0% found this document useful (0 votes)
414 views129 pages

Artificial Intelligence Aakash

Uploaded by

Avinash Jha
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
414 views129 pages

Artificial Intelligence Aakash

Uploaded by

Avinash Jha
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 129
(lease write your Exam Rott Noy Roll No, First Term Examination February 2017 How many eategories do i igent System fall into? Brifly explain horizon too? System in context of Artificial Intelligence for Model Based Reflex Agemt AW ¥Q) 3 (vk) (RV Q)) ‘lo you mean by free and bound variable? Idcuify fcc and bound varia lowing wf YS BYP yA Vx QO) V RO) ST Q2. a) Solve the problem of a simple house her with «wo things oF PUSH, CARRY, WAL! b) Differentiate betwee example ‘othe Wd robot of moving a desk from one too ‘Means End analysis, The operators available ‘UP, PUTDOWN, PLACE (Please write your Exam Rot Ho) Exam Roll No. MID TERM EXAMINATION SIXTH SEMESTER |B. TECHL| FEB-2019 Pips Coe ENCES — Sibi: Anil nligece ae tsofous ___ Maximum Marks: 30 Notes tert any Three questions inating @. No Lwhich is compuliony QNo.1 a) How do you define Artiicil tte nce in terms ofits task domain? }) How production system useful 9 donagaGProaches required or the representation of knowledge ina domain? 4) problem-solving search can proceed factors determine te choice of direct ‘What is At? How do you compare hum: (Please Write your Rall No. mmediately) Koll oll No Mid-Term Examination B.Tech — VI Semester April, 202 Subject — Artificial Intelligence Max Mart s: 30 QNo. Tis Compulsory. Attempt any tno more Question frovs the rest. (2°5°10 marks) Miate Propositional logic and FOPL with ¢ nuple 6.) What ts PEAS? Specify PEAS for an Aerospace System. ©) Obtain disjunctive normal form of PYCPS (av qe 1) > 41 Explain Turing Test, Why Turing Test was eritivised ani! als» explain Chinese room Argument Test. ©) Dilferentiate OR graph and AND-OR graph (92.31) Whit ts heuristic search? Explain Hill Climbing and its lr b.) Explain erative Deepening Depth F (5 marks) 1 Search with example, (5 marks) Q3. a4) Represent the following facts in FOPL: (S marks) 1) All men are mortal 1) Some pet dogs are dangerous 1.) All basketball players are tall 18.) Lapton is a tea Seu ws are sick today ploy: by Cigck the validity of fullowing argument: S marks) ‘o> 161 vet the job and work hard, then L will yet promoted IF yet promoted, then | will be pps. Lill not be happy. Therefore, either I will not get the job or [vill not work hae! H Satistiv tien proved O43.) Solve following cryptographic puzzle using Constr FOUR MICE ‘OUND aL) Discuss forward and backward chaining with suitable example (S marks) (Please write your Exam Ro Na Bam Rll Ne MID TERM EXAMINATION SITHSEMESTER Maxima Mark hich is compass knowiedye ai where backwr chaning (ease write your Exam Roll No. irst Term Examination ‘Siath Semester (B-TECH| February 2016 Max Marks: 30 Subject: Artificial tntetigence {HFee questions including Q.No. | which is compulsory. exs) ee c) Agent 4) Logical Reasoning, or Medical diagnosis system. Knowledge with example ie statement foe “All ats ike Fish, ets eat everything they Does Tom eat fish? liming search ecaique? What ae various problem o oy ° ewe vali a al ca Wi hyo constant satisfaction proce or solving the flowing cryptarithmetic o o (emi your RN, me ” al No FIRST TERM EXAMINATION SIXTH SEMESTER B.TECH, (February 2018) Sul: Artificial tnteligence Max. Marks 30 ‘Note: Question No. 1 is compulsory, Attempt any 2 questions from the res. ad First ler priate logs: igri ofthe following graph i the cos oF io ate ne Ropers SYLLABUS ARTIFICIAL INTELLIGENCE SIXTH SEMESTER Applicable from the Academic Session 2023-24 | Discipline(sVEAB/OAE | Semester] Group _| Sub-group Paper Code CSEATICSTIT 6 PCE, PCE-3 CIE-374T ECE 6 PCE PCE-1 ECE-318T | ‘AUCSE-AIML 6 PC PC “ALBOZT EAE 6 AI-EAE AI-EAE-1 AI-302T EAE 6 AIML-EAE | AIML-EAE-1 AJ-302T Marking Scheme: 1. Teachers Continuous Evaluation: 25 marks 2. Term end Theory Examinations: 75 marks Instructions for paper setter: 1. There should be 9 questions in the term end examinations question paper. The first (*) question should be compulsory and eover the entire syllabus. This Question should be objective, single line answers or short answer type question of total 115 marks. 3. Apart from question 1 which is compulsory, rest of the paper shall consist of 4 units fs per the syllabus, Every unit shall have two questions covering the corresponding tunit of the syllabus. However, the student shall be asked to attempt only one of the two questions in the unit. Individual questions may contain upto 5 sub-parts/sub-questions. Each Unit shall have a marks weightage of 15. UNITI ‘AI Definition, Problems, Techniques, Models, Defining Problem as a state space search, production system, Characteristics, Search methods and issues in the design of search problems. INo. of Hrs. 10] UNIT It Knowledge representation issues, mapping, frame problem. Predicate logic, facts in Jogic, representing instance and Isa relationship, Resolution, procedural and declarative Knowledge, matching, control knowledge. Symbolic reasoning under uncertainty, Non monotonic reasoning, statistical reasoning. UNIT (No. of Hrs. 10] Game Playing, minimax search, Alfa beta cut-ofls, Natural Language Processing, Learning, Explanation-based learning, discovery, analogy, Neural net learning and Genetic Learning, INo. of Hrs. 1 jo. of Hrs. 10) UNIT IV / Fuzzy logic systems, Perception Networks, Kem nd action, Expert system: Inference in Bayesian INo. of Hrs. 10] ins Clustering Algorithm, Machine learning, SYLLABUS (2016-17) ad erie ARTIFICIAL INTELLIGENCE (ETCS-310) Instructions to Paper Setters: Mi 1. Question No.1 should be compulsory and cover the entire syllabus. This question 8 have objective or short answer type questions. It should be 25 marks. hello i four units as per the syllabu, 2. Apart from Question No. 1, rest of the paper shall consists of four units as ps Every unit should have two questions. However, student may be asked to attempt only question from each unit. Each question should be of 12.6 marks. Objective: To learn the basies of designing intelligent agents that can solve general purpo, problems, represent and process knowledge, plan and act, reason under uncertainty and cq n from experiences shoul UNIT-I Introduction: Introduction to intelligent agents. Problem solving: Problem formulation, uninformed | search strategies, heuristics, inforn search strategies, constraint satisfaction. Solving problems by searching, state sp formulation, depth first and breadth first search, iterative deepening. {71, T2](No. of hrs. | ‘UNIT-I Logical Reasoning: Logical agents, propositional logic, inferences, first-order lo inferences in first order logie, forward chaining, backward chaining, unification, resolub [T1, T2}(No. of hrs. ‘UNIT Game Playing: Scope of Al-Games, theorem proving, natural language processing, vis snd speech processing, robotics, expert systems, AI techniques-search knowleé abstraction. [T1, T2}[No. of brs. UNIT-IV Learning from observations: Inductive learning, ee learning theory, Explanation based learning, earning decision trees, computa Applications: Environmental Science, Robotics, Aerospace, Medical Sciences ett 4 ences (T1, T2}{No. of hrs: New Toy Q.1. Wha Ans. Pro! task. It lays d essential contr * Itis als knowledge wh applied to any | knowledge dep Examples + Tying: performing spe This skill is usv + Riding to maintain bal practice and ex; * Operat procedures, suc’ skills are often | © Perforr perform surgery how to close wor Example The followin way: var 12 var b=(]; for(var i=0; i b.pushiali } console logtb Output 11,2,3,4,5) Q.2. What is Ans. Declar: certain task. It la only the knowled knowledge. Thus, declarative knowl New Topics Added From Academic Session (2023-24) Onwards SIXTH SEMESTER [B.TECH] ARTIFICIAL INTELLIGENCE UNIT-T Q.1. What is Procedural Knowledge? PS ) - Procedural or imperative knowledge clarifies how to perform a certain Eee. s down the steps to perform. Thus, the procedural knowledge provides the aad MM: 7| essential control information required to implement the knowledge, Z ion shoul, * It is also known as imperative knowledge. * Procedural knowledge is a type of responsible for knowing how to do something. * It can be directly trategies, procedures, agendas, etc. * Procedural it can be applied, knowledge which syllabue applied to any task. # Itincludes rule: npt only knowledge depends on the task on whicl Examples + Tying shoelaces: This is a typical example of procedural knowledge. It involves performing specific movements with the fingers and hands to tie shoelaces in a knot, This skill is usually acquired through practice and repetition. | ‘+ Riding a bicycle: This involves the execution of several coordinated movements to maintain balance on a bieyele. People learn how to balance, steer, and pedal through practice and experience. + Operating a computer: Knowing how to use a computer involves a set of procedures, such as navigating through software and using keyboard shortcuts. These skills are often learned through experience and by following user manuals. + Performing surgery: Surgeons must have a wide range of procedural skills to perform surgery, including how to use surgical instruments, how to make incisions, and how to close wounds -der logi Example resolutio ‘The following example shows how to copy an array using JavaScript in a procedural fhrs.l way var a=[1, 2,3, 4,5 aa var bel); paaieda for(var i=0; i Use Forward Chaining, Fey, output Hypothesis + Must query for data = > Use Backward Chaining, 5. Backward chaining engines query for new facts, whereas forward chaining relic ‘on the application asserting facts to the rule engine. 6. In backward chaining, classification task nd quite a few other. , age sub goals. Whergy ch is goal directed, so rules can be applied that ar s891Y to achieve the goal ard chaining the whole proces i not diverts towards goal, so when to stop the rules in not known 7. Ifthe facts that: has to be established lead to a large number of conclusion, but y number of ways to reach that particular conclu out rather than information in, then backward chaining should be used. On the other hay ifthe number of ways to reach a particular conclusion is large, but the nurnber af conclusing likely to be reach using the facts is small, then forward chaining, is preferred 44. (a) Trace the constraint satisfaction procedure for solving the followig cryptarithmetic problem: BROWN +YELLOW ee « PURPLE ssinall, then there is eaore informatig 1 2 3 4 5 6 WR NE BY BROWN +YE990W PURPLE Ans. 0 oO Let assume 9 in L so that BROWN +YESSOW PURPOE so that and 0+9 =P Let O is the highest value than 9 which in 8 BR8WN Ba +YES98W Mee SURE eS pemenavae 3 3 W cast be O other wise N Be en so thatw=4 6E9901 —— ex: int val of tr a ur | inorpretation, Forex waity (Teoh) AB Publisher Net NeKat ed otaeas Q.4.{6) What are the properties of propos! ional calculus, explain with wo example. ‘Ans. Properties of propositional calculus statements: Gd) Valid: A sentence is valid, if itis true for all values of inputs or for every ‘PoaP iv valid since every interpretation of P results in a true alue for Pv>P. interpretation for which itis true is called (ii) Satisfiable: A statement having one For ex: If statement P is satisfiable “it will have atleast one interpretation taint ja true, However, P will not necessary be valid beoz itis not patinfiabl of P for which the vall true for every interpretation of P. (ii) Unsatisfiable: It is also called called unsatisfiable if there is no interpretation for hwatisfiable because itis false for every interpretation of P. io) Equivalence: Two statements 61 and s2 are 4) interpretation they have the same truth ‘value, For ex: two statements th have the same truth value for ‘every interpretation of P. equivalent since bot {o) Logical Consequence: Statement’ ot ipgaid tobe logical consequence of 81 iti watinfied by all interpretations ae vich satisfy 81. For ex: out of given sentences Pand P2Q, Pin said to be torical ne aaequence of Pc peeause for every interpret for which Pais true, P isalso true ‘contradiction. A statement or preposition is ‘which it is true. For ex: Pv>P is uivalent if for every sments P and ~(P) are tation AMINATION [APR. 29, SECOND TERM eS eTr R [B.TECH] 1g) SIXTH SEM ARTIFICIAL INTELLIGENCE [ETCS- $10) Time 1:90.30 sutsory and any two more questions. M, hich is com ay bn Note: Auiempt Qo 1h alco tality eutot QL. How does gam je itt is worth continuing as the BOW rate day iswhereyou decides tree that offers litte possibility Forimproye ploration of a sub-tree tha ‘called futility cut off the limitation of expert system. Terminating over known paths Q.1.(6) Exp! Limitations of expert aytems: esa very small range of domain : oaaaeeetaiyTanetion in the domain ofthe extracted, cognitive, logical thi, nly function in the dom: 2, They mai process oo. 3. They are not able to make use of comman sense 5. They are non-self referral systems. Q.1.(@) List out main components of Learning Agent. Ans. Following are the components of learning agent: * Perception * Critic * Learner or learning element * Problem generator + Performance element Q.1.(@) What is morpheme? What is its importance in the language? Ans.Morpheme is a primitive unit of meaning in a language. The importanc: morpheme is that without it one cannot find the crux of the sentence. It helps in gin meaning to the sentence. Q.L(e) How many types of speech recognition system exist and what type signals are used in speech recognition. t Ans. There are two types of speech recognition. One is ealled speaker-depence and the other is speaker—independent. Speaker-dependent software is commonly us {or dictation software, while speake r-independent software is more commonly found telephone applications. Speaker-dependent software works by learning the unique characteristics of a ‘sing person's voice, in a way similar to voice recognition. New users must first “train” t software by speaking to it, so the computer can analyze how the person talks. Speake independent software is designed to recognize anyone’s voice, so no training is involve This means itis the only real option for applications such as interactive voice respon ‘stems where businesses can't ask callers to read pages of text before using the syst Speech recognition engines that are speaker independent generally deal witht fact by limiting the grammars they use. By using a smaller list of recognized words, Speech engine is more likely to correctly recognize what a speaker saad Acoustic signals are used in speech recognition system 2. (a) Generate parse tree of the following: (@ Theard the kids listening to the radio. Aman wanted to go to market with Ayan, LP. University-(B-Tech)-AB Publisher 2016-7 ‘Ans. (i) [heard the kids listening to the radio, | heard the kids listening to the racho Ver . / foun phvase ® i | ~ z E Determinant Mab phrase > | pace the lds Prepositional \ dun @ prose cy ratio Sing i) Aman wanted to go to market with Ayan. ‘agnan wanted to goto market with Ayan e Noun phrase Neces ‘Noun phrase (lo go to market with Ayan) Q Aman Veto a / ~ ‘verb phrase (go to market with Ayan) Prepositional Qi phrase / \ ] Lot 7 2 of Prmneatinal Noun Prepestcnm "noun ( se We (2) | a to ‘Market with Ayan ‘d 7 vi Q.2. (6) Solve the following with alpha beta’ pruning: 6) gl he r 5 owing are the stePs* Ans. Foll Semester, Artificial Intelligence Sixth 82016 Step-2: Need not bev =: pruned f= Step-4: LP. University-(B.Tech)-AB Publisher 20169 Step-6: Step-7: Step-8: 10-2016 Sixth Semester, Artificial Intelligenc, eS / / |acTtis step is prunes Step-9: Step-10: Step-11: Step-12: LP. University-B.0 ty-€B.Tech)-AB Publisher 2016 6-l1 Step- This step is. oe Q.3. (a) Whi the ae = Explain Explanation based rial ete anaci intelligence system? ‘Ans. Role of Learnii e earning: Machine learningis the " improve automatically through experience a Unsupervised learning is the ability to fine stream of input. Supervised is at oe cr to find patterns ina str determine what category s Se ‘and numerical regression. Classification is used to Se eae ry something bole in after sexing nner of examples of ng the relationship between inputs is the allempt woreda ¢ function that Shoufd change as the inputs change. In‘ hate areeeart let and predicts how the outpi good responses and punished for bad ones. These can ee din tar Lorahernie areors, using concepts like utility. Th analyzed in terms of decision theo nd the penformanc am The mathematica ‘analysis of machine learning algor tionat learning theory. ‘of theoretical computer science known 05 within developmental robotics, developmental learning approaches were elaborated for lifelong cumulative acquisition of repertoines ‘of novel skills by a robot, through for Memos self-exploration and social interaction ‘with human teachers, and using fuidance mechanisms such as active eg maturation, motor synergies, andimitaton Explanation Based Learning: Taplanation-based learning (EBL) ss form of smachire learning that exploits a very strong, 0 TE perfect, domain theory to make Tfoneralizations or form concepts from training examples. cm example of EBL using 2 perfect domain theory is a program that learns to play chess by being shown examples: & teecife chess position that eontaing 2” {important feature, cay, “Forced loss of black ‘queen in two moves,” includes many jrrelevant features, reage the specific scattering of Pann Tt reed, EBL can take a single taining sucnple and determine whatare the Welevant features in order to form & generalization. 'A domain theory is perfect oF vomptete if it contains, in PHN, all information needed to decide any question erst the domain, For example,‘ domain theory for hese is simply the rules of ches a Mngt roles, in principle its possible to deduce the best move in any situation: vrowover actually making Such 3 “jeduction is impossible jin practice due to any Tpinatoric explosion. EBL vst training examples ‘make searching in Pyeductive consequences enain theory efficient in Bracke way vo deduce each tains example from the system's existing datahe of Tavinga Doane example extends the irenain-theory database, a aa classify future examples that are ‘similar to the training example very 4 a ce a drawback of the ‘method—the cost numerous—Wwas ana! jyzed by Minton- UIsy aaindusos ayn ur paysiqeysa are pie uone99] B10} 0 ‘S199 ureaq uray ai yo paseq st ABojou HOMJON jeanoy oy, ATVJ0sonbruyoo} ay) oxw SuysoT04 suy Sonbyuyoo} Ty uo op0u p20ys 97124, autionans xe[R|[99 949 Uodn os00, ) Auwow Jo uoneazasqo azour ow £q paquezies poutoe Uoneurieun jo dea aatyeaxo paajoau yng ‘an 30} Som suioisurg ‘Areuonnqosas uae Uy, "Bumuoseas aaniompae jo Ue Woryen “puede uoMRUEZOFUT 949 Sp[els Burwoswas aananpqy "396 ou 20} worou [ue suoryeatasqo 0 498 ayapdutoour uw tm sudo ‘possasqou ue suoneasosgo sry ay yey) ofns & UY pops Toxo} Jo synsox ays pv (oqduaexa sop pa 19 Seat o[p2oU ssedwteD ax pumnos® aye oun UY arqistau Suryom0s yexa eopr oxy Yala poreUsE ouresaq put pre svook 4 ualym ssedutoo yoyood w jo suamonom ot paasonqo 9360 ogy "qduxa 20g SU IYBIUE TENN BouaptAo poxsasqotn oq 2o¥pany ou Sea 729 tU99q sey 21 wae oy eu os ane payin 28 278 AUNTS OH ng omy aaron we wae Pe aud stojouypen Sue Troy x94 © WOE) iy oy woReUsesT! ae nk ono #8 ONS POT AxouornIon eesgaTyoxa38 ws “WORT Fag x anoqe £s9n0281P UNIS ‘aos (VO) wR HOBTY 280420 pqata8 20u 30 8 jospueats J ogy, avepspuvo oPEs# CA ST youd ano jo aaa peoungy azn SOL wey RonngayD powneut ean fe qo puny sontgen an evry HH a TN ‘uw yenqeno of, :sHrTpEsOSTY FN ser peanoxgerea wo as0008 00 SPC PO AT ‘asou) so awoxPHAS, ‘own 950413 oq w 087 0) aTOUTP 1 PU pam, todsag msgs donde ES goa ya soanm Tayndiog © TN spon oye axa 07 28024 ee aay 033909 Pus puow 10 TINTS TE ood AOASEHL SE HONS prow Sa oatp=F SV LE Pa a rae 0 smie ous past HN TS ond omy vase ssoqusN O82 M8 HO fou swine pansy snd uae} pe pred ame {S008 PLL, « ‘puso ad rus 00006 29°23! e Suanydes oyy]) sajdures Aue 9427 senronton oF . ouasu [TP oron enone sa 3584 pe Csqmat 49yndtto Jo kun pe peas se poonpoed ny Gon w 78 ogee or pay az ouoydooqer © Aa pound ‘soavt PuTOs 2 shy ax 44084 ‘8108800010 7 ester, Artificial Intelligence sixth S —— st “high fitness” criteria; fae par ea ee is ; ‘ sJation meets caapernaips a selection strategy an, 2. Ifa candidate in the por erent set using a selection Oe oa, 4. Select the best of the em i. eet the : maintenance, then oe 4. Reproduce using €r08 a create sligent agents (1A), now often known 8 “bots”, are yp, can b ‘ gents: Intelligent a Intelligent Agents user, Using directeor™™ (cost itive asks for @ om echnology that performs difficult or repet e arovided list of instructions kne™™%, Tt de coche oy led timetable, the IAs oxecute a orice ga" from capability nie 8, form script The intelligent agen a ee arity ean then can a eae 3 that can stay att Ins oer A techniques: pect ded tous by computer processors that ca Stay anette “ever diligent” capability ‘ou ve Adwork 24 hours ay 7 days a Weeks anda s Q:4.6) Explain the role of AI in the field of Medical scienc Aero o 2017 (Page No.: 28-2017) ne ») Refer Q9 of End Ferm Exam 2017 (Page : Ans.4 (by Refer Q9 TE T 1 a END TERM EXAMINATION [MAY. 2016] MM, 4! Note: Attempt any five questions including QNo. 1 which is compulsory select one quest, from each unit. . Q.1. Write short note on the following: 26x10) (a) How does the Artificial Intelligence (AI) solves problems for which nor practically feasible algorithm exist? 2 Ans, Artificial intelligent system is based on learning. Due to this reason it give P< best possible solution. Also because of heuristic function in Al one can able to find th. optimal solution. Q.1.(b) How is an expert system different from other conventional software Ans, Expert systems differ from other conventional software as in that the knowledy base in expert system is more comp] lex due to less reliance on the end-user to evaluat: the results. Q.1. (ce) ferentiate between knowledge base and data base. Ans. Difference between database and knowledge base: [Data Base (DB) Know ledge Base (KB) 1. It is defined as a collection 1. Ithas information at a higher level of data representing facts. of abstraction. 2. Itis larger than a KB. 2. Itis smaller than a DB. 3. Changes are fast. 3. ‘Changes are gradual. 4. All information needs to be 4. Ithas the Power of inferencing. stated explicitly. 5. It is maintained for operational 5. Ibis used for data analysis and purposes. planning. 6. Knowledge is represented by 8. Knowledge is represented by log relational network or orrules, |___ hierarchical model. Tit, Q.1.4d) Differ yetween st rentiate en st rong ALand weak AL qeast equal to h ray |] ent puns. a fanay 2 Tecate wath the ceaton ot some | 2 Sere canal ve [ coma ge Weis th etn fe em mputer-based AL tha Mnputer based Al that can reason ff iene Aegean pele aie cr in te dn ag | thomstin the oatanaton (eras —< oing” is sometimes referred oneiog ers ato | 4. “Helping” is called as weak AI Ss. We nave sil to achieve the | 5. w We now aioe 5. We have already reached the objective [objectives of strong i] of strong AL a Qe) Explain Turing test for AI- ] “Ans. In 1950, Alan Tu 1 | Alen Turing proposed “Turing Test, In this we havo tw One person tte ‘ om from machine uit [computers ether perso The intern ator ask question to both. The goal ofthe machine £278 listofeal =) other Peator into believing that tis the person if satisfies we can sexs like human being or itis intelligent, oe 7 ann What is heuristic search? Give an example. =25) Ans. eee sazareh san artifialintaligence search technique tha emPsTo ecause of exponential nature ch no forits moves. ‘of problems. gives [F polynomial number. ‘Example of Heuristic search is Hill ‘climbing search strategy. ithe | Q.1-(g) What do you mean, by reasoning under uncertainty? Ans. The reasoning under ‘uncertainty is focused on the uncertainty of truth value, are? (that is, to allow and procest *Feath values other than “true” and “false” edge) THCeVSIOPS system that reasons with uncertainty msAnt to provide the following ne 7 Soe ‘explanation about the origin and natur® of the uncertainty * away to represent ‘uncertainty ina formal language 1 asetof inference rules that derive ‘anvortain (hough well justified) conclnsic™ ent LP. University y-(B.Tech)-AB Publisher 2016-15 Ans. Difference bet Diff between strong Al and weak AL IC St jf Astrong A Strong AT { wa ee AL calins that computers | 1. Aweal AT week at LS aA | Gh aure re eed oom Heuristics play a major role in search strategies Tehelps to reduce the number of alternatives nom Ot ‘exponential number to 1 mechanism for uncertainty manage teristics of knowledge representation system? for any particular domain should + anefficient memory-control Q. 1.(h) What are the charse ‘Ans. A good knowledge representation system possess the following properties: 1, Representational Adequncy the ability knowledge that might be needed in that domain. 2. Inferential Adequacy the ability to manipulate the representational structures to derive new structures (corresponding to new Treo wledge) from existing StuctotS 'g, Inferential Efficiency” ability to incorporate additional information into the knowledge structure vhich can be used to focus ‘the attention of the inference mechanisms in the most promising directions ae sqnal Efficiency- the ability acquire new info I the seatahould be ee sentra its own knowledge acquisition, but direct insertion of P would be acceptable: {ferent kinds of to represent all the di Information by a knowledse engineer semester, Atiicial Intelligence Wetting 2 arte scareh andi ey ‘ alpha bet hat are Qwh : Ans. ALPHA-BETAP" ctrategy: it reduces the ves that are obviously bad for the Junge eae ted stars with vinfnity ang ge) Qe) Fr the MIN nodes, the SOP" CNT rea with -infinity and increase yt) Ans. H rhe AN pes, compe start wth init and ne eget = tine For Mite alpha-Beta procedure des Saag mney the Al ‘i rithm for -game tree. Vv -splain Minimax algo! a : G17 Bamlnin ipe two players ave called Maximizer ang q, 1949 Ans, In Minin get the highest score possible while the minimizer ty 1945 The maximizer (71 ie while minimizer tries to do opposite ml isso the lowest score possible .d with it. In a given state sory board state has a value associate i Every board stare ore ofthe board will tend to be some positive ya has upper hand ‘hen Cr hand in that board state then it ill end tobe got! 1996 aan nate of the board are calculated by some heuristics which ang ¢™% Salve The va “ every type of game i ‘Example: Consider a game which has 4 final states and paths to reach ers are from root to 4 leaves of a perfect binary tree as shown below. Assume yy maximizing player and you get the first chance to move, i.e., you are at root | opponent at next level, 196 196 Min . 19 Since thi B4 J . ince this is a backtracking b: ithm, it tri ; backtracks and makes a decision USOTthm, it tries all possible move: jug a3 ths it LP. Universi ity-(B.Tech)- AB Publisher ‘Dh la trig nt Q.2.4a) Expl Loy om me @) Explain origin and histor my Serer -y of Artificial Intelligence. . ie Ye Milestone /Innovati “ Si Karel Capek play named “Rossum 7 toa Bs come ‘robot” in English. IRUR} opens in 2 a neural networks tid in sac Asimov, Columbia Unive estes acon A eco Ne U niversity alumni, coined the term Robotics fa Aan Tring Ting at yaar aes : “omputing Machinery and Intelligence. Cla eae ae aon ee eee Chea Pagingaateear = fehnecriny sine scm ified tie Destin ot : ons s nning Al program at Carnegie Mellon Universit — oa ge naeeniraa WT showed that compte understand natural lang e well enouy apne tng wdersta ‘quae well enough tosolve algebra word problems and 1965 Joseph Wei ceph Weizenbaum at MIT bul Yo pl baum at MIT built ELIZA, an interactive problem that ; carries on a dialogue in English. 1969 Scientists at Stanford Research Institute Developed Shakey, a robot, equipped with locomotion, perception, and problem solving, | 1973 The Assembly Roboties group at Edinburgh University built Freddy, th Famous Scottish Robot, ‘capable of using vision to locate and assembl4 models. 1979 The first computer-controlled autonomovs vehicle, Stanford Cart, was butt 1985 Harold Cohen created and “demonstrated the drawing program, Aaron 1990 Major advances in all areas of AT = srations in machine learning + Significant demonsti ‘ Case-based reasoning . ‘Multi-agent planning: + Scheduling + Data mining, Web Crawler * jatural language understanding and translation "Vision, Virtual Reality-Games 1997 The Deep Blue Chess ‘Program beats the then world hess champion, Garry Kasparov. 2000 Interactive robot Pet pecome commercially available MIT iy displays Kismet, a70b0t ‘with a face that expresses emotions: The cabot Nomad expTores remote regions of Antarctica and Tocates meteorites earch formulation of well known water Q.2.(6) Write and explain| state space 5 ‘ad-gallon one and jug problem. in Ans. : --Jug Problem: You are give , ° Ste re inhas ee nited water which YOUCTE see to fill the juss and ee a cigs any messuring maTRINES on ‘dia 104 stint a Pano 249 uo BNF amnt-g on “duo ou 54 Saf asp panos ox uo Batam ox ty ‘dua you v1 Bat ang jy feNI-g 9 30 1n0 sows otOs Ae, ‘Ard you st nf ann BANC oan ow yo a0 s9qwa ous ano Ing 04 8 “Sof anne wnt ang oy y ‘Sn ony on ta ones 09 fou worgoud 91M Uoronpo1g) suopesede snows ey, “ways 332m)8 oumssy) woqqosd Bnf 29}0a Uo Yosees: ‘Ssuourag ‘wro|qoad aaoqe Suyajos 03 ayes wononpoad om oatosoC (EO. Ozxpurgs c+2>9 Bat e3-¢ 941 Bak (oer qe «s oo, (oppepye wn wompuos perzue 764 guysn spaovend 8 1nd ST got om ose 9 POURS 2 A rom 9791080 ON TE ren (on nr 191404 Seon 14rd ‘an3s cron onouse 388989 ASUS ganssurpatn pate ‘anyon ae gaan we space NIV HA ANE ra #4 Bu apps a re PE aN soe + eemasvonn ro sa ond 8-8) | aS an ged | sc a gon nyse fo wot Sean sepa sone soon es ET [ ) ao ||| sel youeas nr se ‘se100q3 9439s ‘SuuoRKe O49 of soodsaz ‘portas on nau 49 (stots poten on ae d D2s0224 600 any 1 2pou sure1309 py n opou tr spuea aN ba sed wow grad yt « (onstimay + 98005 49) dant aso opea 080 rn at esd S208 8 7 sony we 09 208 aran 80 4206 ca or mes oud d sumone xn auneey “SHY ou Ht ssuortsodaad. ° 1 TE fay goud Bursof109 O° omar wononpay 1, Artificial Intelligence mPa . iat a e (iv) F a using lines (d) and (e) and rules f thes ST: premise i J of lines (e) and (6 (@)S*T using modus ponens of {h): using rule of simplification Je of simplification io) Euplain the rus to unity two predicates os ee Jes in an inference system (such, nification & inference rules i ‘ ; ee ee ee ee t son tichsars bose on PROLOG, where the unification also play an important ru sel variables must be universelly qualified. Whenever you see a variable ina. variables must be univ f rapesioal dar -y-h era expression, assume that it is universally qualified. This allow us 8 substitutions, : , + Existentially quantified variables may be eliminated by replacing them with. tants that makes the sentence true. For example, Ex mother (x, bill), we can replace Xwith a constant designating bill's mother, ann, to get mother (ann, bill). Q.5. (6) Represent following facts as predicates and convert them to clause form, explaining the steps of conversion. (8) @ Any student who is intelligent or hard working will pass the exam in good marks, (ii) Any student who is passed with goods marks gets a job. Ans. “Anyone passing his or her artificial intelligence exam and winning the lottery is happy. But anyone who studies or is lucky can pass all his exams. Pete did not study butis lucky. Anyone who is lucky wins the lottery. Is pete happy?” Step 1: Change sentences to first order logic. 1. “Anyone passing his or her artificial intelligence exam and winning the lotte winning (¥X) (Pass(X,ai) v WIN (X, lottery => Happy(X)) 2. “Anyone who studies or is lucky can pass all his exams" (VX) (Studies (X) v Lucky (X) => Pass (X,Y) 3. "Pete did not study but is lucky" udy (pete)*Lueky (pete) 4. “Anyone who is lucky wins the lottery” (VX) (Lucky(X) => Wins (X,lottery)) Steps to convert predicate form 1. Eliminate all implications = 2. Reduce the scope of All - to single term.* ‘happy to clause form: LP. University-(BTech)- AB Publisher 2016-28 $. Make all variable names unique 4. Move quantifiers Left* 5. Fliminate Existential Quantiti @ Fliminate Universal Qunatifiers® Convert to conjuction of disiunets te clause for each conjunet 8. Create sep: Stop 2: Convert all sentences int Stop 2: C ences into clause form: 1 (WX) (Pass(X.ai) WING, lottery)=> Happy gives: = (Xai) v WIN (lottery) v Happy 0) Studies(X) v Lucky (X) => Pass (X.Y))gives! 2.(¥X¥Y) Studies (Y) Snicky (W) v Pass 3. Study (pete) s Lucky (pete) gives: adies (pete) Lucky (pete) 4 (eX) Lucky(X) => WINS (X, lottery))gives: LuckyiU) v Wins (U, lottery) Step 3: Add negation (in cla Happyipete) Q.5.(e) Assume use form) of what we want to kno. ‘e that ‘Mukesh is neither hard working nor intelligent”. Using resolution prove that Mukesh does not get ajob- @ ioe ap 4: Use resolution (and our negated goal) to build a resolution refutation graph to prove a contradiction. tation graph ig the graph to prove the contradiction. it shouldbe the cae that the | negative of the goal gets used somewhere inthe proof. S Sontraction occurs when resolve clauses like A, of + result = () (Null d false at the same time! | set) since A cannot be true an JR IWIN (xiottery) OR HAPPY (%) wIN(U Jttery) OR | 1) OR | LUCKY (U) HAPPY (pete) 1, ai) OR I LUCKY (pete) 1 PASS (pet LUCKY a ¥ 1 PASS (pete.2i) LUCKY (v) PASS (W") pass (U.2i) OR HAPPY(U) 1LUCKY (dete) LUCKY (dete) ) { UNIT ibe various components ofan expert Q.6. Whi y system. Mention 50° advantage cnet 1s epee 7 aman expertise: i st of progté ‘ ‘An Expertaystemis at shat requires | | encoded nowledge al Intelligence sixth § sticles 4. Databases 5 Jauenal Arti ngineering desi, 1. Sp cine (b) System ve program that has expert-level jn, Mul ar and knows Mie task is being performed. Ldeally iho, nchon pert award Feirenbaom ef Stantord Bas yg 2 a velligent computer program that uses kgs compu bt svn incligent computer ProBeam UNat Ses kn ei eee lems ficial intelligence inte’ a4 Hee ranch of art ut Ieie branch erect radar ree enh untae a ree pases athe level ofaiuman expert The source fkaayad wih vaya rth ‘specific problems is called knowledge domain of the expen. a sae Smponents; knowledge base and inferertee engine ledge base contains the domain ss age witch i used by the inference engine to draw conclusions, The inference engine isthe generie -—"~—}_ass_ control mechanism that applies the axiomatic | user _| inomlndge to the taak-epecfic data to arrive at i ‘ome conclusion. When a user supplies facts or . se seca, ensues mos fae co he receives advice or expertise in response. That is given the facts it uses the interen engine which in turn uses the knowledge base to infer the solution, Characteristics of expert systems: A High performance: They should perform at the level of a human expert. Adequate response time: They should have the ability to respond in a reasonat\ amount of time. Time is crucial especially for real time systems. Reliability: They must be reliable and should not crash. Understandable: They should not be a black box instead it should be able ‘explat [he steps ofthe reasoning process. It should justify its conclusions in the seine way human expert explains why he arrived at particular conclusion, Shell: A shell is a special Particular applications. User should supply the knowledge base to the shell. Examph for the shell is EMYCIN (Empty MYCIN) shell Advantages of expert systems: Availability: Expert systems are avails Cheaper: The cost of providing exper Reduced danger: They can be used j i ean hte in any risky environments where hums Permanence: able easily due to mass production softwar tise is not expensive, The knowledge will last long: indefinitely, LP. University-(B1 ty-(B.Tech)-AB Publisher 201 016-25 Multiple ‘expertise: It can by Multiple expertise: Itcan be designed tohave knowlege of a tare capable of explaining in detail the reasoning that il the reasoning that led vo a Fast re a “sponse: They can computers ove hhey can respond at great spec eee (speed due ta the inhers vantages of Unemotional and Apa response at all times: Un Escor Gani and werk aa aily dirlag ances tinea a ee ee Q.7. What do you Ni P; ing (NLP)? E " you mean by Natural Language Processing (NLP)? Explai LP)? Explain its using a natural language suth ae English Processing of 3 Natural Language is required w ae yor instructions, ied wen ou want an ntallgen agen ko robot to perform is nt to hear decision from a dialogue based clinical ‘The field of NLP involves making compute natoral languoges biimans use, The iayut and outp Oe ae eae put and output of an NLP system can be * Components of NLP : There are two components of NLP as given- Natural Language Understanding (NLU) Understanding involves the following tasks- + Mapping the given input in natural language into useful representations « Analyzing different aspects of the language. Natural Language Generation (NLG): Itis phrases and sentences in the form of natural language Tt involves- «Text planning - It includes retrieving the + Sentence planning- It includes choosing required words, forming phrases, setting tone of the sentence. «Text Realization- It is mapping sen' Applications of NLP: 4, Natural language int 2, NLP research would help in pre ‘the process of producing meaningful from some internal representation. relevant content from knowledge base. g meaningful tence plan into sentence structure, terfaces to databases paring texts, reports, summaries for people who need it. 3, NLP is used for 4, We use NLP in inquiry! markets etc. ; 5, NLPis used in language 6 Tris used for business eonferent’ 7. It is used in text understanding & Gener" sng? Explain with examPle- (0 a system that enable the gystem to do the same machine translation centres of airports ation counter, share o, railways reserv the different ty] Q.8. What are P denotes changes in Ans. Learning « task more efficiently next time. : : Learning is an important feature of ‘Intelligence artificial Intelligence oe cin examples of ita input, + Reinfor Spears JIves learning a function from €3 ples of pang, nto . learning involves lea al + Supervised eee noes ves learning patterns in the input when no pong, mpare the r y nsupervised learning inval pare ato ae supplied inforcement (reway, «Jn reinforcer7 feedback than in supervised learning) eee iene on ving we are given a few labeled example andd must may In semi-supervised learn! eoaten fais ‘a large collection of unlabele ve mem RE eC are ey of the learned information plays an important role yas read o sesermining how the learning algorithm must work “ee on i shine learning? e subjec Why do you require machine] : 1 subi « Understand and improve efficiency of human learning. a + Discover new things or structure that is unknown to humans. ee i ification about a domain Pn + Fill in skeletal or imeomplete speci ; — Learning agent: An agent is anentity that is capable of perceiving and do action “"" * An agent can be viewed as perceiving its environment through sensor and acting. 1° upon that environment through actuators. a The oy De De ader cor’ sensors environment actions actuators! ; Paradigms of Machine Learning: + Rote Learning: Learning by memorization, © one-to-one mapping from inputs to stored representation; Association-based storage 'h and retrievel. * Induction: Learning from example; A form of supervised learning, uses specific examples to reach general conclusion; Concepts are learned from sets of labeled instance. he * Clustering: Discovering similar group; Unsupervised, Inductive learning in which natural classes are found for data instances, as well as ways of classifying them, oll * Analogy: Determine correspondence between two diffrent representation that come from inductive learning in which a system transfers knowledge from one database se into another database of a different domain. * Discovery: Learning without the help from a teacher; Learning is both inductive . ey LP. University-(B.1 * and deductive. It is ded ty-(B-Tech)-AB Publisher Kincorema, It le inductive wien lt eee ene a al not given. hen it raises conjecture. liseovers conccpts about tho goal not i ecture (guess) Ivis uneuperciend spe + Genetic Algorithm: et Ie Genctic algorithms search h the spac inspired by evoluti pation; In the natural atural world, the organisms, are poorly suited for an e an environment die off, while th , while those wel-suited for or it prosper. of individual dividuals for good candidates. Th “goodness! of an individual is measured by some fitness, Search takes place in paral by som 'ss function, Search takes pl Mel, hg with many individuals in * Reinforcement each generation, function using these rewards. ke Rote le ol Rote learning technique avoids understand que avoids understanding th on memorizing the ‘standing the inner complexities but focuses ng the material so that it can be recalled by the learner exactly the way it in was read or heard. ‘a good move or a poor one. good mo Poor one. The job of reinforcements learning is to find a successful + Learning by Memorization which avoi jon which avoids understanding the inner complexi the subject that is being learned; Rote learning instead ear Saas as material s0 that it ean be recalled by the learner exactly ‘the way it was read or heard + Learning something by repeating over and over and over again; thing and trying to remember how to say it; it does not Jearn a poem, or a song, or something like that by 1. Learning from example: Induction: A process of ule from a set of observed instances. extract rules and patterns out of massive data sets. d learning, does classification and & tries to induce a general ‘The learning process belong to supervise constructs class definitions, « Winston's Learning program ¢ Version space: Decision Decision trees repret en directly use ina rticular category ™* understand them or ev [records falling into a Pa Description: each node is either a lea! # Leaf node- Indicates # decision node- S} ‘The conditions are: # Attribute-value collectio # Predefined cla’ | be defined (ie supervise * Decision tree is n of properties or att ‘ss: Categories to which & {or decision node. description: ributes having .d data). the target attribute Object or case discrete value- rote learning. called infuction or concept learning. 1s * Decision Trees trees are powerful tools for classification and prediction. sent rules, Rules are eas! database acces: ay be retrieved. ‘a classifier in the form of a ly expressed 80 sxample are to be as must be expressible as ‘saying the same help us to understand, like we learning by example. The system ‘The learning methods that humans can language like SQL so that tree structure where (class) values of example. specific inductive approach to learn knowledge on. classification. a fixed signed must already Artificial Intelligence sixth Semester, 2.2016 tree ple decision Asim CED Precision nace c=yue nye, Haman teen i accomplished by form of hnown general pring,t> cal ee at ieteekground ‘knowledge in the fo! reste ce ied eglatalhe Beet Le BL See sneral concept from a particular training ol ro anid . ulate general concepts on the basis ofa speci fetnine sree hc nro fi ere . is janation ure, nab sce Ra tempers cesta ae eens nstance of the goal concept. The explanation-structure i ve 3rmulating the general concept = . rise The overall architecture of the EBL learning method. Inputs .8.(b) Explain the construction and work ‘ing of Nicol Prism Ans. Refer Q9(d) of End Term Exam 2016 (Page No: 31-2016) 9. Write short note on the following: Q.9.(a) Learning Process in Artificial neutral networks- Ans, Learning in ANN: The artificial neural networks are inspired by the biolo: the hat System, in particular, the human brain. The most interesting characteristi the human brain is it's ability to learn. It is believed that during the learning pr the brain’s neural structure is altered, increasing or decreasing the strength o! synaptic connections depending on their activity. This is why more relevant inform: #8 casier to recall than information that hasn't been recalled for along time. More rele information will have stronger Synaptic connections and less relevant information gradually have it’s synaptic connections Weaken, making it harder to recall. Art LP.U networks can model this le ee = | ee rea of the synaptic conn ote This emul he sensiennand | weal ening of the connections eee ‘The aongening ond ieee ee aes nt algorithms that can be used mre ind of learning alg area ewe the desired out, m wutput for the n on) network. By providing wetwork is also provided wi ig the neural network with both an pend ty ‘while training the nS fan, to calculate an error b: ne aa aca pnd recit doutpat pair itis possi fo) focataate ance asedn tan ata enn Unsupervised Learniny = prork by Pea ae _ sing, the pute provided withou jork’s responsibility to find some kind of pattern wi wef roca coon nd tg) ability to predict a user's preferences by many recommendation algorithms due to their ni hon eroupe ee ‘on the preferences of other similar users it ed Reinforcement Learning: Rei ‘i ® Jearning in that some feedback is given, ae learning is similar to supervises reward is given based on how well the 93 cae a re vcorent reward i en pamize the reward th stem performed. The aim of reinforcement ning is r w system receives through trial-and-error. Thi eal rel ates sere with how learning works in nature, ‘for example an ani al ign erat hor for example an anim might remember the actions t's pre) aly taken which helped itt find food (the reward (6) Learning process in Genetics algorithm. 6) Ans. Learning) rocosein genetic algorithm: Genetic alotuns re onbiological ‘evolution. Genetic algorithms can be used to solvea wide varie) ‘of problems. Given a problem ‘agenetic algorithm | generates a set of possible: solutions and evaluates each ' jn order to decide 8 grtp solutions are fit for reproduction. articular solution is more fit then it wil have | more chances to generate new ‘solutions. Finally we can. find a real solution. | Genetic algorithms are so powerful that they can exhibit more efficiency if | programmed perfectly. Application “nclude learning Robot behavior, molecular structure optimization, ‘automated. "jesign of mechatronic systems» ‘and electronic circuit design Genetic algorithms provide computers with a method of problem-solving whieh is based upon implementations sf evolutionary processes. The ‘computer program begins ‘vith a set of variables which ve ernally resemble the chromosome ‘which store the genetic jnformation in humans. Each genome of these digital chromosomes represents a trait of whatever the data structure is supposed to represent; this jnformation can be stored (5) either in bitfield form, in ‘hich each genome is regsified as being on or off 0 oF 7 i i i i ‘tring in which each sgical| respectively). “Alternatively, they oP be stored in @ character stt ics ‘t character represents a> ‘integer value which describes the magnitude of a trait fos veess) example, it could be number from 0 to Jp, with 0 being a total aBSene ofa trait ani fit’) 255 being 2 total presence of the trait, and all its weights. neural network is only given a set ‘numbers in between representing & fioa} gradient between the 170 polarities. - van ‘The computer program first creates these digital enromosomes a ‘will (random) means: ce pi eT s a : cPhis is a kind of netic selection in which Sixth Semester, Artificial Intelligence 40-2016 J use a model or procedure to test the fitness of the, HRs eal ue for its fitness in comp: See tial camara 800 Lo oth ie urnament selection. This form of selection invaiya” om hvomonamesagninet each other in some kind of modeled environ’, hy serie competition are deomed to be the fittest mi ee 10s the fittest chromosomes ‘ ‘The computer program then takes t _ joneration through the se of some kind of genetic operator. That a hy irchromosomes ean be ereated in either (or both) of two ways, mn + "Genetic recombination. Thisis analogous to sexual reprodictign are created from the fittest chromosomes of the previous generation" "vy * Mutation. This is analogous tothe simulation of genetic mutatin offepring are identical to their parents but have random, stochastic sy structure (and thus their traits are somewhat modified) ee These two genetic operators can be used in different co Producing different results. Using both would imply first genetically oe Chromosomes and then mutating them and would most closely approxima reproduction pattern of humans. Using mutation only would simeg 4 repreuction, in which not as diverse a gene pool of chromosomes are crest genetic crossing occurs. cated ban mbinations, gy Sp. level fy flight i u: compu provid possi c input progr ste ng. df 60! onan incase Aust ame pny lige You can trode 1 sve ary aa hi th ous bate free computation —h Canela 1 world eb vane ampion by brute forve 1 pension acaba million yaelona pa sean eae ese wt soit wera eee Peis Wi Unting speech recogmition of Might yar ie lly wets cca aha heen se compar wing one wer neg Yr we ovr he z nde cui tl net et ii a a anna ‘arsing: sentences is not onough either, Thee stb ae oe wate ivi so seadimensionall abject posed of thre anal, Some seful *V cameras sve two dimen vision: Tho world tuts Lo the human eye and compute they are not as good as what humans evidently use 7 sees Expert systems: A “knowledge engineer” inte! vledge in a computer program f intellectwal mechanisms F ror carrying; out some task and tries to embody their ke How well this works depends on whether the euired for th ow we hi he prosont sate of AL When te twee 4 4 yam, ere were many disappointing r walts, One of the first expert systems was MY IN in 1974, which Fiaygnosed bacterial infections of the blood und suggested treatments, Heid Detter than denial students or practicing doctors, provided Hee eritations were observed Namely its ontology included bacte ee entaents and ih not aelude patients doctors, hospitals, death, tame. Ia interaetions depended on a single patient be tated by the Knowledse ny about pat rar that the knowles ced what the nod framework. In the TAL, this has to be true. Th resystenis depends sense. : One of the most vo of ALis to put some inform ‘information, An examples adv Information 1s available about the own Jaleo sbout tbe item he te Bure and , about whether there have be symptoms, covery, and events occtirrini ‘considered. Since the experts cons ints, doctors, death, recovery, ete it” ‘experts told them into a predeterrt sefuulness of current exper engineers kne' engineers fore present state o! on their users having common Heuris sent knowledg using several sources 0! credit card purchase record of payment ani from which he is buying it frauele at this establishment) tem given tegories hy feasible kinds of expe tion in one ofa fixed set of sing whether to accepLt proposed ‘er of the credit card, his about the establishment con previous credit cart

You might also like