IFICIAL INTELLIGENCE July = 2021 Paper Code:PCC-CSE-304-G sion No. Vis compulsory. AI! questions carry equal marks. Q.1.(a) Differentiate between informed and uninformed search. 8) ‘Ans. ¥ Informed Search d Uninformed Search “ruses knowledge for the searching Tt doesn’t knowledge for searching, ‘process. process, Itfinds solution more quickly. It finds solution slow as compared to informed search. Ttmay or may not be complete. Tt is always complete. Cost is low. Cost is high. R . 2 Tt consumes moderate time. t providesthe diregtion regarding the ‘No suggestion is given regarding the solution kf & solution. ug _ | init : Ttis more lengthy while implementation. Bx: : Depth’fitst search, Breatdh first search Q.1.(b) Write a short note on Semantic Network. @) Ans, Semantic Nets > Itis another form of representing the knowledge. ‘The major idea is that : 4 ~The meaning ofa concept comes from its relationship with other concepts. ~The information is stored by interconnecting nodes with labelled arcs. Representation in a Semantic Net The physical attributes of a person’can be represented as in Fig.(a). ; These values can also be represented in logic as: isa (person, mammal), instance(Mill- ‘all person) team(Milk-Hall, Cardiff). | Features of Sementic Network + : i eee |. Sementic nets were used was to find relationships among objects by spreading acti- ion out from each of two nodes & seeing where the activation net. This process is called ‘section search, 2. Itis a natural way to represent relationships that would appear as ground instan nt’ Predicates in predicate logic. “that It is particularly useful for representing the contents ofa ty| “scribes several aspects of a particular events, int ce of i ical declarative sentence Bae @ scanned with OKEN ScannerSementic N eset. therefoetoatempto mode a yg eats commorsesere. J. Lin nic logics formal ogi whose cy 2 pnt EWU raonset on corn DStth reine ee grt jes bog a sen ee 2 ed to formalize plausible believable reasoning 5 Defaulat reasoning 5 circumscription 2h nantenanc ses. fa short note on genetic algorithms. ® Aigo Get st) ede ies computing, GAs ae inspired by Dis ise exon corre cometiion among indivi ny the fitet’ In at se ange a en ‘Genetic operators are analos Reprod : (eessetion) i Fig (b) = Using Partitioned Semantic Nets re vod ).1.(¢) Write Se Genetic fon); and Mutation. Algorithm is presented below: Q-14@) What do you mean by term Heuristics ? j sor rae Hewrisie ‘sa rule of thumb that Probably leads to a solution, Heuristic yt |. helptoredan ie strategies because of exponential mature of the most problems. Heuiss Avtifiialineg efltematves from in exponential’ ‘umber toa polynomial nun ‘meaning, nantehe® heuristic search has a general ‘meaning, and a more specialized textil isnot he ni a een Sense, the (erm heuristic is used for any advice that is often effetvs lemurs sua S¥ETY se. Within the heuristic search arhitesture, hovers The allowing to SPecial case ofa heuristic evaluation function inte tS make use of heuristic evaluation functions. (2) Constraint Satisfaction (4) A* Algorithm © Beam search Search Algorithm , a motonie reasoning. f fevaldconchsions, trees The conlisions dered fromthe mono! Hnceasthe amount ofkaens "8 NEW axioms(Fat and avfone) inte knOWe paul 2 sino Tone go sined in the knowledge bas ae tt me 7 venue and invading reasoning Inthe wc —- on icture, AS new ino pie ues ne eT ees knowledge in the ratte: AS new facts came into Pi Each eycte in Genetic Algotins ROME ei ret "ation ted to nonnenotonie gowih en ald thereby requiring further rete Sap atescin eet ryan Ful eal worl. itis nt resound tnes, | mijelen tae spas ann Pe firsttime itis not rescmiblewenes kent the Fst pass ‘complete or valid, Thy i yee iia thatthe knowledge acquited 8g ir yt oe a ll @ scanned with OKEN Scannera EES ty ofthe AL computer growing he inet. popula ofthe plc bs ao es The perfomance of the strings oe ri pe Maitsh and BM into bit-string, called chromosome The pe called wit! is forthe Apple b compatible computer, such as voice an init lp ome us epesentng te ons othe mae egg lO ecomevalbl Abo Alec bs chest eat xm oth croesames thy seed fr subsequent geet megs | sash MY ry re Gemnd or Actes hema ested etd that sls poss isan sponsible raga |e 2S, ity i nlite ph eatin us, After selection ofthe population strings is over the genetic n't’. inst proses ou Ine fis. eth cesoverosation hat reconins hepa Poh ar a: Aplcatonaf ATs lo teach to sles sings (chromosomes i exe gg pe ail ili snipe ein ‘Unit-1 a ties. gst Pl cea eile nop 24) Dee trm Ail Iniigence. What are appiatins ot tg, |, Wrarenstaaton ane ror ne shat various fields ? inthe ake ast sry: Robots have becom onan yids They ao As iil lcligec(A ste intigene of machines andthe branch of year nn dangerous humans. Raks ae oven isin science that ims to create it comp | josthat Core Ty ead to mistakes or aces doo lap incnettin Intelligence isthe computational par ofthe ability to achieve goals inthe wor ey especies WIEN Ty find degrading: ‘nds and dares of ntlience occur in people, many animals and some mene SS | “spjobs which nurans my Lae a sey tency. Wil scot ton McCarty, who coin h tem Ail Intelligence in 1956 defines (a) MESE? TH ne cathe is fhe A “te since and engineering of making intelligent machines, especialy intelligent compas stare eying to make the COP ond procesing of i so aes 68 Late eon, performance, music Hs ee re csi Tid nd rl SS ite Te aa ara me) Game plang: aes aT AE et jst ea = ayenemies, partners, and suppor chara oi 0.2(b) Explain Hill ‘Climbing stratezy ya are he rls cing is stratery sh wean i SS td white applying ths eric Sexe bie Te Ans il climbing i n= aie Tt mts nee the seach and ee So et etd sn caren ing a ee Soke er expansion i nie oe = thar beter han ay 618g a ay an eager, but blind mount 2 Ames a make sec Atal eligence [Nobert wiener was one ofthe fist] mgono further. _painbaso sere Simon develope The Lage ns sil of edback hry. nate 1955, Newel 8 eoblemsin Hit Climbing: BOP nga rogram, represating each bio eed BY many to be the fist AL progam" |] asian, ihe govt mony EIT a hap branch tht wuld moet ely rec OU tempt a solve iby sees | tame no btersttescan Dee ield of ‘the following states : - FeldofAtbas made itacrcial stepping stone in develop" Heidel set Aslent and expertse of others inte ther ofA, organized a conference to dre — ns oO ai me (0 Loca Maximum # At @ scanned with OKEN Scanneror more rules before doing the test. This coresponds to movingi ticularly good strategy for dealing with ridges. —, ly! yi Ts APH ss 1000 1s inefficient in large proble ‘ sis ery se ined with other methods, 0 oy What do you mean by Game playing in AI? gis an important domain of artificial reaowledge the only knowledge we need to provi vg of wining or fosingte game, to win the game. So, both of them ty to make the best move Jung techniques ike BFS(Breaith Firs Seach re not acute ch turn. Seat ranching factor is very igh, so searching will take lt oftime So, weneed procedures that improve aeedure so that only good moves are generated that the best move can be explored first sech technique in game playing is Minimax search procedure. ‘Taemgsteommen sete) depth) fiited search procedure. It is used for games like chess antietac- WY aie expla “Puddle win the et of rain 8) um. Tan area of isis bats ncncenag Mate, ‘Ans. Alpha-Beta Pruning number of game state increases ings ‘moves. cgonentaly, To reduce search, the pruning is \unsintans two threshold values one is called 2 “These threshold values are defined as follows: Lowe’ bound on maximum value of wilt 03 ans.came Peyin Fig. Gi) Plateau : sai ah m sues Aff he Sach sin which al pconl lawl compaiens. etrmine the best direction ino a th oth players 119 hich tombvebyaare | ble Mt nisasthe we “Generate “est procedure 5° In MINIMAX search, done. Alphabets is one such Pr jgha or‘) and oerbeta (8 V svesea ridge by single, funtion. antnebesteccepabe valet ® iy funtion in maximizing ply. ee es {= Upper bound on minimum value of Mee sr 7 ; of uliity function incase minimizing pl. ing vty ale es mae seihtgene eons tenn very prec Node ms a i : . he cdeeee cine: eee termine ere | an alpha wil be pruned. That means. i oe, Sin . : 1 Similarly, if utility val ig tii iy vaso oer neereene sesth best iest 0 iy fumet beset future al the nod ore than eta incase mx jiu LES fee will be pruned ons @ scanned with OKEN Scanner _4+En SRr terrier men titer Unit It : ane between propio ogl amd pees, Predicate Logic etogicthat | Predicate Togic isan expression consistingof| fdectarative -variables with a specified domain. It consists truth value, objects, relations ‘and functions between . the objects. Ttisan extension of propositional logic covering predicates and quantfcation “Apredicte's th value depends on the Here the posi variables value Jee the possbiltes are considered : ee es aa tp 02. Amin py the best va fi Pediat loge els nd analy ihe sare tei $..0 Ts liek propagation rota a mxiizng mo yee “ofthe subject verte predate. Tera ily vale 7s fir more then secepted (siti mininss gt thee quantifiers: Univesol Quantifer() imizing ply) Sp ep for all Existential Qantfier) further note Evil not be ot be explore nthe sitaton wh tcl willbe prned on when more plies are considered bho ieing there exists some and Uguees Se — Aa ipae dis a |= 5, all notes and related sut ir me ho oe =e a tp tener sub tees having valu of uy ib ‘Quantir (3) deniaing nse? finan earhsa els —- Fig: 8 pring ing ply wl be pruned Te je Procedure ae “opositigns combined with Logical . it Peet BPE G4 TCA } peice Tike, ancien inar(Positin Dest ge} LON eon ie aon), DUNE J Exclusive OR®). 1. Ifreached spy (Post snr desc St Ceounction(Ay 2a. Path = nil, \LUE = STATIC (Position, Player) Teptiton(=) sion ot oe Doxbe Implication). = Tris a more: specialized represet vert one reurs more p < oe ly ofthe and set SUCCESSOR inthe iit | [> nerd rpreseiaion Tene dome SOs emp then return he Ty Theannot deal with sets of entities, Tecan deal wil 4 evn, mie tl same structure ofquantiies Arama ese tone, o VALUE = BESTacon na sum the struct ication? ray ee : QAR) What do you mea by Ko ay ed ie aussie by Sen i ‘Ans. Skolemisation : In pred 1d Skolemisation (after the name of 8 "aif. nthe proces of skolemisaion fF PATH = Bes == Shoe isthe ms ere nes na am ses reuivesst aoa tha for rented Po the order in whi Soe ee ithe cat Mert ae rset owned tear avastumed into pater ma, : Hor becomes vised on pn fr the a B search needs examining ot "tint auch avaible aha esse ie cor ee at otenitentalvarisble #168 EN ‘on font ite See 0 Tat means, effetve braced | factor in nme ter the value of b+ te univer into a I becoee eh val f b"is 35, with lpi bt ese ate a Hm an e8ees 107 = Ths iniates as —— cates a significant amount significant aout ‘The Skolemisetion is summarized ss: wich 00 U0 har catoneh vail arene @ scanned with OKEN Scannerot possible to answer simple question suchas itis < oa is the heaviest player?” for finding heaviest player is provided then these facts wl ease 5 for (if) Replace éach existent function are universally quanti (Gif ovo different ur lly quanifed variable bya Futon The ey variables which include the exsten " men alin ther agg hy rally guatid variables have same name name, ipa procedure Tor answer. We can ask things like who “bate” and “tows, 0 fo Pre pute 2 = i ov : 1] pat col : : et (Reach nivel gud aa by Spy ie, Yee — ute se Knowledge : Relational owes ade eet coiig bys 2 QS. Explain the various ways of knowledge representation in Ay L putes jngassociated values, . ‘ fm Koon rpescratin tema ed enced kre ay ina Fo 8 corso ments inert abs om ees etn en's knowles bse. The abject of nowledye representation isto ome iowa et int noe es Wine jew ene ort: compu ancabe form, such hatitcan be used help ineligentsysompere "Sar stint go a era al attributes ofthe parent element Repesentton of Knowledge: As knowledge consist of nee concent af crane many eases Me ‘Boa ean be represented asin diferent foms as mena mage asspohe vwaiten ttt epee is sn ald seman evr ec ‘ some language, sega oh irs anda character sings or colecon eh in [pale ial structure i fromsofinferenceis propery inher, spaccond te on ae ae pe mete nef foc ie ec strucre, Te The objects OF ae ina genered tiem. Sine he rmre general classes : cf atibutesof objets. aes | ‘phic represents objects and vals _ WaITEN TEXT Bax 000 al mr oF + nt gm abject IE, ra f ; | aero in esata flrs FER Ff cumcrersrane sind é PUBLICATION a nas Aes ‘MAGNETIC sPoTs = 4 9 i: Dierent Levelt Knovedge Keowledge Represeatton Schemes aon) Beinn Knowledges Ta nowedge associates elements of one domain wit ‘The table betow s o¥2snpe wy oy = _ acts about a set of obj, See, : brie about a set, ie Ae put sysemataiy ‘in columns, if aos te oppor rn : Player Height Width, Bats-Throws a Shyam 325 1 Rightright 5 @ scanned with OKEN Scannerdata, 8 ‘gical Methods : Statistical methods provide amethod for representing beliefs ©) Sain (or ane.) but for which there may be some supporting or contador) suis reset 8 onages of Statistical methods in two broad seenarious ae sivmane Randomness: Playing Card games are a good example. We may not be able cist oncres wih en but we have knowledge about the ikthood of etn ee ike bring dealt an ee) and we ean exploit his ‘ceptions: This advantage etn be represented by symbolic methods. However ifthe ito exceptions is large, sch system tend to break dow. For example many common pan expert reasoning asks Satstical techniques ean summarise large exceptions without orig enumeration. {G) Fuzzy Logic methods: Fuzzy logic also deals with uneerainyy problem. Fuzzy tajeisafoxm of many-valued ogi; teal with reasoning hats approximate rather than ive artic ® patur specie onl eff snlexct In contrast with tradvonal logic theory, where binary ses have to-valed ogi tue MS Becordingly cant wedee that i> | Vany logic hasbeen extended to handle the concept of partial tut, where the ruth vale may ‘all routines thar anu" hi hat know i ‘ange between completely true and completely false. Furthermore, wien linguist variables re wed these degree may be managed by specific funetions. 6,0) Explain Denier shafer theory with the help of example. (10) ‘As. Dempster Shafer Theory :There i some information that probabiiy cannot describe, For sxanpl, ignorance. Consider the fllowing example: If we have absolutely no information about te con, in probability theory, weve assume that it would be 0% hed and SOP al However, in angler scenario, ye kow the eon i fi, 50 we now for aft that ‘ould be 509% heed and SO% tal Therefore athe Sno diferent scenarios, weave atthe resolveambiguic iguitiesand cont Troe (RUSE BS able to update cna conta ay Symbolic method: thos hat eon ©) Statisticat mene, (3) Fuzzy logic eth - @) Symboticmesnege THe concluion, How we present ital proranceinrebablly hear wacom problem. ‘enter shater ear en eel ses problem h Dt Safer ery, tetheignermnee setae, hebelcrotteadantebsteter hil would Forte ia ‘cna nebiof eat would Othe see Ta olla be ‘antrratve netration cn tht he roby of head span sis pen gerne, wen tap ery seb on ("ine Upon sn Scone we uy hat p05 wi potty ant hash var ih pee a Zoot tnd oighoraec gh belt phat some ea dsroaon on heer 0 m “Tebase en nroresening weeny ie model ods The (Symbol S)Seupaeanidie nea Aneel fobs within whe eae rob- ‘The (Symbolic) methods basieal site “A : > Tue, ally represent uncerngy bleh > Passer ‘bly es witha certain confidence based on the Belief B and plausibility PL provided some ‘evidence fora proposition P. <> Neither True nor False, Some methods also had problems with > Incomplete Knowledge - > Contradictions in the owledge, as being, ‘ip The belie brings together all the evidence that would Lead ws to belive in Pith some extn Gi) The plausbiiy brings together the evidence that is compatible with Pand is not ‘wcnsisent witht “This method allows for farther additions tothe set of knowledge and does nat ‘hisatoutcomes @ scanned with OKEN Scanner1 isth ; rember ofthese 20 and ae a ee hea “othe et 20 and takes values inthe vy mass probabil Misprobbilty demi fear oe So.istheset(Ma(E) Coane ones tio 2 but FPCPLELCP). (©, Pneumonia (P) then 20 isthe nn Subse, istheset (or P.O, Me ‘ ‘The confidence interval is then defined as [2(€), PL(E)) where ill the evidence that makes us believe in 1 correct PLE) = 1-808) “cae ness ofp ang =1-EM where “LEE 1 c i.e, all the evidence that contra: : where AcE rye ae pla plan for baking a cake nf : re cake might 88 eg05 ae Nour, get noe =pay forall goods =e nna Botibekchen the agent enn Pata plan because . Me bentcan wander around the soreness ae chy conane eB our an inntgsie tems on its shppingt 2 Expl @ ave sree ae Representing sates Actions and Goa ns, Ree sme mdel ofthe work, ana mo si een a paral neon Fam Kae idk neve te. Forex sn ear object iit is nowhere neat the object. The precondition o secon bao The let of 20 ree cra delivery abot wort with mal and eofee Is : To reason about what t0 do, an sdel ofthe consequences ofits actions. esto states. Its partial because not every ‘robot eannat carry out the ation to pick: fan action specifies ecifies the resulting state to deliver. Assume @ st ust ‘Example = santifed domain with fourleation x shown in Figure. Coffee ‘shop ; Sam's Office Lab Features tovdescribe states RLoe = Rob's location RHC = Rob has coffee swe = Sam wants coffee aw = Mailis waiting RHM = Rob has mail Actions = move clockwise move counterclockwise pue = pickup coffee de = deliver coffee @ scanned with OKEN Scannerpm as -picp a = deliver mail Figure : The delivery robot dom: The robot, called Rob, can buy coffee atthe coffee shop, ‘move, and deliver coffee and/or mai ek Up mail in them ams office wil sop su, detiveredto ofthe prob Delivering the coffee to; Samsoftce ny 2cionsandinplanting. sin represen Tye ne can be described in tems ofthe following features je rob's loston (RLe), which sone ofthe cafes ot (3), Sam’ ‘hema room (nr) orthe laboratory (ab, 97, Seog whether the robot has coffee Reb does not hive cafes = whether Sam wants coffe. Sam doesnot want coffee, ~ whether mail is wait ‘nd ~moe meanthereisno mat. (RHC), Let rhe mean Rob has coftee ‘nd “rhe mean (SWC), Letsire mean Sam wants coffee and “swe mean ing atthe mail room ( (MIP), Let mv mean there is mail wea Waiting "9-1 mean theres mai wig ther the robot if carying the mai at ig he thane ei mall (REN), Let rhm mean Rob has mai, Suppbse Rob has sx actions Rob can move ockvise (me) FON Rob can ren’ eiiterclockwise (egy, | OS Carrying coffee and i is at Sam’s office. Let de mean OF dis rhe p Roses SA'S office, Let de se Assume tha * itis only possi efor Ro lercnimpes e Rob © €o 000 ection att i Et theseactinn O° tion at atime, We ass th Unity tis an By Aer System 9 . ‘Ne various domainy,” E*P1IN it architecture su sen, tile RR GTIESB Sisley Sob ppers By -202 tape Sptem: Anenpesten sof Ans. blem sited Seem A cine = Knowiedge™ Exper system 7 in history. (ae N and DEN two expert system DENDRAL ate Mivoussomponets of an Expert stem: cot aul and heisicknowe canine are that attempts to provide an answer fy uncertainties where normally one or more human experts would need 10 i int mn or clarify un ipared with traditional computer. E computer) ‘Data structures = Program in traditional computer) ledge Base : The knowledge base stores all relev jon, data, rules, all relevant informati ige of expert systems Aone used bythe expert system, i the knowledge base of exp ships u User nverface Fig) is wie rically cask dorsi thats widely shared, peal stat knowledge of he jose knowledgeable in the particular Fatal kno ony eed upon Kmowledge oral ound intext orc experental more judgemental knowledge peless aru Mpeursicknowledge's rarely discussed, ands F A yest anenienen pepe soe fo I en coed guessing re woes 1 under en dedige of multiple human experts. Aru largely nthe knoe Ap bowled of multiple hun inthe field. eis the can combine I oactons sa 2 Aedes oe ni ae : @ scanned with OKEN Scannerw A frame's another approach used to capture and store knows... Inrelates an objector item to various facts or values, A fre ee a sue for bjectorented programming techniques, Expert syotene eset hy | sore nove ae alo called fame bsed expert systems, "S MAking ug op gy | Hens ©) none The purpose othinfene caging id bas | meading on the eror rte bet rd gation algorithm asa tions, and ty, must find the right facts eesti intl ENN is shown in Fig Fence method are common) Pins et ea Hidden Output i : partes the other Iyer. ion, anda hidden layer i an intermediate ayer which separ 4 tem seton om the input ayer to output lye are connected through acyl ars raining algorithm to learn the dtases which modifies the neuron wei etween target and actual output. In general, ANN uses the bac oackit ining algorithm to learn the datasets. The general structure of Fig. + Property Inheritance Hierarchy Here, the knowledge isa mapping process between (4) Procedural Knowledge: Here, ening pst domains that specify “what todo when” and the representation is ofhow to make” rather than whats”. The procedural knowledge batt @ rye isc eae nc ny {i are represented ssl programs that know howto do specific things, proceed. : Example: Apaserinenst contain articles, adjectives and nou articles, adjectives and nouns ge has the knowledge that a noun phrase may rallanguoge has th : thus accordingly call routines that know howto process (a5) known as production systems is Arulebased system uses rules “Thedsfintions ofrale 1 nto the system. jie the rexpert systems) a6 eer kngtedgecoedine Hesse tha ict syste a shelved eee nxt HEN teeing based system depend ngakowledies ert reasoning of human €XP Se. em ith exam e-based systems (080 fax intelligene® ule Base-Sys system: Ro simples formal Q5. Explain Rt @ scanned with OKEN Scanner: Artificial Tatehigaies snowtedee ina declarative static way as ase of things Which ae tue ue-based ya stem vremet knee ens of eof Us that lls what td oF What 6 cone de in Afferent situations. va rule-based system i a way of encoding a human expert's knowledge ina fan nly arrow area into an automated system. A rule-based system can be simply created by using vcerions and a set of rules that specify how fo act on the assertion set. Rules are set of apressed asa setof ifthe statements (called F-THEN rule or produetoin rule) IFPTHENQ which is also equivalentto: PQ “Acrule-based system consists ofa set of IF-THEN rules, a set of facts and some interpreter contelling the application ofthe rules, giventhe fats. The idea ofan expert system istouse the knowledge from an expert system and to encode it into a set of rules. When exposed tothe same dat, the exper system will perform (or is expected to perform) ina Similar manner to the expert. Rule-based systems are very simple models and can be adapted andappied fora large kind of problems. The requirement stat the knowledge onthe problem satea can be expressed inthe form of then rules. The area shold also not be that large because ‘thigh numberof rules cartmake the problem solver (the expert system) inefficient Elements ofa rule-based system : Any rule-based system consists of few basie and simple elementsas follows: 1. A set of fats. These facts ae actually the assertions and should be any thing relevantio the beginningstate ofthe system. 2, Asetof rules. This contains all actions that should be taken problem specify how tact on the THEN part. The system should contain only and avoidtheielevant ones because the number ofrulesin thé system will affects performane® 3. termination erterin, This is acondition that determines that solution bas beet found or that none exists. This is necessary to terminate some rule-based systems: that find themselvesin infinite loops otherwise. Faetscanbeseenasacallection of data and conditions. Data associated the val of characteristics witha thing and conditions perform tess of the values of haracteis ies © determin somethingisofiteres,perhapsthe comet classification of someting we anevent has taken place. For instance, if we have the fact: temperature <) then temperature isthe data andthe condition is <0. i Rules do notinteract directly with data, butonly with conditonseithersinelyor et (oinedby lopial operators as shown below. thin the scope of Jevantrules | asics | \ | rene 3 fete apment on cos c is is called inference. es oineatonls the domait- Bai shots to actin; ‘when specific informati : i see mes in, TH e: Ifthere isa power fa lure then ( = The inference is} | lest _ al uncertainties wales pot ea me based systems is tha often the connections reece! by vem e ahoely oe in(ie ainsi) and the gathered information SOND i renotabsolutely certain eet cases, acertanty measures added tothe premises as wells the conclusions theres ofthe system : a " ad faint descies how much achangein hee 7 sepreise will change the certaty ofthe conclusion ead SIF product in order to reduce te probabi Initssimplest form, this looks like IFA(ith certainty x) then B (with certainty £()) ‘This is anew rule, say ule 4, added to earlier three rules. ‘There are many schemes for treating uncertainty in rule based! systems. ‘Themost common ate : = Addingertainty factors. — Adoptions of Dempster-Shafter belief functions. = Inclusion of fuzzy logic. Applications : Two major applications of probal — RiskAssessment = Commodity Markets. Govermenttypcally apply probabilistic methods in environmental regulation where itis ‘aed“pathway analysis”, often measuring well-being using methods that are: stochastic in on choosing projects to undertake based on statistical analyses of their probable effect thepopulationasa whole. sn inifcantaplcation of potailtythery ines te eee such as automobiles and consumer electronics, utilize reli of failure. The pro bility theory in everyday life are in reliability. Many ity theory in ailure @ scanned with OKEN Scannera A Seetion—C ici agaaee Q.6. Explain in detail Dempster-Shafer information about the coin, in probability theory, wevill eng ee Ee have 0% tall However, inanother scenario, we know the cots ica ae eos Dempster-Shafer theory can effectively solvetisproblens Inpro one for the ignorance scenario, the belief of H 's problem: In Dempster-Shatter Theor i ‘of Head and the belief of Tail would be 0. Fe vi scenario, the belief of Head would be 0.5, the belief Tail would alsobe og oO ‘Analternative parameterization can say thatthe probability of hends i -p-Upon gnorane, we sty tat pisuniforly disebuton the, Hiner Gre coin assumption, We say thatp = 0.5 with probability 1 and ha other values with peta The! ‘basic idea in representing uncertainty in this model is: . sanyo SeRESRSgRS tara: Anna of sei vinnie ability ies witha certain confidence based on the Belief B and plausibility PZ provi some evidence E fora proposition P. eee ae Gi) The belief brings together all the evidence that would lead us to believein P with some certainty. , : Gi) The plausibility brings togetherthe evidence thats compatible with Pand isnot inconsistent witht “This method allows for further additions tothe set knowledgeand doesnt assume disjointouteomes If. the set of possible outcomes, then amass probability, Mis defined for each ember ofthe set 20 and takes values inthe range [0, 1} “The Null set, 0s also. memebr of 20. Misprobabiity density function defined not just for © but fr em all subsets ‘So Q isthe set (Flu (F), Cold (C), Pneumonia (P)) then 20 is the set (8, (7),(C) Pd FOE PC PLE CP) ‘The confidence interval is then defined as [B(E),. PL(E)| where BE)= 2M Where A.¢E i.e.all the evidence that makes us believe in PL(E)=1-BCE) _>M a: bility cannot absolutely no the correctness of 7,44 tdence that contradicts P | io eee pers, July -2022 asi ie jing under Uncertainty ? probability Theory in FMS of Reasoning un' es i Ba vwsthe relation between two conition eae Thomas Bayes and is theorem isnamed for ste Be depresses tecondiional proba Ns vented Bay" rule. Bayes’ theorem alba Crea Bf oa ee ) Sceposteroe probability” ofa vera of the “prior probability” yofE,and the conditional probabil hi fy” of He peor pratt of Band OS temmsof the “pr ity” of He i itt was more unlikely re given H eimplies that evidence ha astronger confirming effect it was before being observed. PUI seh Sn qa. whatis Blaborate i Bayesian Theorem Bayes SOIC each other. Pee) PH) SPE PH) A “This statement/equation can be read as that gh athypothess Hs trues equal tothe ratio ofthe probal sea ror evidence on te probably of Hand and thesum ofthe probability of over ne scofall hypotheses tes the probability of these hypotheses 1 Ft set ofall ypothese must be mutualy exclusive and exhaustive, That finde examine medical evidence to diagnose anillness. We must know atte prior probabilities of findsymptom and also the probability of having an illness based on ceinsymptomsbeing observed. ‘Simple Statement of Theorem : Thomas Bayes addressed both the case of discrete otablity distributions of data and the more complicated case of continuous probability {xibuins Inthe discrete case, Bayes theorem relates the conditional and marginal probabilities tfeventsA and B, provided tat he probability of B does not equal zero (ap) = LED PCB) InBayes’ theorem, each probability hasa conventional name ¢ | ‘Paste prior probability (oF “unconditional” or“imarginal” probability) of 4. tt 1 ‘s*prior” in the sense that it does not take into account any information about 2; however, the |} Seat Bneed not occur after event A. ~ P(4|B) isthe conditional probability of 4, given B. Itis also called the posterior Wbability because itis derived from or depends upon the specified valuc of B. i ~ PW|Aisthecontonal probability of B given tials calle the Hkstiboo GOs rr ormarial proba 8 andacsesa nmairing constr Be hse nis fo gvesamathenatcal representation theca ity ofeventA given 2 is related to the converse conditional probability of B viven + “This sates wven some evidence E then probability bility that E willbe true given 7, times @ scanned with OKEN ScannerArtificial i ita Section -D *eence! Q8. Explain in detail Expert systems. “Ans: Artificial Intelligence isapiece of software that simulates the| bctaviouery jnigementofa human or an organization that has experts ina particular domain isknovnasen xpet pte. Ie does his by acquiring relevant knowledge fom ts knowledge base 5 interpreting it according tothe user's problem. The data inthe knowledge base i added humans that are expertin a particular domain and this software is used| byanon-experere squire some information, is widely sein many areas such as medial diagnos, coun coding, games ete . ‘Anexpert system is Al software that uses knowledge stored in a knowledge baseto solv problems that would usually equiea human expertthus preserving human experts owledgein itsknowledge base, They can advise usersas wel as provide explanations them about how they reached a particular conclusion or advice, Knowledge Engineeringis the term used to define the process of building an Expert System and its practitioner ere called Knowledge Engineers. The primary role ofa knowledge engineer isto make sure thatthe computer possesses llthe knowledge required to solve problem. The knowledge engineer must choose one or more forms in which to represent the required knowledge asa symbolic pattem inthe memory ofthe computer. ‘Example : There are many examples ofan expert system. Some of them are given s MYCIN--One ofthe earliest expert systems based on backward chaining. Ica ‘identify various bacteria that can cause severe infections and can also recommend drugs based conte person's weight « DENDRAL Iwas nati infeligencebased expert system sed forces analysis. tused a substance’s spectrographie datato predict its molecular structure: *RIUIKCON- Itcould select specific software to enerate acomputersysen she? by the user, + PXDES-~Itcould easly detennine the typeand the degree of lengeancerina patent based onthe data # CaDet-Itisacl palients. ‘© DXplain Itwas also actinical suppor system that could su hased onthe findings ofthe doctor. cal supporsystem that could identify eaner ints ea" stages gostavarey ce ‘coms a rules, Itconsists of .edures and intr pure of an Expert Syste” represents facts an proc cavehitect - + Knowledge Base insi domain. “the knowledge base esto solve problem, cular domain as wellasrul knowledge ina particu! ee relevant ction of the inference engine s (0 fetch the : aan aitand to finda solution relevant f° the user’s «knowledge base and applies them f0 ;planation and gatarelevanttothe d ~The fun knowledge from the knowleds % problem. The snference engine aoqire therules fom I safer new fasts. Inference engines C2 also include an ex! ine known faetS cebu abilities. _/ : e ny i Learning Module—The function of this component is «« Knowledge Acquisition and: tallow the expert system towequire shore and more knowledge from various, sources and store it in the knowledge base: «= User Interface ~ This module makes it po with the expert system and find asolution tothe problem. « Explanation Moduie ~This module helps the expert system f0 give the user a explanation about how the expert system reached a particular conclusion. “The Inference Engine generally uses two strategies for acquiring knowledge from the ssible for a non-expert user to interact Knowledge Base, namely ~ Forward Chaining ~ Forward Chaining is a strategic process used by the E: St i fe stem to answer the questions ~ What will happen next. This strategy is mostly used for ‘anaging tasks like creating a conclusion, result or effect, Example — prediction or share market acme eS @ scanned with OKEN Scanner— Artic ig eg Feet a ved papers, July -2022 procedural code. pase = bs « Expert systems eansolve complex problems by deducing new facts through existing << 7 | facts of knowledge. represented mostly as if-then rules rather than through conventional rect2 |_—I | ‘+ Expert systems were among the first rly successful forms of artificial intelligence C Les] | uname Faas | ‘ Limitations: | + Donothave human-‘ike decision-making power. mA ‘Cannot possess human capabilites. ‘Cannot produce correct result from less amount of knowledge. Forvard Chania | ‘* Requires excessive training, ard Chaining | Advantages: Backward Chaining Backward Chain + Low accessibility cost Why this is strategy is most aN li : Toot aus or reason behindit, considering what has already h cam tutte sarong. onsuntkehnas stomach pain, blood cancer or dengue, ete. tappened. Example diagnosisof + Low error rate. | * Capable of explaining how the reached soli \ Disadvantages: Fast | «# The expert system has no emotions. | ‘* Common sense is the main issue of the expert system. Decision 1 + tisdeveloped fora specific domai ‘ Itneedsto be updated manually Itdoes no learn tsel. « Notcapable to explain the logic behindthe decision, ‘Decisions Applications: The application of an expertsystem canbe found in almost llareas of Fast | business orgovernment. They inehide areas sich s— | "Different types of medical diagnosis ike intemal medicine, blood diseases and show oct Dee | | oe Diagnosis ofthe complex electronic and eletromechanical system, {Kove diteteemeticeien ee “ine tomas nines eenh noe en mentanrehonnses ‘Sl ee - Gnggi theo of ensuing an expert for various dons suchas med A So - eet *They ‘use.a knowledge| ‘base and inference enoine. Teas specie @ scanned with OKEN Scanner0 ‘ 5 Artificial Inuetigengs Expertsstemshaveevolvedso much that they have started various debt far ofhumanityin te face of such intelligence. with authors suchas Nick pannel the cfPhilosophyat ioe University), pondering fcomputing power has rofessoe tocontalit nied ourabilty Q9. Demonstrate 2 discussion of AI, its current trends, ti its current trends, limitat roe imitations and ___Ans. Artificial Intellgenee: Artificial Intelligence (AI) isthe machine-g a inligene that simulates human behavior or thinking and canbe trained a solve spectng problems al saconbnatonof Machine Leaning tehiques and Dep Lening heed afc intelligence models are rained using vast volumes of data ity toate sie ‘ 1s of data have the ability to make Al Trends: Artificial intelligence adoption in different . ilferent enterprises has grown due billnby 2023 Snethe COVID-19 panes hithe worth potent vale fafa iti Past town. Asuvey published the Mekinsey State oF In Naver 2020 suggested that at least half ofthe onganisations have ado ; alf ofthe onsanisations have adopted Al functions in their organization. This article on Ariicial intelligence trends wll help you understand more about psomingArticial Intelligence trendsin 2022, Al becomes increasingy i ions conti important as organisations continue to automate day-0- ‘fzreandniesiandCOVID fected ass Businesses are more i ialycomet sin the skdownan or fom bone wee implemented, ‘us oFAl adoption i restricted to improvinggthe efficiency of operations orthe Greater Cloudland collaboratic fe atleast eaten collaboration Rico Buen the diet ofclentinnaaon Daladoni rete shat Aric lligeneewllpay asian Wilke pesblcenant Sitonsin 2021 Trout deployment of ania netizens “soltenebe oaloa sources andthe vastanourofaviate das cin 02 Copper grtenumbet of Alsouos ht arebeing developed fr ill prblens onto a Sion pets that solutions atcan detect omen teupconinggeae Thea gett snllanctions rises wile an ice tea reduce downtime: teams in an organi = Wishes poetandtescete allow te teamsin an organizationto wok ps becomie more popular i feet Po rae Ove last few years, the complexity of IT systems ae dance onthe acess ‘pline, suchas eppication, infrastructure. and networking. T increas piezh Semesier Solved papers, July-2022 i operation ober teams can improve tee process, deion-makng ands wih ‘tops solutions and improved analssis ofthe volumes of ata comings way. Forester advised {NaI eadesto find AlOps providers who will empower the cressteameolboraton hough | Fadio-end digital experiences, dataconelton, and integration ofthe operations management tpolehain “Alwill help in structuring data : In the future, we will see snore unstructured data srvctured with natural language processing and machine leaning processes. Organisations will pverage these technologies and create data that RPA or robotic process automation technology canuse when they wantto auiomae transactional ativityinan organization. RPA is oneofthe fastest-growing areas in the software industry. The only limitation that itfaesisthatitcan only tsestyuctured data. With the help of Al, unstructured datacancasilybe converted into structured data, which can provide a defined output. Thisis one of the most important Al trends. Artificial intelligence talent will remain tight: The supply of talents expected fo be anissue in adopting artificial intelligence in 2021. Tere has been a persistent gap in Al talent and organisations have finally realized this potential Iti essential to adéress this gap and ensure tata wider group of people lear artificial intelligence. Ensuring that a broader set tsers have access artificial intelligence to focus on technology, learning strategies, and supporting’ change inthe working environments essential in 2021. This sone of the most important Al trends. ‘Large-scale adoption of ATin the IT industry : We have seen cont the adoption ofA within he industry. However, Simion pedis that orgarisations wil use ‘Alin production and start using them ona lage scale. With the hep of artificial intsigenee, 29 crganization ean get ROL neal me. This rears tha organisations willsee their efforts eine paid off. This is one ofthe most important AI tends. “(LEthies is the focus + Natalie Cartwright, co-founder, and COO of Finn Al. an AL tanking platform, pedis that in 202 organisations wil deliver expertise on howto erage artifical ntelgene acsinst major global probes, simalteinmavation and economist and ensure inclusion and diversity. As AL ethics become more important to organisaions. tansparency of at and algorithm kimessaetwo ofthe ses hata inthe spaih gente Processes have Become inreasngy popular Aca nelianceas data science will bea part of the bigger picture regarding innovation an automsation in 2021- Data ecosystems are scalable and lean andals0 provide data on time to heterogeneous sources. However, providinga foundation to adaptandfsterinnovation snes According toAna -Malobert | abigdataengineerat Globant, companies will go furtertoopinize theranemened ‘esineas and development processes. Using Arif nelgene, software dee omen ‘Processes can be optimized, and we can look for wider collective intelligence and improv lah tec wt fester adt-diven clean gor ot ofthe esperiinn aE uous growth in Me a eivenlel Thisis one ofthe most important Trends, © scanned with OKEN ScannerArifcil ineligence will become more explainable : The senior director of atcustomer datalhub Tealium, Dave Lucas, says that there willbe bigger focus on ex 3 ay Searca regulations come into play, tustin Al willbe pivotal. To clearly nian articulate ow each characteristic wll contribute tothe ee ie willontibutetothe machine leaming modesend prediction ____ Hoiceand Language Driven inteligence: Particularly in customer ear centers th increase in remote working has driven a great opportunity to adopt NLP or ASR oi speech recognition) capabilities. Les than 5% ofall customer contacts are routinely checked for quality feedback, according o ISG's Butterfield. Due tothe lack of one-on-one coaching, ‘organisations can use artificial intelligence to complete routine quality checks on customer Understanding and intentto ensure continued compliance. Limitations of Artificial Intelligence : |. High Costs: The ability to ereate a machine that can simulate human intelligence is ‘nosmall feat. I requires plenty of time and resources and can cost a huge deal of money. Al also needs to operate on the latest hardware and software to stay updated and meet the latest requirements thus making it quite costly 2. No creativity :A big disadvantage of AL sth box. Als capable of learning over time with pre-Fed data and past experiences, but cannot be ‘rcative in its approach, A classic example isthe bot Quill who ean write Forbes earning reports. These reports only contain data an facts already provided to the bot. Although itis impressive that abot can write anarticle on its own, itacks the human touch present nother cannot learn to think outside the Forbes articles. 3. Unemployment: 1 application of artificial intelligence fs arobot, which is displacing ‘occupations and increasing unemployment (in afew cases). Therefore, some claim that heres always chance of unemployment asa result of chatbots and robots replacing humans. Forinstance, robots are frequently utilized to replace human resources iamanufacturing businesses in some more technologically advanced nations lke Japan, This is notalway3 ase, hough, a it creates additional opportunites forhumans to work white also replaciné ‘humans in order to inerease efficiency. (Gfgeon @ Semester, Solved papers, July-2022 DS 4B | fen given jobs that reconsidered dangerous to humans. Robots have proven effective 4, Make Humans Lazy :Al applications automate the majority oftedious and repetitive tasks. Since we do not have to memorize things or solve puzzles to get the job done, we tend touse our brains less and less. This addiction to AI can cause problems to future generations. 5. No Ethics : Ethics and morality are important human features that ean be difficult to incorporate into an lL The rapid progress of AI hasaised anumber of concems that one day, ‘Alwillgrow uncontrollably, and eventually wipe out humanity. This momentis refered toas theal singularity 6. Bmotionless : Since early childhood, we have been taught that nether computers sorothermachineshave feelings. Humans function as. team, and team managements essential focachieving goals. However, there isno denying that robots are superior to humans when fintioning effectively, but itis also true that human connections, which form the basis of teams, cannot be replaced by computer. 7.No Improvement : Humans cannot develop artificial intelligence because itis a technology based on pre-loaded facts and experience. Alis proficient at repeatedly carrying cutthe same tas, bit if we Want any adjustments or improvements, we must manually alter the codes, Al cannot be accessed and utilized akin to human intelligence, bu itcan store infinite data, Machines can only complete tasks they have been leveloped or programmed for: if they are asked to complete anything els, they frequently fil or provide useless results, which can have significant negative effects. Thus, weare unable to make anything conventional Application of AI : Application of Al are as follows : () Finance : Banks use anifical intelligence system to organize operations, invest in stocks, nd manage properties. (@ Hospitals: A medical clinic can use artificial intelligence systems to organize bed schedules make a staff rotation, and provide medic Gi) Heavy ‘obots have become common lustries, They are is that are very respectvies which may lead to mistakes or accidents due 10 lapse in ‘oncentration and other jobs which humans may find degrading. (Gv) Musie: The evolution of music has lays been affected by tectinoe smulate the activities of the skillful musiciin, With al, Scientists are trving to make the computer et @ scanned with OKEN Scanner44 ificial Intelligence Composition, performance, music theory, sound processing are some of the major areas on hich research in Music and Artificial Intelligence are focusing, (v) Game playing : Games are made by creating human level artificial intelligent entities e.g, enemies, partners, and support character that act just like humans. @ scanned with OKEN Scanner