注：不含主觀題第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】練習(xí)題第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】第2題ABCD第3題ABCD第4題ABC第5題ABCD第6題ABCD第7題ABCD第8題ABCD第9題ABCD第10題ABCD第11題ABCD【作業(yè)須知】第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】作業(yè)第1題Whatisthewrongclaim?AAlocalminimizerisalsoaglobalminimizerforconvexoptimization.BAnisolatedlocalminimizerisalsoastrictlocalminimizer.CAglobalminimizerisalsoalocalminimizerforunconstraintoptimization.DAstrictlocalminimizerisalsoanisolatedlocalminimizer.第2題Alocalminimizers.Bstationarypoints.Cglobalminimizers.Disolatedlocalminimizers.第3題Whichofthefollowingclaimsisnotcorrectforunconstrainedconvexoptimization?AThefirstordernecessaryconditionisalsothesufficientconditionforxtobeaminimizer.BIfxisalocalminimizer,itisalsoaglobalminimizer.CIfxisalocalminimizer,itisalsoastrictlocalminimizer.DAnyconvexcombinationoftwoglobalminimizersisalsoaglobalminimizer第4題AtrustregionstrategyBexactlinesearchCinexactlinesearchDsteepestdescentdirection第5題ABCD第6題AlinearBsublinearCsuperlinearDquadratic【作業(yè)須知】第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】作業(yè)第1題Whichofthefollowingsteplengthsisnottheinexactsteplength?AbacktrackingBWolfeconditionCstrongWolfeconditionD第2題WhichconditionsaretheWolfeconditions?(multiplechoices)ABCD正確答案：AB第3題Steepestdescentmethodwith________linesearchisgloballyconvergent.(multiplechoices)AWolfeBstrongWolfeCGoldsteinDbacktracking正確答案：CBA第4題Undercertainconditions,theconvergencerateforNewton’smethodisAlinearBsublinearCsuperlinearDquadratic第5題TheconvergencerateforsteepestdescentmethodundercertainconditionsisatbestAlinearBsublinearCsuperlinearDquadratic第6題Theremaybe________phenomenonforsteepestdescentmethodtowardsolutions.AZigzagBfastconvergenceCellipsoidsDincreasing第7題ABCD【作業(yè)須知】第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】作業(yè)第1題Whichisnotthekeypointintrustregionmethods?A

thesizeoftrustregionBthemodelfunctionmkCthesteppkDsufficientdecrease第2題Whichmethodscanbeusedtosolvethesubproblemoftrustregionmethod?(multiplechoices)A

DoglegmethodB

Two-dimensionalsubspaceminimizationC

CauchypointD

KKTcondition正確答案：ABC第3題ABCD第4題ABCWemaynotmakegreatprogressatthecurrentiteration.DWemaymakegreatprogressatthecurrentiteration.第5題AshrinkthetrustregionradiusBstayatthecurrentiterationCreproducethemodelfunctionandsolvethesubproblemwithinasmallertrustregionDenlargethetrustregionradius正確答案：CBA第6題TheCauchypointcanbeviewedasAsteepestdescentmethodwithespecialsteplengthBDoglegmethodCNewton’smethodDtwo-dimensionalsubspaceminimizationmethod第7題Indoglegmethod,whichdirectionisnotinvolved?AsteepestdescentdirectionBNewton’sdirectionCthecombinationofsteepestdescentdirectionandNewton’sdirectionDconjugatedirection第8題AlinearlyBsublinearlyCsuperlinearlyDquadratically【作業(yè)須知】第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】作業(yè)第1題Any

set

of

vectors

satisfying

conjugacy

is

also

.

Alinearly

independent

BorthogonalClinearly

dependent

第2題Which

of

the

following

statement

is

not

correct?ABCWhen

applied

to

solving

quadratic

convex

optimization

problem,

the

different

versions

of

nonlinear

conjugate

gradient

methods

are

reduced

to

the

same

linear

conjugate

gradient

method.

DWhen

applied

to

solving

quadratic

convex

optimization

problem,

the

different

versions

of

nonlinear

conjugate

gradient

methods

are

still

different.

第3題ABCD第4題For

linear

conjugate

gradient

method

with

exact

line

search,

which

is

not

true?ABCD第5題AKrylov

subspace

Bhyperplane

Corthogonal

subspaceDrange

space

第6題ABCD第7題ABCD第8題ABCDifferent

variants

including

FR

method,

PRP

method

lead

to

different

iteration

sequence.

DDifferent

variants

including

FR

method,

PRP

method

can

terminate

within

finite

steps.

【作業(yè)須知】第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】作業(yè)第1題ABCD第2題ABCD第3題AItisstronglysemismooth.BItissemismooth.CItisapiecewiselinearfunction.DItisasmoothfunction.第4題ABCD正確答案：BAC第5題Supportvectormachinecanbedividedinto_______ASupportvectorclassificationandsupportvectorregressionBSupportvectorclassificationandclusteringCSupportvectorregressionandclusteringDDeeplearningandreinforcelearning第6題Thetraditionalmethodforsupportvectorclassificationinclude________(multiplechoices)ATRONBDCDCCoordinatedescentmethodfortheprimalproblemDModifiedNewton’smethod正確答案：DCBA第7題OnecanapplysemismoothNewton’smethodtosolve_________(multiplechoices)AL2-lossSVCBε-L2SVRCL1-lossSVCDLogistic-lossSVC正確答案：BA第8題WhenappliedglobalizedversionofsemismoothNewton’smethodtosolveL2-lossSVC，wehavethefollowingresults________(multiplechoices)AThesequenceconvergesgloballytotheoptimalsolutionofL2-lossSVCBThelocalconvergencerateisquadraticCEverymatrixintheClarke’sgeneralizedJacobianispositivedefiniteDTheobjectivefunctioninL2-lossSVCisstronglysemismooth正確答案：DCBA【作業(yè)須知】第1題【注意】單擊題目圖片即可放大查看【此題不計(jì)入成績】作業(yè)第1題

Whichofthefollowingstatementisnotcorrectforconstrainedoptimization?A

Constraintsmaymakeitverydifficulttofindalocalsolution.B

Constraintsmayalsomakeitdifficulttoidentifylocalsolutionsandglobalsolutions.C

Isolatedlocalsolutionsarestrictlocalsolutions.D

Strictlocalsolutionsareisolatedlocalsolutions.第2題ABCD第3題ABC第4題ABC第5題ABCD第6題Iflinearlyindependentconstraintqualificationholdsatx,thereisAThelinearizedfeasibledirectioncoincideswiththetangentconeatxwithrespecttofeasibleset.BThecriticalconeisthesameasthelinearizedfeasibledirection.Cotherconstraintqualificationsalsohold.DThefirstorderoptimalityconditionshold.第7題A{1,2}B{1,3}C{2,3}D{1,2,3}第8題ABCD第9題Whatistherelationbetweenlinearizedfeasibledirectionandcriticalcone?Acriticalconeisasubset