Selected Topics CSE 891
Popular in Course
Popular in Computer Science and Engineering
This 14 page Class Notes was uploaded by Donnell Kertzmann on Saturday September 19, 2015. The Class Notes belongs to CSE 891 at Michigan State University taught by Laura Dillon in Fall. Since its upload, it has received 15 views. For similar materials see /class/207406/cse-891-michigan-state-university in Computer Science and Engineering at Michigan State University.
Reviews for Selected Topics
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/19/15
Computing research CR Research Oxford Concise Dictionary is la the systematic investigation into and stu NauweofCompmmg Resear ch CSE 891 Sec 4 8809 httpwwwcsemsued udioncse891 L mimosuaxm swam A view of m quotmum mummy hwy ungmma l A Culum From Gordana Dotting mm A Seismic Methods in Co 5 Sec 5 mkovm mpulel Science L mien csEaet mg 2009 3 the classic sciences Three main groups 0 Logic amp math 1 0 Natural sciences 2 0 Social sciences 3 Broader context 0 Connections to humanities 4 O lmpregnated by cultural O Sciences have specific areas of validity 0 Inner regions are prerequi sites for outer ones 0 Influence decreases with distance f aterials sources etc in order to establish facts and reach new conclusions 1b an endeavourto discover new or collate old facts etc by the sci entific study of a subject or by a course of critical investigation Computing encompasses Phenomena related to computersinformation Use of computersinformation processing across the spectrum Dijkstra said that to call the eld Computer Sciencequot is like calling surgery Knife Sciencequot 4 A view of the classic sciences OBJECTS MAJOR METHOD SClENCEDlSClPLlNE proposltlo u b rs Logic amp math mathematical structures Deduction physical bodies elds and interactions living or Scienti c method ganisms Natural science human individuals Scienti c method Social science groups societies Hermeneutics man ideas actions and Hermeneutics h relationships language Humanities Fr D g kaovic Scienti c Methods in Computer Science L nimcsstamsge 45pnngzooe Implications ltgtHow does the field of computing fit into this view ltgtWhat does this view imply about the nature of computing research CR L Dillon CSE s01Sep A Spring 2000 Scientific method The logical scheme used in searching for answers to questions posed by science Le in developing new theories L Dillon CSE s01Sep A Spring 2000 Computer science differs so basically from the other sciences that it has to be viewed as a new species among the sciences and it must be so understood Computer science deals with information its creation and processing and with the systems that perform it much of which is not directly restrained and governed by physical laws The failure of computer science to conform to the paradigms of physi cal sciences is often interpreted as immaturity of computer science This is not the case since theory and experiments in computer science play a different role than in physical sciences computer science has the potential to give deep new insights and quantitative understanding of the computing paradigm and our intellec tual processes and thus just maybe a possibility to grasp the limits of the knowable From Juris Hartmanis Turing Award Lecture on Computa tional Complexity and the Nature of Computer Science L Dillon CSE 891SecA Spring 2000 0 Scientific method The logical scheme used in searching for answers to questions posed by science Le in developing new theories pose a question L Dillon CSE 891SecA Spring 2000 5 Scientific method The logical scheme used in searching for answers to questions posed by science ie in developing new theories Deduce consequences iolmulate a make plemclions pose a question L automcsEaeksec 45Dring2009 Scientific method The logical scheme used in searching for answers to questions posed by science ie in developing new theories pose a M W lepose ine question Selem among Consistency achieved competing theolies From Gordana Dodigrcrnkovi L automcsEaeksec 45Dring2009 Seenime Methods In Compute Science n Scientific method The logical scheme used in searching for answers to questions posed by science ie in developing new theories meme consequences iolmulate a make hypothesis pledldlons pose 3 mmquot lepose the question L BlamesEaeksee 45pnnezooe Scientific method e OKnowledge is always changing P 39 Igj It is subjectto continual examination and W selfcorrection OThe logic of science is recursive Each iteration starts with a hypothesis based in the preexisting body of knowledge OThe method is impartial A theory is accepted based on results obtained through logical reasoning observations andor experiments The results have to be reproducible If the original claims are not verified the causes of discrepan cies are exhaustively studied L BlamesEaeksee 45pnnezooe Science and technology Aristotle s Science contra Technology Science Technology Object unchangeable changeable Principle of motion inside outside End knowing the general knowing the concrete Activity theoria end In poiesIs end In something Itself else Method abstraction medalmg comma complex innovation form discovery invention Type of result law mke rulelike statements statements Time perspective longterm shortterm L DilionCSE89tSec AtSDrIHQZOOS 13 The dialectic of research Types of contributions to logical debate in a eld 0 Thesis present the original statement of an idea g tnes signs early in the development pioeessquot Fuchs 1992 O Antithesis present an argument to challenge a previous thesis e g entitnesis clients often choose reetuies of the prototyping environment Without eonsi genng possible eiteineti yesquot Hayes and Jon es 1989 0 Synthesis form a new argument from existing sources Computing and technology The computing discipline is more complex and hetero geneous than disciplines of Aristotle s time Computer Science does not deal merely with computer use technology or software program and data It encompasses abstract mathematical thinking and includes an element of engineering What is speci c for CS is that its objects of investigation are artifacts computerrelated phenomena that change concurrently with the development of theories describ ing them andwith the growing practical experience in their usage L uipmssaotsee 45piiiigzoos 14 16 Modeling Basis for all research Modelingabstraction always occurs in science 0 Description ofa phenomenon in a symbolic language 0 Accounts for relevant featuresomits irrelevant ones 0 Enables prediction of observablemeasurable consequences of changes to a system eg synthesis environmentsquot Gravell and Henderson 1996 From chiisiohhson Whalis lesealch In computing solence7 L niiioiipssaoxsee Spinoza Is REAL WORLDquot MODEL Pro ram ASG CFG call graph 9 function tracefailures set Computer hardware simulation Database relation L Dimesnnsss Summons 6 Questions for modeling 0 How to model What to take into accountneglect What formalism to use in modeling 0 Does the model serve its purpose Do we have the FigurciiMaclmg right level of abstraction O In what ways do the featuresbehavior of the model differ from what is expected 0 How does the model differ from reality or from other models or experimental results ltgtAre the results valid L DillonCSE89lSec 45Dring2009 w Proof by demonstration Build an artefact that stands as an example for a more general class of solutions Olterative re nement to move closerto a good solution 0 Demonstrate feasibility proof of conceptquot OLearn practical limitations 0 High risk may ignore formation of hypothesis artefact may fail before you reach any conclusions difficult to know what conclusions to draw L DillonCSE89lSec 45Dring2009 l9 Forms of argument in CR ltgtDemonstration ltgtEmpiricism OMathematical proof OHermeneutics From Chris Johnson What is research in computing science is L mimcssaeissc 45Dnng2009 E m pi l lClS m Grounded in scienti c method OGenerate hypothesis devise method collect data draw conclusions eg find performance of new algo on accepted benchmarks eg determine effectiveness of a test method at finding faults OAttractions impartiality repeatability delineation of variablescontrols OChallenges probabilistic measures are often inappropriate experimentalactual operating conditions may affect outcomes difficult to know when a set of trials is sufficient eg when to stop testing L mimcssaeissc 45Dnng2009 2o Generating ideas CSE 891 Sec 4 8809 httpwwwcsemsuedulldillonlcse891 L Dillon CSE 5913 A Spring 2009 1 Obstacles to creativity Standard mindset Connect the dots by drawing four straight continuous lines and never lifting the pencil from the paper Try to Think outside the box 0 Challenge assumptions will they still be reasonable 1025 50 years from now 0 Look at same problem from different angle 0 Try a completely different approach Discovery consists of looking at the same thing as everyone else and thinking something different Albert Szent Gyorgi Obstacles to creativity Standard mindset 39 Connect the dots by drawing four straight continuous lines 39 and never lifting the pencil from the paper 0 C I Obstacles to creativity Fear of failure or ridicule may keep you from O Pursuing risky ideas 0 Challenging common assumptionsasking obvious questions 0 Suggesting a radically different approach And the trouble is if you don t risk anything you risk even morequot Erica Jong Try to Have confidence in your ideas If at rst the idea is not absurd there is no hope for itquot Albert Einstein The silly question is the rst intimation of some totally new developmentquot Alfred North Whitehead Try to Develop a thick skin Professor Goddard does not know the relation between action and reaction and the need to have something better than a vacuum against which to react He seems to lack the basic knowledge ladled out daily in high schoolsquot 1921 New York Times editorial about Robert Goddard s revolutionary rocket work Obstacles to creatIVIty Impatience fixing too quickly on a solution 0 It takes work and time to internalize difficult problems Go some distance away because the work appears smaller and more of it can be taken in at a glance and a lack of harmony of proportion is rapidly seenquot Leonardo da nci 0 Once you invest in a solution you can become to blind possibly better to alternatives Try to Systematically consider alternatives 0 morphological analysis 0 comparative analysis Try to Leave a problem and revisit it periodically Try to Come at it from varied perspectives Try to Be open to competing views no place for ego Obstacles to creativity Environment can get in the way 0 Too many interruptions or distractions make it difficult to get into a state in which ideas flow 0 Need a comfortable and welcoming workplace Try to Protect sufficient blocks of time for research Try to Arrange workplace to your liking 0 Do you work best in silence with music other noise 0 Do you need a blank wall photos inspiring quotes Try to Isolate your workplace from social and other distractions Obstacles to creativity Can t tolerate uncertainty and ambiguity ltgtSolution of a complex problem is a messy process ltgtMust usually wallow in misleading and illfitting data hazy and difficulttotest concepts opinions etc ltgtProblemsolving is bringing order to chaos the ability to tolerate chaos is a must Try to ltgtdocument issues that bother you and move on ltgtreturn to them later as other issues that they depend on are resolved Things you can do No one can teach you research creativity We learn to generate promising ideas by ltgtObserving how others generate ideas PhD study is essentially an apprenticeship with your adviser As you learn new ideas ask what gave the researcher the ideas 0 Becoming expert enough to internalize key issues behind problems develop intuition amp see opportunities There are activities that can help and prerequisites for success L Dillon CSE 591SecA Spring 2009 5 Achieve basic competency If you are constantly working on just getting byquot your mind won t find the spare cycles to be creative Put in the time to master the mechanics 0 Math and statistics possibly another sciences 0 Programming skills 0 Humancentered skills 0 Technical skills familiarity with MSU computing system facility with word processors drawing programs etc ability to search library and other available resources Also need expertise in your research areas L Dillon CSE s01Sep A Spring 2000 0 Use questions lists 0 Put to other uses ltgtWhat s similar 0 Modify ltgt Magnify Minimize 0 Substitute ltgt Rearrangetranspose ltgt Reversenegate 0 Combine ltgt Etc L Dillon CSE 5013 A Spring 2000 11 Creative browsing Consciously look for inspiration by ltgtSkimming research papers in top conferences can you think how to approach limitations open questions ltgtBrowsing web sites of leading researchers ltgtReading calls for proposals from funding agencies 0 Reading survey articles provide an overview of a field and discussion of open problems ltgtAttend research seminars ltgtTake topics course ltgtDiscussion with colleagues L Dillon CSE s01SepA Spring 2000 i0 Morphological analysis Discover relationships betw dimensions of a multi dimensional problem ltgtList different attributes issues facets of the problem ltgtBelow each list as many alternatives as you can ltgtLook at and assess different combinations L Dillon CSE s01SepA Spring 2000 12 Comparative analysis Systematically compare ideas partial solutions strategies etc ltgtList different attributes issues qualities as columns of a matrix ltgtList competing ideas partial solutions strategies etc as rows ltgt Rate each row with respect to each column L Dillon CSE s01Sep A Spring 2000 13 15 Work from concrete example ltgtConsider a simplified instance of a messy problem ltgtUse it to clarify issues and document ideas ltgtSolve the simple instance and then try to generalize the solution ltgtBe prepared to throw outquot the solution as you scale up the instance and discover better ones L Dillon CSE s01Sep A Spring 2000 15 Combining good ideas ltgtlf approaches have complimentary strengths amp weak nesses can they be effectively combined ltgtCan approaches that are effective in one domain be applied to problems in another domain ltgtGo wild Force yourself to connect two seemingly un related ideas somehow L Diiipn CSE s01SepA Spring 2000 1A Draw analogies ltgtEstablish an abstract sense in which your problem or situation is like some other one ltgtThen list the attributes of the metaphorical version and make correspondences ltgtCan approaches that yield results in the metaphori cal domain translate effectively to your problem domain L Dillon CSE s01SepA Spring 2000 16 Critical reading CSE 891 Sec 4 8809 httpwwwcsemsueduldilloncse891 L Diiion CSE s01Sep A Spring 2000 Quick overview First build a context for understanding Read ltgtTitle abstract amp introduction ltgtSection headings 0 Conclusions Skim ltgtTables amp graphs and their captions ltgtDefinitions and theorems ltgtRelated work section and bibliography L Diiion CSE s01Sep A Spring 2000 m depends on why Reasons for reading research papers include ltgtQuick overview of a new methodideatechnique approach or of a new areafield ltgtPresent a paper in a seminar course ltgtApplyimproveevaluate a specific idea or result ltgtUnderstand relationship between your work and that of others ltgtReview a paper for a conference or journal L Diiion CSE s01SepA Spring 2000 2 Reading for basic understanding Next read paper from top to bottom ltgtHighlight importantilluminating passages 0 Mark areas with questionsconcerns ltgtTry not to get bogged down in details yet Jot down preliminary thoughts in your words ltgtWhat is the goal of the research ltgtWhy is the goal importantdifficult ltgtWhat is the key idea behind the solution ltgtWhat are main benefitsdrawbacks of the solution ltgtHow might ideas be improved ltgtHow do they relate to your interestsideaswork L Dillon CSE 891SecA Spring 2000 A Reading for deeper understanding Comprehension questions Re read to answer questionsfill in detail ltgtWhat is the goal 0 May first need to fill some holes read cited Why 393 th39s 903 39mportant What are the main difficulties in achieving it papers technical reports the3is textbook OWh t th I d I b I 9 ltgtConstruct an example to clarify ideasmake them a are e Qalme Con 7 u IOnS39 A new or improved finding or method concrete Better understanding of a complex phenomenon or of other OH in details missing from the paper reSU39tS i an area I I I ltgtWhat are the conclusions Add torefine your notes from preVIous times What lessons can you take away from the paper How might you apply these lessons elsewhere Evaluatechallenge what you have read ltgtWhat is the support for the conclusions By what method were the conclusions validated What assumptionssimplifications were made L Dillon CSE 8913e0 A Spring 2000 5 L Dillon CSE 501 Sec A Spring 2000 7 8 Evaluation questions DeCIding what to read ltgtAre the assumptionsmodels reasonable ltgtlnformation overload too many conferences jour Do they reflect reality nals magazines technical forums special interest Has the problem been simplified awayquot groups technical archives ltgtAre the methods for drawing conclusions sound OHighly variable quality re39eVance importance S the ogic correct 0 Journal papers relatively complete mature presentation have been vetted by experts Textbooks survey articles may give more accessible presenta Do authors measure what they claim using appropriate controls testing procedure statistical analyses etc ltgtAre the condusmns userISurprismg tions provide broader picture of context amp contributions Do they generalizescalesolve a real problem Colleagues other papers can point you to seminal papers Has the wheel been improved Is the delta sufficient 0 For bleeding edge topics Search relevant conference proceedings ltgtWhat are the threats to validity other plau5ible explana Askfor pointers from experts tions for observed effects Search libraries internet but be discriminating L Dillon CSE 8913e0 A Spring 2000 7 L Dillon CSE 501 Sec A Spring 2000 Who am I Laura Dillon Research Procedures come 3115 EB 0 Office hours Tu Th 200330 pm In 0 Email ldilloncsemsuedu I 0 Phone 3534387 Computer Scrence Education 0 BS MS amp ABD in math University of Michigan 1974 1976 amp 1978 CSE 891 Sec 4 8809 0 MS amp PhD In CS University of Massachusetts 1981 amp 1984 httpwwwcsemsueduldilloncse891 Teaching positions 0 Instructor Eastern Montana College in Billings 197879 0 Visiting professor UMASS Amherst 198485 0 Professor University of California Santa Barbara 198597 0 Professor MSU 1997now L Dillon CSE 8913e0 A Spring 2009 l L Dillon CSE 591 Seo A Spring 2009 2 Who am I Who are you Research interests 0 Formal modeling specification amp analysis of concurrent systems 0 Design programming amp verification of concurrent systems 0 Effects of concurrency on design and maintenance activities Professional experience 0 Past editor of TSE TOSEM CACM ACM SIGSOFT Exec Comm 0 Programorganizing committees ICSE ISSTA FSE ICDCS FASE ICFEM GHC SEES Doc Symp CRAW career development MICWIC OAdvisoryfunding panels for NSF NASA ACM l have a life 0 Husband is professor in FampW Department 0 Son works as software developer 0 Daughter aspiring artist motion amp paying off college loans 0 Lots of outside interestssports scuba X skiing jogging L Dillon CSE 8918eo A Spring 2009 3 L Dillon CSE 591 Seo A Spring 2009 A Goals Objectives Intended audience Learn about different paradigms for research in CS ostudents in ea y stages of PhD research ltgtWhat paradigms are most appropriate for your research area 0 How are results evaluated in each An introduction to the basic concepts and ltgtWhat ingredients go into good research in each skills involved in computer science research Deve39op your reseamh quottastequot ltgtHelp for successful completion of PhD Read and critically evaluate technical papers OHGIP for SUCCGSSfUl career in Exercise creativity in generating new research ideas academia Develop technical communication and writing skills national research laboratory industrial research laboratory Provrde an Indepth research experience any creative technological endeavor Become aware Of standards Of research ethics L Dillon CSE 891SeoASpnng2009 a L DillonCSE 891SecA Spring 2000 Syllabus Course reqUIrements Jan 20 Introduction Research presence Jan 27 Nature of CS research Feb 03 Reading assessing literature Recognizing good ideas Feb 10 Generating ideas Feb 17 Empirical methods Generating new ideas Feb 24 Writing goals venues Mar 03 Writing content organization Initial report Mar 10 SPRING BREAK Mar 17 Reviewing amp refereeing Mar 24 Research presentation Interim report Mar 31 Writing workshop style form Apr 07 Research ethics Apr 14 Time management Draft slides Apr 21 Great research Hamming Apr 28 Open Final report May Miniconference Research presentation L Dillon CSE saiSep A Spring 2000 7 L Dillon CSE 8918ecA Spring 2000 A PhD What Why ltgtThe extended process 5 years Prepares you to do research Is a onceinalifetime opportunity to become an expert ltgtCertifies that you are capable of conducting research ltgtNeeded to join some communities L Dillon CSE s01Sec A Spring 2000 0 Research ltgtThe creation of knowledge ltgtDiffers significantly from taking classes A producer S a consumer of knowledge L Dillon CSE 5013 A Spring 2000 11 APhD is NOT ltgtLucrative At least not immediately Certainly not sufficient or necessary ltgtA chance to take more classes ltgtA meanwhile activity ltgtWell defined No assignments and checklists If you only do what your advisor asks and no more you will have missed the point of the PhD L D1llonCSE s01SecA Spring 2000 10 Desired end state ltgtA successful career Ability to have real impact A lifetime of learning and advancing knowledge A job you love Flexibility to pursue what interests you ltgtHigh quality research is a prerequisite You will be evaluated on your publication record and research contributions noton your dissertation or course grades Dissertation is a minimal requirementthink BIG ltgtGood reading A PhD is Not Enough L D1llonCSE s01SecA Spring 2000 12
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'