec.6.5).Often,abstract hierarchical representations are natural by-products of data compression(Sec.4.3),e.g.,Sec.5.10.4.5Fast Graphics Processing Units(GPUs)for DL in NNsWhile the previous millennium saw several attempts at creating fast NN-specific hardware(e.g.,Jackel et al.,1990;Faggin,1992;Ramacher et al.,1993;Widrow et al.,1994;Heemskerk,1995;Korkin et al., 1997;Urlbe,1999),and at exploiting standard hardware(e.g.,Anguita et al.,1994;Muller et al.,1995; Anguita and Gomes,1996),the new millennium brought a DL breakthrough in form of cheap,multi-processor graphics cards or GPUs.GPUs are widely used for video games,a huge and competitive market that has driven down hardware prices.GPUs excel at fast matrix and vector multiplications required not only for convincing virtual realities but also for NN training,where they can speed up learning by a factorof50and more.Some of the GPU-based FNN implementations(Sec.5.16-5.19)have greatly contributed to recent successes in contests for pattern recognition(Sec.5.19-5.22),image segmentation(Sec.5.21), and object detection(Sec.5.21-5.22).5Supervised NNs,Some Helped by Unsupervised NNsThe main focus of current practical applications is on Supervised Learning(SL),which has dominated re-cent pattern recognition contests(Sec.5.17-5.22).Several methods,however,use additional Unsupervised Learning(UL)to facilitate SL(Sec.5.7,5.10,5.15).It does make sense to treat SL and UL in the same section:often gradient-based methods,such as BP(Sec.5.5.1),are used to optimize objective functions of both UL and SL,and the boundary between SL and UL may blur,for example,when it comes to time series prediction and sequence classification,e.g.,Sec.5.10,5.12.A historical timeline format will help to arrange subsections on important inspirations and techni-cal contributions(although such a subsection may span a time interval of many years).Sec.5.1briefly mentions early,shallow NN models since the1940s,Sec.5.2additional early neurobiological inspiration relevant for modern Deep Learning(DL).Sec.5.3is about GMDH networks(since1965),perhaps thefirst (feedforward)DL systems.Sec.5.4is about the relatively deep Neocognitron NN(1979)which is similar to certain modern deep FNN architectures,as it combines convolutional NNs(CNNs),weight pattern repli-cation,and winner-take-all(WTA)mechanisms.Sec.5.5uses the notation of Sec.2to compactly describe a central algorithm of DL,namely,backpropagation(BP)for supervised weight-sharing FNNs and RNNs. It also summarizes the history of BP1960-1981and beyond.Sec.5.6describes problems encountered in the late1980s with BP for deep NNs,and mentions several ideas from the previous millennium to overcome them.Sec.5.7discusses afirst hierarchical stack of coupled UL-based Autoencoders(AEs)—this concept resurfaced in the new millennium(Sec.5.15).Sec.5.8is about applying BP to CNNs,which is important for today’s DL applications.Sec.5.9explains BP’s Fundamental DL Problem(of vanishing/exploding gradients)discovered in1991.Sec.5.10explains how a deep RNN stack of1991(the History Compressor) pre-trained by UL helped to solve previously unlearnable DL benchmarks requiring Credit Assignment Paths(CAPs,Sec.3)of depth1000and more.Sec.5.11discusses a particular WTA method called Max-Pooling(MP)important in today’s DL FNNs.Sec.5.12mentions afirst important contest won by SL NNs in1994.Sec.5.13describes a purely supervised DL RNN(Long Short-Term Memory,LSTM)for problems of depth1000and more.Sec.5.14mentions an early contest of2003won by an ensemble of shallow NNs, as well as good pattern recognition results with CNNs and LSTM RNNs(2003).Sec.5.15is mostly about Deep Belief Networks(DBNs,2006)and related stacks of Autoencoders(AEs,Sec.5.7)pre-trained by UL to facilitate BP-based SL.Sec.5.16mentions thefirst BP-trained MPCNNs(2007)and GPU-CNNs(2006). Sec.5.17-5.22focus on official competitions with secret test sets won by(mostly purely supervised)DL NNs since2009,in sequence recognition,image classification,image segmentation,and object detection. Many RNN results depended on LSTM(Sec.5.13);many FNN results depended on GPU-based FNN code developed since2004(Sec.5.16,5.17,5.18,5.19),in particular,GPU-MPCNNs(Sec.5.19).5.11940s and EarlierNN research started in the1940s(e.g.,McCulloch and Pitts,1943;Hebb,1949);compare also later work on learning NNs(Rosenblatt,1958,1962;Widrow and Hoff,1962;Grossberg,1969;Kohonen,1972; von der Malsburg,1973;Narendra and Thathatchar,1974;Willshaw and von der Malsburg,1976;Palm, 1980;Hopfield,1982).In a sense NNs have been around even longer,since early supervised NNs were essentially variants of linear regression methods going back at least to the early1800s(e.g.,Legendre, 1805;Gauss,1809,1821).Early NNs had a maximal CAP depth of1(Sec.3).5.2Around1960:More Neurobiological Inspiration for DLSimple cells and complex cells were found in the cat’s visual cortex(e.g.,Hubel and Wiesel,1962;Wiesel and Hubel,1959).These cellsfire in response to certain properties of visual sensory inputs,such as theorientation of plex cells exhibit more spatial invariance than simple cells.This inspired later deep NN architectures(Sec.5.4)used in certain modern award-winning Deep Learners(Sec.5.19-5.22).5.31965:Deep Networks Based on the Group Method of Data Handling(GMDH) Networks trained by the Group Method of Data Handling(GMDH)(Ivakhnenko and Lapa,1965; Ivakhnenko et al.,1967;Ivakhnenko,1968,1971)were perhaps thefirst DL systems of the Feedforward Multilayer Perceptron type.The units of GMDH nets may have polynomial activation functions imple-menting Kolmogorov-Gabor polynomials(more general than traditional NN activation functions).Given a training set,layers are incrementally grown and trained by regression analysis,then pruned with the help of a separate validation set(using today’s terminology),where Decision Regularisation is used to weed out superfluous units.The numbers of layers and units per layer can be learned in problem-dependent fashion. This is a good example of hierarchical representation learning(Sec.4.4).There have been numerous ap-plications of GMDH-style networks,e.g.(Ikeda et al.,1976;Farlow,1984;Madala and Ivakhnenko,1994; Ivakhnenko,1995;Kondo,1998;Kord´ık et al.,2003;Witczak et al.,2006;Kondo and Ueno,2008).5.41979:Convolution+Weight Replication+Winner-Take-All(WTA)Apart from deep GMDH networks(Sec.5.3),the Neocognitron(Fukushima,1979,1980,2013a)was per-haps thefirst artificial NN that deserved the attribute deep,and thefirst to incorporate the neurophysiolog-ical insights of Sec.5.2.It introduced convolutional NNs(today often called CNNs or convnets),where the(typically rectangular)receptivefield of a convolutional unit with given weight vector is shifted step by step across a2-dimensional array of input values,such as the pixels of an image.The resulting2D array of subsequent activation events of this unit can then provide inputs to higher-level units,and so on.Due to massive weight replication(Sec.2),relatively few parameters may be necessary to describe the behavior of such a convolutional layer.Competition layers have WTA subsets whose maximally active units are the only ones to adopt non-zero activation values.They essentially“down-sample”the competition layer’s input.This helps to create units whose responses are insensitive to small image shifts(compare Sec.5.2).The Neocognitron is very similar to the architecture of modern,contest-winning,purely super-vised,feedforward,gradient-based Deep Learners with alternating convolutional and competition lay-ers(e.g.,Sec.5.19-5.22).Fukushima,however,did not set the weights by supervised backpropagation (Sec.5.5,5.8),but by local un supervised learning rules(e.g.,Fukushima,2013b),or by pre-wiring.In that sense he did not care for the DL problem(Sec.5.9),although his architecture was comparatively deep indeed.He also used Spatial Averaging(Fukushima,1980,2011)instead of Max-Pooling(MP,Sec.5.11), currently a particularly convenient and popular WTA mechanism.Today’s CNN-based DL machines profita lot from later CNN work(e.g.,LeCun et al.,1989;Ranzato et al.,2007)(Sec.5.8,5.16,5.19).5.51960-1981and Beyond:Development of Backpropagation(BP)for NNsThe minimisation of errors through gradient descent(Hadamard,1908)in the parameter space of com-plex,nonlinear,differentiable,multi-stage,NN-related systems has been discussed at least since the early 1960s(e.g.,Kelley,1960;Bryson,1961;Bryson and Denham,1961;Pontryagin et al.,1961;Dreyfus,1962; Wilkinson,1965;Amari,1967;Bryson and Ho,1969;Director and Rohrer,1969;Griewank,2012),ini-tially within the framework of Euler-LaGrange equations in the Calculus of Variations(e.g.,Euler,1744). Steepest descent in such systems can be performed(Bryson,1961;Kelley,1960;Bryson and Ho,1969)by iterating the ancient chain rule(Leibniz,1676;L’Hˆo pital,1696)in Dynamic Programming(DP)style(Bell-man,1957).A simplified derivation of the method uses the chain rule only(Dreyfus,1962).The methods of the1960s were already efficient in the DP sense.However,they backpropagated derivative information through standard Jacobian matrix calculations from one“layer”to the previous one, explicitly addressing neither direct links across several layers nor potential additional efficiency gains due to network sparsity(but perhaps such enhancements seemed obvious to the authors).。
大数据分析与挖掘课后习题参考答案
（2）使用等宽划分时，将其划分为四个箱，16 在第几个箱？
（3）利用等深分箱法，将其划分为 3 个箱，平均值平滑法进行平滑处理，第
2 个箱的取值为多少？
（4）利用等宽分箱法，将其划分为 3 个箱，边界平滑法进行平滑处理，第 2
个箱内数据值为多少？
数据清洗：负责解决填充空缺值、识别孤立点、去掉噪声和无关数据等问
题；
数据集成：负责解决不同数据源的数据变换：将原始数据转换为适合数据挖掘的形式。包括数据的汇总、聚
集、概化、规范化，同时可能需要对属性进行重构；
数据归约：负责搜小数据的取值范围，使其更适合数据挖掘算法的需要。
df=spark.createDataFrame([(Vectors.dense(3.2,1.78,130,6000),),
(Vectors.dense(3.5,1.76,122,7000),),
(Vectors.dense(3,1.73,135,5500),),
(Vectors.dense(2.8,1.80,120,4000),),
model.transform(df).show()
print('MinMax')
miScaler=MinMaxScaler(inputCol='Features',outputCol='Feature_MinMax')
model_=miScaler.fit(df)
model.transform(df).show()
7000
3
3
1.73
135
5500
4
2.8
1.80
120
ANTLR3_handouts
ANTLR 3‣ANother T ool for Language Recognition‣written by T erence Parr in Java‣Easier to use than most/all similar tools ‣Supported by ANTLRWorks‣graphical grammar editor and debugger ‣written by Jean Bovet using Swing‣Used to implement‣“real” programming languages ‣domain-specific languages (DSLs)‣http://www.antlr .org‣download ANTLR and ANTLRWorks here ‣both are free and open source‣docs, articles, wiki, mailing list , examplesANTLR Overview3T erI’m aprofessor at the University of SanFrancisco.JeanI worked with T er as a masters student there.Books‣“ANTLR Recipes”? in the works‣another Pragmatic Programmers book from T erence Parra = 3.14f(x) = 3x^2 - 4x + 2print "The value of f for " a " is " f(a) print "The derivative of " f() " is " f'()list variableslist functionsg(y) = 2y^3 + 6y - 5 h = f + gprint h()The value of f for 3.14 is 19.0188The derivative of f(x) = 3x^2 - 4x + 2 is f'(x) = 6x - 4# of variables defined: 1a = 3.14# of functions defined: 1f(x) = 3x^2 - 4x + 2h(x) = 2x^3 + 3x^2 + 2x - 3ANTLR 3BA15ABsautomatically supplied root nodeKey:provided generated writtenLexerANTLR Documentation17http://antlr .orgGrammar Syntaxgrammar-type ? grammar grammar-name ;grammar-options ? 3 types: lexer, parser and tree;defaults to combined lexer and parserComments use the same syntax as Java.must match the filename with a “.g” extensionRegular expressionsANTLR 3Parse Tree drawn by ANTLRWorks29AST drawn by ANTLRWorksvalue: NAME | NUMBER;NAME: LETTER (LETTER | DIGIT | '_')*;NUMBER: '-'? DIGIT+; // just integers fragment DIGIT: '0'..'9';fragment LETTER: 'A'..'Z' | 'a'..'z';RELATION: '<' | '<=' | '==' | '>=' | '>';TERMINATOR: ';';WHITESPACE: (' ' | '\t' | '\r' | '\n')+ { $channel = HIDDEN; };if (a < 10) {Parse trees show the depth-first order of rules that are matched.to refer to non-unique elements.ANTLR 3ANTLRWorks ...39lexer rule syntax diagramrequesting a grammar checkT ests parse tree creation (not AST)41doesn’t support use of predicates (discussed later)ANTLRWorks Debugger Simple when lexer and parser rulesare combined in a single grammar file‣press Debug;ANTLR 3ANTLRWorks Debugger ...67Some Morelike a gated sem. pred., but only used when syntax is insufficient to choose { target-language-expression }?。
新GRE数学真题(OG+机经)
1.y=2x2+7x – 3x y2.The figure above shows the graph of a function f, defined by f(x)=|2x|+4 for all numbers x. For which of the following functions g defined for all numbers x does the graph of g intersect the graph of f ?A g(x) x–2B g(x) x 3C g(x) 2x – 2D g(x) 2x 3E g(x) 3x – 23.Each employee of a certain company is in either Department X or Department Y, and there are more than twice as many employees in Department X as in Department Y. The average (arithmetic mean) salaryis $25,000 for the employees in Department X and is $35,000 for the employees in Department Y. Which of the following amounts could be the average salary for all of the employees in the company?Indicate all such amounts.A $26,000B $28,000C $29,000D $30,000E $31,000F $32,000G $34,0004.Working alone at its constant rate, machine A produces k car parts in 10 minutes. Working alone at its constant rate, machine B produces k car partsin 15 minutes. How many minutes does it take machines A and B, working simultaneously at their respective constant rates, to produce k car parts?____________minutes5.1)If the dollar amount of sales at Store P was $800,000 for 2006, what was the dollar amount of sales at that store for 2008 ?A $727,200B $792,000C $800,000D $880,000E $968,0002)At Store T, the dollar amount of sales for 2007 was what percent of the dollar amount of sales for 2008 ?Give your answer to the nearest 0.1 percent.___________%3)Which of the following statements must be true?Indicate all such statements.A For 2008 the dollar amount of sales at Store R was greater than that at each of the other four stores.B The dollar amount of sales at Store S for 2008 was 22 percent lessthan that for 2006.C The dollar amount of sales at Store R for 2008 was more than 17 percent greater than that for 2006.6.A certain store sells two types of pens: one type for $2 per pen and the other type for $3 per pen. If a customer can spend up to $25 to buy pens at the store and there is no sales tax, what is the greatest number of pens the customer can buy?A 9B 10C 11D 12E 207.A list of numbers has a mean of 8 and a standard deviation of 2.5. If x is a number in the list that is 2 standard deviations above the mean, what is the value of x ?x=______8.Frequency Distribution for List XNumber 1 2 3 5Frequency 10 20 18 12Frequency Distribution for List YNumber 6 7 8 9Frequency 24 17 10 9List X and list Y each contain 60 numbers. Frequency distributions for each list are given above. The average (arithmetic mean) of the numbers in list X is 2.7, and the average of the numbers in list Y is 7.1. List Z contains 120 numbers: the 60 numbers in list X and the 60 numbers in list Y.Quantity A Quantity BThe average of the 120numbers in list Z The median of the 120 numbers in list Z 9.The figure above shows the graph of the function f in the xy-plane. What isthe value of f ( f (–1)) ?A –2B –1C 0D 1E 210.By weight, liquid A makes up 8 percent of solution R and 18 percent of solution S. If 3 grams of solution R are mixed with 7 grams of solution S, then liquid A accounts for what percent of the weight of the resulting solution?A 10%B 13%C 15%D 19%E 26%11.Of the 700 members of a certain organization, 120 are lawyers. Two members of the organization will be selected at random. Which of the following is closest to the probability that neither of the members selected will be a lawyer?A 0.5B 0.6C 0.7D 0.8E 0.912.Line k lies in the xy-plane. The x-intercept of line k is –4, and line k passes through the midpoint of the line segment whose endpoints are (2, 9) and (2, 0). What is the slope of line k ?Give your answer as a fraction._______13.In the course of an experiment, 95 measurements were recorded, and all of the measurements were integers. The 95 measurements were then grouped into 7 measurement intervals. The graph above shows the frequency distribution of the 95 measurements by measurement interval.Quantity A Quantity BThe average (arithmetic The median of the 95mean) of the 95 measurementsMeasurements14.The random variable X is normally distributed. The values 650 and 850 are at the 60th and 90th percentiles of the distribution of X, respectively.Quantity A Quantity BThe value at the 75th750percentile of thedistribution of X15.If 1+x+x2+x3=60, then the average (arithmetic mean) of x, x2, x3, andx4 is equal to which of the following?A 12xB 15xC 20xD 30xE 60x16.Parallelogram OPQR lies in the xy-plane, as shown in the figure above. The coordinates of point P are (2, 4) and the coordinates of point Q are (8, 6).What are the coordinates of point R ?A (3, 2)B (3, 3)C (4, 4)D (5, 2)E (6, 2)17.Let S be the set of all positive integers n such that n2 is a multiple of both24 and 108. Which of the following integers are divisors of every integer nin S ?Indicate all such integers.A 12B 24C 36D 7218.The range of the heights of the female students in a certain class is 13.2 inches, and the range of the heights of the male students in the class is 15.4inches.Which of the following statements individually provide(s) sufficient additional information to determine the range of the heights of all the students in the class? Indicate all such statements.A The tallest male student in the class is 5.8 inches taller than the tallest female student in the class.B The median height of the male students in the class is 1.1 inches greater than the median height of the female students in the class.C The average (arithmetic mean) height of the male students in the class is 4.6 inches greater than the average height of the female students in the class.19.A random variable Y is normally distributed with a mean of 200 and a standarddeviation of 10.Quantity A Quantity BThe probability of the event 1/6that the value of Y isgreater than 22020.In a graduating class of 236 students, 142 took algebra and 121 took chemistry. What is the greatest possible number of students that could have taken bothalgebra and chemistry?__________students21.The total amount that Mary paid for a book was equal to the price of the book plusa sales tax that was 4 percent of the price of the book. Mary paid for the book with a $10 bill and received the correct change, which was less than $3.00. Which of the following statements must be true?Indicate all such statements.A The price of the book was less than $9.50.B The price of the book was greater than $6.90.C The sales tax was less than $0.45.22.Which of the following statements individually provide(s) sufficient additional information to determine the area of triangle ABC above? Indicate all such statements.A DBC is an equilateral triangle.B ABD is an isosceles triangle.C The length of BC is equal to the length of AD.D The length of BC is 10.E The length of AD is 1023.The fabric needed to make 3 curtains sells for $8.00 per yard and can be purchased only by the full yard. If the length of fabric required for eachcurtain is 1.6 yards and all of the fabric is purchased as a single length,what is the total cost of the fabric that needs to be purchased for the 3curtains?A $40.00B $38.40C $24.00D $16.00E $12.8024.The fabric needed to make 3 curtains sells for $8.00 per yard and can bepurchased only by the full yard. If the length of fabric required for each curtain is 1.6 yards and all of the fabric is purchased as a single length,what is the total cost of the fabric that needs to be purchased for the 3 curtains?A $40.00B $38.40C $24.00D $16.00E $12.8025.In the xy-plane, the point with coordinates (–6, –7) is the center of circle C. The point with coordinates (–6, 5) lies inside C, and the point with coordinates (8, –7) lies outside C. If m is the radius of C and m is an integer, what is the value of m ?m =_____26.What is the least positive integer that is not a factor of 25! and is not a prime number?A 26B 28C 36D 56E 5827If 1 is expressed as a terminating decimal, how many nonzero digits1/[(211)(517)]will the decimal have?A OneB TwoC FourD SixE Eleven28.Eight hundred insects were weighed, and the resulting measurements, in milligrams, are summarized in the boxplot below.(a) What are the range, the three quartiles, and the interquartile range of the measurements?(b) If the 80th percentile of the measurements is 130 milligrams, about how many measurements are between 126 milligrams and 130 milligrams?29.The figure shows a normal distribution with mean m and standard deviation d, including approximate percents of the distribution corresponding to the six regions shown.Suppose the heights of a population of 3,000 adult penguins are approximately normally distributed with a mean of 65 centimeters and a standard deviation of 5 centimeters.(a) Approximately how many of the adult penguins are between 65 centimeters and 75 centimeters tall?(b) If an adult penguin is chosen at random from the population, approximately what is the probability that the penguin’s height will be less than 60 centimeters? Give your answer to the nearest 0.05机经题：8月5选11. 3个数的平均数是30，把这3个数翻倍之后也加在这个list中。
argparse用法choices用法
Argparse用法及choices用法在Python编程中，argparse是一个用于解析命令行参数的模块，非常有用而且易于上手。
本文将介绍argparse的基本用法，并重点讨论choices参数的使用。
一、argparse的基本用法1. 导入argparse模块要使用argparse，首先需要导入它。
在Python中，可以使用以下语句进行导入：```pythonimport argparse```2. 创建ArgumentParser对象接下来，需要创建一个ArgumentParser对象。
这个对象将帮助我们定义命令行参数及其属性。
```pythonparser = argparse.ArgumentParser(description='Description of the program')```在上面的代码中，description参数用于描述程序的功能，将在打印帮助信息时显示出来。
3. 添加命令行参数通过add_argument方法，可以向ArgumentParser对象添加命令行参数。
该方法的常用参数包括name或flags、type、help等。
```pythonparser.add_argument('name', type=str, help='Help message for the parameter')```在上面的代码中，name是参数的名称，type用于指定参数的数据类型，help是在打印帮助信息时显示的帮助消息。
4. 解析命令行参数使用parse_args方法解析命令行参数，并将其存储在一个命名空间对象中。
这样就可以方便地使用这些参数了。
```pythonargs = parser.parse_args()```二、choices的使用choices参数用于限定命令行参数的取值范围。
当设置了choices参数后，用户只能输入预先指定的值，否则将抛出错误。
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
34Implicit Arguments RAJESH BHATT AND ROUMYANA PANCHEVA
1Introduction2Implicit arguments in passives and middles2.1Implicit arguments in passives2.2Implicit arguments in middles3Implicit arguments of nouns3.1Optionality of implicit arguments of nouns3.2Control of implicit arguments of nouns3.3Differences between implicit arguments of nouns and passives3.3.1Controlling implicit arguments of passives3.3.2Differences in Control by implicit arguments of nouns and passives4Null objects5Implicit arguments of evaluative predicates6Arguments of modals7Conclusion
1Introduction Reference to non-overt arguments has been made in the description of a widerange of syntactic phenomena. Some of them (PRO, pro , A/A ′ -traces) are rela-tively well understood and there exists a certain consensus regarding their ana-lysis. There is another class of non-overt arguments, often referred to as implicitarguments, for which no such consensus prevails. Implicit arguments do notseem to form a unified class. To appreciate this, let us examine some cases whichhave been argued to involve implicit arguments:
(1) Implicit agents of passives (vs. middles and unnacusatives): a.This ship was sunk [PRO to collect the insurance]. (passive)b.# This ship sank [PRO to collect the insurance]. (unaccusative)c.*This ship sinks easily [PRO to collect the insurance]. (middle)
BCT2C34.fm Page 554 Monday, August 15, 2005 9:58 AM Bhatt and Pancheva: Implicit Arguments 555(2) Implicit arguments of nouns: a.the negotiations [PRO to achieve a peaceful settlement]b.the use of drugs [PRO to fall asleep]c.the playing of the game [PRO to prove a point]
(3) Null objects (cf. Rizzi 1986a):
‘This leads (people) to the following conclusion.’‘This leads people to conclude what follows.’(4) Implicit arguments of adjectives (from Roeper 1987a): a.It is necessary/*inevitable [PRO to go].b.It is wise/*probable [PRO to go].
(5) The bearer of the obligation of a deontic modal: a.The books can be sold [without PRO reading them].(from Chomsky 1982 via Williams 1985)b.*The books might have been sold [without PRO reading them].(from Kratzer 1991)
(6) Implicit agents of agentive suffixes (e.g., -able ): Goods are exportable [PRO to improve the economy].
The above list includes the implicit agent of a passive (section 2), the implicitargument of a noun (section 3), null objects (section 4), the implicit argument ofan adjective (section 5), the bearer of the obligation argument of a deontic modal(section 6), and the implicit agent associated with agentive suffixes like -able . 1 What unifies this class? It is felt that all of these examples involve a missingnominal element. The evidence for this missing nominal element comes from thefact that (1–3) all involve an infinitival with a PRO subject. Something, it isargued, must be controlling these PROs. There is no NP argument in the relevantstructures that could be doing so. The element held responsible for Control is theimplicit argument.In principle, null subjects (PRO, pro ) could have been called implicit argu-ments, given that they are non-overt and indisputably arguments. Furthermorethere have been analyses in the literature according to which PRO/ pro are notsyntactically expressed (for PRO see Partee and Bach 1980; Chierchia 1984b; Kleinand Sag 1985, among others; for pro see Alexiadou and Anagnostopoulou 1998,etc.). However, by convention, PRO/ pro are not grouped together with the casesof implicit arguments in (1–3). This is why in this chapter we do not discussPRO/ pro .
a.Questoconduce(la gente)allaseguenteconclusione.thisleadsthe peopleto-thefollowingconclusion
b.Questoconduce(la gente)a[PROconcluderequantosegue].thisleadsthe peopletoconcludewhatfollows
BCT2C34.fm Page 555 Monday, August 15, 2005 9:58 AM 556 Bhatt and Pancheva: Implicit Arguments From its inception, the literature on implicit arguments has defined them assyntactically active elements that nevertheless do not occupy a syntactically pro-jected position. 2 Consider, for example, the following definitions for implicitarguments that have been proposed in the literature:
(7)mplicit arguments are not the mysterious shadowy presences they aresometimes made out to be. They are really nothing more than the argumentslots in the argument structure . . . A ‘weak’ theta-criterion is all that isneeded to give implicit arguments, since these are nothing more thanunlinked argument roles.(Williams 1985: 314)