You are on page 1of 56
ERI Semester-7 CS Rit 2014-15 » 2015-16 * 2016-17 * 2017-18* 2018-19 ry rods ntexuston fn History of Ariioal Igece Tligene,nligne Agents Structure Comper visio, Nara Language Possessing, ‘UNTE2: INTRODUCTION TO SEARCH. Teuton Seth Searing fr soon, Uriformedsesteh ities lnformedeearch tes, Loc arhalgoritmsand ‘Spt proses, deel Search, Seach for games Ap ets pruns “rage Repseniton Reser Propceoelige, Tony tf fagtopde ogc, Inference in Fst ore loi, Forward fSuckwordchshurg Resolution, Probable reasoning, Uiliy theory Hdlden Markov hel (ENEM Bayes Networs UNIT: : MACHINE LEARNING @1Ata16A) ‘Machine Learing Suprise unswpensed lesan, Dec frees Sassi md, Lennig tk complete a= ‘Nave ayesrres Lem nth iden dats -Sfalgoethn, entesmeriooming UNITS PATTERN RECOGNITION (S-1A%05-19) ‘ate Recgnton: nkoction, Desig principe of pate rcoguon syste, Statstial Patter recognition, Parameter ‘Station mets - Pcp Componant Analyst (DCA) and Tinar Decriniunt nlyt(LDA), Classi Techniques ‘Nearest Neigh (XN) Rie, eyes Chasis, Support Vecor / Machine (SVM). k-means hi ‘SHORT QUESTIONS (S@-1A 080-164) ‘SOLVED PAPERS (2014150 2018-19) (GP-1 AtoSP-15) cia none (ADs mae orp aces a a gent machines tht worcsadeactntic Oa ‘ius teome an even prfthe tele esse acted with rial intligenee igh tech Seer ca ae cre rllens of artical intalone tea optmelagemytes cerned ‘awl Eine Frotemohine Perception ae lamin “Abide manipulate andor jst Goals ot At: Ir To erate expert aytems The ne hich aii ities TiStbur barn densest eat and aie srs 2 Toimplement human intligrace nmachine: Crating seme Taunt ink, le se har Hama RBI tart cttret rans erica nage? Dace sme ofthe branches aed progreve ade intr el {Ptole achieved by using sw rales. f 1 Language understanding : It moans to uodeesand nat Tengu meaning. Alanguage isa )stemotsensbaving ‘enremign The aigsningy-eonvention it distineive of Tanga oe very differnt fom natural esi [BST] tan Tarn tentin eta 1 The tet conducted for competerinteligencs i known ag he Turing ‘Tis ivoves three partisipants, the compute, a human stern hi abuman a "Te intervgstor attempts to detain, By asking questions of the her porsiapats, wich i he cmputsr All communication i 8 eytoard and sercen "The narogatr may ack tab x penetrating and weaning ‘Deo sdhelies andthe compart permit od everyting pestle A lore wrong entiation, ‘Tho fui ust belp the iterogatr to make coret entiation. ‘oc example fhe human interrogator Roam Cis et sleet sri is troom Aoi loa thes the mohine priceses tlic man or Machine? Haman or Machine aman nterogator 2 FETE oirereatt etncen bomen nsigence and machine aligns scansat Bure out plete abject eareectly one pat fAT began nround 1948, AT is silo, bt on be sce ‘Facial recognition : 1s Fee ee Unmupervied learning 2010): Unsere arningithe tmoperried oing where the core anor re 4. Clustering algorithm + Uso fo find patorn in datasets se {ton wouptaatct oent core cates 4 Market scmentation: Toe cutamars bed on ans Ie dls shen the coum ake puri, Tis tennikrnaned marttng ‘Gonti programing (010) ssa programms ania that Sorereliotry rcs improve iri: ater of F019 werd Tere are ay cal-epovninicking ‘Min elgene Hore ce common tenes tht esta thant acon atonal Por example, the oar ofa ere te [ESR] pescrine ne pptetons of ata able or] Applications of artificial intelligence: 1 "Gaming Alps erusalzlen tate ams sch a chess, ple, tietac te, fe, lee machine eae hin large numberof posse ations sed on heute kale. ‘Natural language processing + I: posible to interact withthe ‘Sompater tha derstand natal lnguegeapke By Romans “Expert systems Thor arse apolctos whic inxraterseing ftvare a spin nfrmation t impart rekeoning ad agen. ‘Thay prvi xplanacion nnd the wes, ‘Vision systomss Thess yam understand, reat and competed “ronal inpat on thecompatee Speech recognition Sane intligentaystems srecupableotheasing nD sonprehending the lnguoge i ett of selects and their ‘Sent a saa 1c band ite ‘Tang wors noe ot hakartd bangin sant cubes |m. Iacteseonding to.» role whose condition matches the erent state, as defined by the parcept 2 ModeLbased-reflexagent: 4. The model-based agont cas work in a partially observable ‘visvoment, nd rack Ue custom 4 Amdl-based agent has two important fctrs 1h Internal state: It xa representation of the current tate aso on erent history. | Theco agent have the model, “whieh is knowledge ofthe world” ‘sd base on Use model thes perform ations tons well ae rtf oa ite eying ievental ead the An age an be rpthing thn pores ta event ‘tnd ctuoua hat evionmant hr cata A ele ofperesiving thinking seine "Tounderetand th structure ointeliget gents, we hol ‘with eto ani gant pen “Agen = Arteta = Agoat Programm Arsitetore nthe machine hat the agent executes on ‘vith aentoe and alutr forsale arate omhow te agen doing and lemons shoul lined Groene Warn an ope mes some fr ofr Pa Sc ictenment then ers me eto change ‘Nuon: Thropn caret meas that ages eb 2etial ates lneryents am misuse the opal {raat hes aias Coupots ot ter sows enema flee ‘pei Ths scat carer mene tha la capes cle Day ve changes within semua Tee abletoacep tilted initiate when appropriate al isalso eapableoflearig [fom ie own exprionea, envirotment and intaration with others. 4A. Soclability: This type of characteristic means thatthe agentiscapable ‘interacts nae to-peor manner wits other gents or mans. Applications oflearning agent: ‘Clustering toes Paton aes cme Setar ea Hesheenil [BERTIE] How to specity tank environment? eae ‘PEAS description is uscd to specify taskenvironment / cial ntaigeee LA CSITSem7 ene cogether oF can im difleeatiated from each oer. Bach gent has thee {slowing properties define fo it ‘Performanco:The output which eget rom the agent. llthenoveveary ‘Ripler hat an agen gives ater processing comes Under its performnce Environment ll he eureusding things and conditions ofan agent fall, {thie Itbascally consists of al the things under which ho agents ‘work. ‘Aetuators: Tho doves, hardware or sofbrae through whieh the agent ‘orf aay aeons or proseses ay information te produce a result are The sevatoe oF the wet. ‘Sonsora ‘Th devices through which the seat observes and perceivesits ‘uvirnmnent one tho essa of tho agent Example: Letustake an exaiple oft driven car. Aste name sugsts,iis.acar ‘phih even ait ot, by aking al te nacssary decisions while diving Tete any help fom te mer uster. nether words, we caeay tht {ivesrdrveraniteersend equees no driver. The PEAS desertion for {ai gent willbe allow Performance : The perarmanc fete fo a of driven ext wil be tht ped, sat while diving (bth th ears tho rer), Ts ate to Actas hv deve anh wih be ntl of he ca areata For exp, he Sern Acea, ‘Branks, Horn, Musleaystem, etc. ™ “ Sensor Al toe devin hugh wich hear gots anetinate about itera nd Mca raw cra preptonset of are entre For example, Caer, Secdometer, GPS Ole Sana, 7 al a Pie eich re tt pi Somucrrashnmansemecae * Sc eet ‘eueterabe othe otra toon 14a CTT Sem-7) Introduction 4 Dynamie/ State: the sovcoment ie hanging fr apa ation ‘hen te environmen dynamic forth gether vata 4 Single agent /Mattpl agents As get operating by tio an Strom issingl gn Mtl got wher other get ace resent: Ouberoget ie anything that ange om top tome. [5 Accessible/Tnacossie: 1th ages sensury apparatus can ave cess tohecomplte stat af thecnroanea, en he event {Scr to tht agen ober trioeesae & Deterministic / Now-deterministc: Ifthe next stale of the fovironment is completely detormoed by the caret sais and the ‘Hinoratheagent, fen hc eernent determining others seer 14 Bpleedie/ Non-plsedie: 1 the age epeions i ied ins “nic todos an ineah peer get reeves epoca and ‘henyorbrns ase aco, then tncomveormont episod heros ‘Niseonei “oa oR = omptr Vin Neat Lage Intligence. eee] 1 Computer vision ean nterisilinary sient eld at dans wit fw camaters can made co gin high-level understanding from. Bk ugnnor idee Frm the prapetiv engineering oka [Elona take thst the Boman ual pete eso, 12, Areasorticl italigence da wit uineroas plana orbs ‘Scns te navigate through a cevieren. 4 Adetied undertanding ofthese environments isroqureto navigate ‘hou tiem. 4 formation about ‘seompator visio yates tu envcoument so sho rbot end be provided by ting eo vin cepor ond prow igh LeT Etemation about the eviconment andthe robe numa develope eglatnasnd leno 4 The fundatinn Und the opproach ein generally regulations af speech with spe I [naerlied and weirdod experts 2 Staten! approach = ot {LThestatiteal appooch to natural language processing i tn observables persia arp of na Fe | Mode whieh ae bod on ation enti persatent theres ‘hvouh matheratiaiterpreistion of th age ex ‘i Byres trending sue of et he ‘develop ts om sti aut wl ua erp future npe vrables andthe davelopen flange outpt Connections sppreachs 1 Tho crnectionst approach to aataral language pcearng i = ‘isla of he re and saan appre Tis apres: stars with gency approved rules language Ad emer the to oes spin rb apt posed ‘Sool nonce [HHRITIE] Write short mate on conceptual dependency ‘anseptal Dependency (CD ister afar guage processing ‘chal das wih represent tensa Sf sasguas aowledgeirereertedin Cy eeent ich slow an conepual_ Consol pring ia strategy ford bth te strut ond ‘eangfasnateicetn on sep ‘Advantege af CD 1. Thee prinitives nave oer neem rl, 2 Mary inference ule ru eeu epreneied CD strane, ie AIA CSI Seq, Gs itl strstr eto few on te Pins toy, Disadvantages of CD P Kaowiege must be deeompttedint fury low level primitives, 2 tisiaporible or diticlt to find corect et of primitives, 13. Alotofiaference may ail bo required. 44 Representation cana comple oven for relatively sme aeons, ‘GERTAIT] What i the role of NLP in AT? llusteate the vari (ARTORUTE TO ary Role of NLP : Refer Q.1.18, Pago 1-15A, Unit. ‘There are five phnses involved in NLP: 1. Lexical analysis It inveiresdentfying ind ansiyaing the strate ‘of words: Leaieal aaalyeis ie diving the whole hunk of tet ao paragraph rentences, and words 2 Syntactic analysis (Parsing): 1 involves analysts of words ic the ‘Stolence fr rammar and arranging words in amaanae that eos ty ‘lationship among words. & Somantic analysis: It dams the exact meaning or the dtioary ‘beanlog tum Te txt. The eu is chesked for meaningtuless i dove mapping syalactie structures and objets nthe ask domain 4 Discourse integration: The moaning of any sentence depemis gen ‘the meaning of te eestenc just bare Th adden, als bg about the meaning of tho sentence that lows & Pragmaticanalysie: Dering pragmatic analysis, what was ait = ‘neared on what actallymesst. It ielves deriving thos asp of language which rouir real world knowledge. ABA CCST Sem 1.1, Deseribe the role ofcompater vision in artificial intelligence. ‘aac Rotor Qf 2 Deseribe the role of artificial intelligence im matural 3. What Is intoligent agent ? Describe basle kinds of agent programs. same Rotera. 13. iscuse various approaches in NLP. ame Reer@ tt 5. Whats theroleofNLPin AL? Mustrate dhe various phases inn? ‘ame Rotor a... @6. What are the different branches of artificial intelligence ? Discuss some ofthe branches and progress made in helt ‘felde same Fofor@ 12 980 EET ete eorehing proc on parameters used to evaluate a search What are the different technique? on List the criteria to measure the performance of different search strategios. 1 Searching isthe sequence of step that transforms the inital state ta the gol tate. 2 The process of search includes Initia state description ofthe problem. Aut of egal operators that change the state. The final or gal state Following are the parameters used to evaluate a search technique 1, Completeness : By competences, the algorithm finds an answer in fome Bite ie So, thalgorithm isn tobe eompleteifitis guaranteed torn: there son, 2 Space and time complexity: With space and tine complexity, we ‘breathe memory required andthe time factor in terms of operations ‘arr out ‘8 Optimality: Optimality tells ushow good the soltion So, an agri ‘ratsearch process said to be optimal, fit gives the best solution. BERET] esrive the type of sare strates. searching tree. 1 starts at the tee rot (or some arbitrary made & ‘sometimes referred fo ava search ke} and explores all ofthe ‘neighbour nodes at he present depth pier to moving on ‘odes atthe next depth eel. Algorithm + 4, Cheats avril aed NODE:LIST and ot tothe ntl state. ‘2 Unlila goal state ie fund or NODE LISTisempey: ‘Remove the fist eloment fran NODE-LISTandeall it IENODE: LIST wasermpt, qi. bb Baceach way that exch an atch he tat dsr ind 4. Ayply the rule to generate anew tate ii ute new statejnngoal stele cut and return ties ‘i Otherwise athe now tate to the end of NODEALIST. 4%, Depth First Soureh (DPS): 4 Deofh Fis Search (DS) sn algorithm for traversing or searching ire 12 ‘The algorithm starts at he root node selocing some ain ode prot node in cau gro) ond explore asf a8 [onribealoageach branch before eckteaing. Algorithm: 1. then eate ea gral state, it amdetarn sce, 2. Otherwloe, do the allowing ntl sacess or fue signaled. 1 Genetateasucesor Bo Us ntl eat. Ifthe are no more soceesors i fale, ‘a. Gale iret Search with asthoiniiaLeate fences returned, signal nsceas, Otherwise continve in thi Iie GTR wee ie adenine an advantage f HES and [BES is more suitable for etc wer tothe ven vous, ee the current noe, Ofte ‘Ta tie complexity of BFS is OV = 8). where Vstands | OU 2), where V xtande Taryari and B sandr | versone and and oe ears ot production system. Production system 1 Scurch rms the cone ofmostntlligent poses aresrotied inva ready al peo a ‘ process becomes ey. Production sytem provide mth tructures. ‘A prociction system cunt of: 7 ach sence in production systems unigne sthey use" THEN” stricture provides simplicity inknowedge representtien ‘roduction system improves the readability 6 coletion of ndapendent ts odie frm sues vices Ti pa aes ar uta any pe “i Bash production rlee normaly written asa the problem of semantics is colvod by the et repreveatation Issu0y involved in the dexiga of search programs : 1. The direction in which to conduct the xeareh (lorward oF Dackward reasoning) The tre can be searebed forse fom the | “kal odo fo the goal tae or bacinars Fom the goal state te Ii ate 12 Hoye to select applicable rales (matching) To let spplesble ils itis enc fo have an fiat pracetore fr ratehing Fes | How to represent each node ofthe search process: This tho ‘mowed suprecctation problem a tho frame problem, or peo like chess a ode an ful representa simp ‘ty, lnaors compl problem sling, however, ts posable fo represen al ofthe fat ‘Whats heuristic soareh? Give the deslruble properties of heuristic searehalgorthon 4A ural unetion is fontion that ranks alternatives in search lorithme at each branching step based on available formation to ese which branch a kom thangs rom depth bread onanalyis these re ae etor thn the rarch method evel "stents te Fetch aed proceed the fo the formal defnition af poben {Wore af the nodes hove generates nodes that areas prom nal sate othe onl tate gie |e t che ot oe by 83) aque fame board wha 10H One equate omy alow ie the pul Sth as oat Sorting configuration to 9S 4 2s | Planning agunst us “The environmeds ith ore than one agent ite vironment in whi each agent 643 ener) G00 ° ered ‘Thum alower bound onthe cot ofany tour “he ah and wud sear satay apple o poles (Gaph seach space where ire tha aw atratre pth Fetwens eo nodes, This sentogy saves all path lengths or costs from anode to all generated ‘etd hoarse srt pat fr farthe xan @ preset position with PINAL VALUE - the reach ma tree we aes 4 5 “Masianer goes LER 1 a peers iekeee Being he mxiniter he wou! chaoe the ager value hate; Hoos Xe optinel move fo tbe musimer in tage LEFT ord opti sous ? ig 2.182 shows two pssbe sores when maniminn make eft and ight eve Give the production rules for travelling salesman. problem. Consider that four cities A, B,C and D are given. 41. Toole this problem, we can apply the nearest alghtour heuristic, “which worst seleting the lealysoperiralernative at each step ‘We can prodse the flowing procedure 2 a. Arbtearilyaclect starting ty 1b Tose the mest ty, ol at al ies ct yet vital andslat the ‘no loscat to the current city. Go tener Repeat step buntlalleitis have been visited. “Production rules Sa iy A tote (Dao) i cyinCtoa Te]. ice ctyisD en: tofmcety A FE ee 10" resto -AO* algorithm = 1, Inlilise the graph to start wade. > me 4 rare ro ileal ore yak re eae ‘havo not yet boen expanded or selved. ee HR Fed ey ole mois acpi ae _xalue FUTILITY otherwise caleulate only /”far each of the sueceso0rs. 4 Affi hen mart noe es SOLVED, 5, Chane th velo fhe ytd mda oes i rere Syleet prepation 1. Werrerpenewe hoa prone re if si dl IISOIVED en nak hope men SOLVED, 41 tating aa ixSOLVED ovale reir than FUTILITY tne eee (BREET neptin A0* grit with mtable example Sate the imitations inthe aleoihee. —— (AEESOTETE OTT] ‘AO™ algorithm KeferQ. 2:22 Page 2-204, Unit, “Limitations of AO” algorithm: : Te elgorihm inserts lithe nde’ ancestors ito hue, whi ny ‘eit the propagation of che ent change Dock up through a Tere umbec of pte whieh at 004 12 Samctimes,iecannet find optinel pth for anacvable nodes. ‘Geeta rapa he conscnint sittin rocoto sare ‘he eryparithmetic problem. OS ONS ON TTS] we] ont saisacton serch rede tat operate na pac ‘Autre al sate contin Ch ores tae orinaly ‘Srennthe problem dosrion ‘he enral oreo enntraint snefction posed: 1 Sect an uoxpced ode th earch rh 2 Tvgeceatal posible nw snatsits, ply the conan infec Fab Gackscd ent the et ofemnstraint contacto, than oper that thi ‘asin doen ‘Troon evn, a fesse nceptraton deeibnt oop niatn instr ontraition or acomplte solution has bee oul then ‘apie poam spss er peer a aril ‘incon wih tw voces comtrnts Ime these pertel “olmons int the mare ap “Numerical r Constraints: aden wey, "Tees reansvesponeh explain be peobioms: ‘snowded ond osticcomiousin osteo ica srematienavof tering aut dite picncest swofind one that works [Tt Inconsderod on nprtant ee igus to sve cata alien ‘eas und purl ie uo considered grea oegue or peog ‘sod ale forms tha bso many le pogtunring engage BEEBE] wy i gameplaying god cantata of AL? Hain 1. Agpod candidate fo Aha following haractoritice: 4. leeontine lage amous of dma spesfc knowledge, 4 eontins computationally complex probien, canbe developed ar raptors fr the knowledge of coveral enpete ‘et usconsder an example of ume lang, an intelligent aystem that iy ssid toate ican (020,(4¢+8)= 10; Seanbod! Batic, 1 The roche ar ay to nar, et ay game aan ‘0c nt et any er ae ae pe 105 poe mont chew tty re deraaen, stutter te legen etdoman sects ‘Tose pou saeofehs hreemuly lar og svt aw thes psc pn kage ‘The chess program inhi asd on things om several hens player It has enormous aro ca ‘doma-speis knowing) tate tb a porte te deisog ‘aking proces, Thi aati th hid dhatacterctes oh ‘undid fora ‘mee ne cen say that ne aig gwdcanilae or AL 220818 iitottiemse “FaeTHE ve an eam tenet Wht he Peon agumetsco? (RETOBOTETE Ba raphbone dese positions in gare agg sminimax procedure al ee cn toni ee sacar lesen iii ag 2 Ba cen es tah pl ee eaters trates mainte ae carmaker + eng a neni spc arte Reema ere \ 44 i\ AK RE XaaRE ie ABEL Te el i ee ce Taunt esos in the cmli une eee ee nob of sible diferent waysthe gam can be played. we 7 For example th gue tre fr tic tactoe has 255168 lat nodes. Purpose ofminimax prooedare: 1 Per may ‘pasa ur gol ‘pore Somplex mins such ap eh 38 chess, search to termination i irda eannt be obained ecng mck gare tee might bo to fnd a good fst & Misged fates Petpese of ve canbs erst by miimy ae proce ymax pveedure. Thisis the duteinagaine ten ee | JACSIT Sem) ee —____ntractionto serch nie eh [GEIB naotnn Win.tarprcetes, Perot thrsdiernstiasant toimproveits performance, ian Snax procedure: 1 Tapmieoneagein reine Zoe for prgrenan ter 2. Theninseeltin opt tn fn oe Te 4 Tease simple eur computa ft mina yeh ‘Min-Max algorithm Refer @ 229, Pegs 2-196, tne _Atpha-bota pruning: 1. Alhasets pruning samedi wesc ofthe in-car is ‘an optimization technique fr the min-nex agora, 2 Tomin-maxsearch algorithm th aun fede guests thet it 2s to examine is exponential indepth ct ince Now ne camo sliminste the expanent completly bat re anette ‘8. There isa technique by which without ceding each ode fhe pane (oae we can compute the correc min-max dso th tens called pruning. This ivalves two theshld parameter lsh and eta Torfuture expansion, soit istaled alpha pang. 4 Alpaheta pruning ante apled at any depth fare, ard vomiting ‘not only prunes the relevent alin entre abiore |S. "Thotwo parameter canbe doi ‘Alpha: The best (highest vslae) hice we ba ound sa ‘any pint along the pho Masini: Th inal val ofp 1h Beta: The best owestvlue)chice we hae found oer a any point along the pathof Mlalaiser hein valve ofbtais Moditications to mia-mas + Thre ae sone heart rch sete other than alphe‘bota prosing inethod which are used to improve Derformance of min-max procedure. Tey a= 1 Greedy hill climbing method 3. Atiseialimmune algorithm Describe alphabets che Min Max procedure waned et nd leith pial Brlefly describe the meaning of knowledge representation and knowledge scquirition. What procedure it nlowed for knowledge acquisition ? Explain, Knowitge representation twee repesnatio they of wae of pitta otc ah acl emcee repens of Pere crea aed rerio ate Horexamp eet Chace dog Representation et wing mathematical oc: Dg (Celie 44 Kaowledge epresensation shoot possess flowing character 4 Representation scheme abrold have wnt of well ened ea ar 1 atould Rave gn expressive cost Wesabe effestive ‘Knowledge acquisition: 1. owed sulin is he oes of sessing Knoles Hoorleige Rope ein ren tte SSAC Seq, |, senantontnutioem aiegious ee eet Procedure for nowledge aesuisit 1 entiation Break he prolons into parts, Frame representa 2 Gonceptalisation Identify the concepts, Seite adr geht ttn ‘& Pormalisation : Represent the knowledge. | at ae tbe anti inthe won 44 Implementation: Progenming. | > © Bape welt Ale rte wid tod [5 Testing: Valdateofhnowedge. | perctrcear es estate ermine |” Si Tomi of colstinof dats nd dt ais The dt ofany type and ses. Ss hae aanes ante aE fear robes lad Ei |" 4 Prodvetion rules syst const fe ‘mea, "IFeonditon thea eto ain atin pis ih t i 3 roa ee 1. epoatin alee rena eens pose 2 [ae |r se | Separate teen wie | | omit) | widge| | aE Thocondion pat ofthe eerie ic yen toaproblom An theacionpatcarranatiecsoeeeae, stad Fealon Nee solving steps This complete poste sled arseqaint oe, Cemspodiaise Tplenesttn BEET] eerie the techatquesctinowiedge representation, | ‘Techniques of knowledge representation \ 1 Logical representation : Topical representation alanguag with soma conereto rules wih | -Agod stem forthe repent o knowin per dou ‘held pout the flloring our popes 1 Representational adequsey he iyo ovat alin ‘wieght ee nde nt oma 1 aeretremeicnmdtaremby npn ‘Tafel adequacy Tommi team > eins deceetaten bene an verte ‘new knowledge inferred from old. ‘ 3. Inferentii jedency :"The ability to incorporate into the knawledge ma a eaten pao - be transleted ‘of the inferenee mechanizms ia the met promising directives, ae aes + Acquisitional iter Ta iyo emantic sc Ives insertion, by apereO arte PSterowiec | Saket aicoesnem tu networks are zt | Ls represanttion, of predicate loge knowl? us woot saying ae ips ews tat | 8 This networ conto a Tecmo ening jt aad arch Rowesiteof proce defied aie a syntax and seman which appa ——— ee farious inferencs rules: an “Modus Ponens : The Mo Ponon rule antes thal P = © Pos Gis trae hen we eon infer tht ib oe | Salant 1am ses then lento 1F-+@ [tterent2:-1acsleny 1 “Cetin 1g abn (Q) any that iP Qiu enim “Prot ruth tale: Begin extn orto der lie ssi xine tee tb Aiuto sitementeinnconcne way rier og ik natural lng des al iy a et he onan ts he pop ata ca ons function § Asanatural language, first-order logic also bso main prsalas ‘and vomantics, GT sepa interne ae ft ong atrence roles or rt order log: a 1 Uateeral generalization . “1 Universal yeneruization aval inference rule which statesthat ‘ifpremise Pic) is true for any arbitrary elementcin the universe af Aen, then ve sun haves cian VAS Tisanerepreniad a Bal "ve Pw 5. Thisrulecane waif wana bowtie Slallr pope nn instantiation “ | Uniera instanton alo ald sie “Savolidinfrearerle eae senrencee foe esi, x) fod apni “ood (chicken? yea (yelled (fo in) Bet Peanut Bil Sh Avec a... “ cerule in frst order ose BERET] trast towing etc in rma nena Jog and canal orm: ‘ a shart i ood fetta ome eats and it docs not All food © ai ens everyting that Sue oes “GL. Kin eas penn and ial {Sohn wil ary Mary tary Toves Joh Kever_| Predicate logic: | FroMuton! Dh vs Vy Balaey) 4 Kies Gh >t) “© Bata, =) enta CRA?) “Bate ipeanuts, in) + ave (5) 8 ie ¥y Marry t,y)-> laren? foe can infer that thee he property P lass membersip wih cay mth "x Dogs 2)» Bask Bir eee os a 2 mal 2)-> Dog ‘contains representations hat on FY one, whose Fest argument ae nts aclins to which the object belong, then I will understand the Therefore, I will 7 bt weeanesplckSApedint: | Gagne nt preteathenfwifenaesod ieee whale night on this problem, then Tean solves. fase the reprosontasins that ne eth te Bethe oliowing sybils or neh sentence 4 iflwork whole might on this poten. the prcblom, then will desta then © Tt ork bol nigh nti problem te wl corteeponding tthe arguments Gop B 03420 entance eval we ey t fad nn sats (A 2B) vB =O), Palen = ut To anit (A >B) iB > when es ir rented. 5. Foresamlesuppece tat berate ‘Bnngien tt heros andes iy sa The flowing br rae Were and cats fle Then Xing, Winn fiog Then Xie re * a sacked chaining ‘ura chining starts wih st fg aye kw toee fee dat alah wl oo tainkrence engine we backward nog wo mar ‘Rls unt Hodson wc bs Then cose at mate ms {Fe Ieeuge fat inference rte eat awn ee tad othe of gals or cxaple, spose tha he ols to anata Inne git that ecole net cna then ~ tefllowing wo rules: 4 UXronks and ets ies -Then ea Una frog Then Xie re TERRE ws Son ss rept ale Properties of forward chaining: Tess diagnosis meted | Teucestop-down proces | Backward chaining En Ahatsupprt agen psi ‘ara sng [iisetiiost search a proof procedure that rproceses inva edinreasaningith ion operatesom statements tha have , mei have not ben cove vnc eth ne ee niin ap ee ein Gpotse contraction tbe awa tis tae of resolution in predicate log Shon ran nfrnce ep rie ho acpi isee i “system for predicate logic inelause fora, 1 Aplin lef rstin th clan ah ei Ee hgh table mpi ‘ FREI Dion reatation in ropontona and predete eslution of propositional logic In rslaton of proposition le the procure for producing a proof by renaton cf proposal S “epactn ast of som Fis a allows 1” Convert all he propositions of Fo clause for, 2 Negato§ an convert the result to cence form. Alt isto these of ‘Isives obtained in step Repeat uo ethers contraditons fou or no progres case mes 4. Seles to clauses all these parent low, esol them oyster The ruling cause aed he essa ‘il be the dgunetion af a of ha ara of bh te re lauses with the following excopton «Whee Te any PES Tiere Zand ~, ach atone of the pare inet i ‘other ~ then select one such pir and eliminate bth from the rsoient. ‘fehe rosolvant i the empty clase then ound Tei ent, Un adit othe set of a rosedure, Lf willy theory Which sone ofthe i Begotten le ith the tera man) wie that for x = Marcus at 0 model oe te the HMM PPeeOh revecnnceravoty peach sen) ‘models to model speech a ofpeech research isthe use of datateses ng sneechint ext is between Markov Model and Hides Morn ([RCREICMERS ‘modele are farily its various 1 | HMsate asec mayo fool. Tes the | Markov models, Its the pplication chow to | HMB defines boat howto kar | “snot | the weights the Markoe sa wate | ln aHidden Markov anil u stato ate onto Ose probit. _y now tbe oats. ‘Therv ar hidden statesin atin Fig 3:29. 5 Inmachino losing sense, oboreation i ou iden refers (Oiwervation refers process i shown by the intent HibeBig 329-1 and each ofthese are ations are know data end refer toWal Qld oss , = (0,1,..,M-1) set poral |A= tate transition probabil _B = observation peobabilty ie inal = (04 O42 eet dterine the iif ra fps oerrid of aad Ca feather ba theo ster: ug, Cloudy, ad ai Fe elas othe prone: Wetor Or s 2 Thi ior rye a mie “Toye cu take greet y pouring wes om ties Cages at to prnaa we. Tica be watcad EEE A feemoon- Feat ieraning Sly egal Unig the Bay’ ul, we can dads the met contig tae feresatie wet ra a Gitte wae Besant i Bayesian network possesses the following merits in uncertainty. representation: : 1 Bayesian tock can envenientyhandencmplee ia, 2. Rayorian network eau learn the casual relation of variables ‘aac elton flr el held ras Ss attrac pettonrenad : ec ontinatin fea network an baesi oe eetintn tia fei nowiege nd nisi pas 28 eto / roan roils ay, nal pry at eed ei ey a. fad aeting in the real world i ae Frac, Seer eth these dnsn uve i Sword appieatios. = : aaa lanl te asta et wor a ply tig feria tra inchoate be. es 2 Ingenta ein nied i E a im enn da lon pe ]Fxplain he method of handlingspprosimateinoene | fase cmpstion oh ia e 4 Reiners regen alin Multiagent planning ie necessary when tere are ota ages the ‘eavronment with which to coperst, compete orcas. {tounacenplabs computation times becuse the th densely connected approximate inference: This ue the network to geeratt onditinal potty dictrbuton Pet and onal probabiteofinterost when the amber large, This mothod expres the inforense os al optimization problem and then Fad pers oft pebble of nterest by sana concerned with the decigs aj the aystem to Team roma he that machinos ean earn fp decisions using algorithm ge desig inh ayy automatically. oy Idea that mochin ‘om pat dats, ident pater nd make decisions ss lgorthins erie baring sds norm weal oa ‘Seve ia rine by poviing i lp an mati ug Ee te th ordain sep cn oeirece rercatet brio eee oer cornu wih he ceeds eee Ea tran Ep Sre« rbien ot nipeas caine see ee Determine th pe ofan ee, We cicogate wicog fe Detrmin he nel care resect Er Determine aacerofc ara ao ee tern casi fe Gonplttho dela Bupervived easing can blaine 1 Cnnsicaion i open Camperiae earning 11 Unoupervised learning, an output units tainedtarespondte clusions 4 irate a Tnths metal ttn hen rear iil year a out the reefs dice patna abet froeplonks ote whi gop amenberbng INomber ofelasses are ont keg sult. court, ball, opponent, and os system and proprioception. inmorving the ball nt sit ———______ | DModerate accurate and wake os ead te pr dh ht oe ‘Reg decans/ predictions based upon daly lerningte relationship “Bioen inp ad output variables ‘ | Taragessinn, the relationship btwn input and ouput raise recommen ceeeasere Feqessionis no the mast rportant sttictial and machine earning tos egvsslon is frequently sued fr prodietion of pices, economies, | yarstions, and 002. ends atest ofthe the ae aelaeit age value bo returned ite, to Walt for atabe st ese, forthe goal predicate Wit. detest Prone aan, Ths) innate by hos 019% Nie hen cen ie ea ai indefned ete redwtion ner. The Hacer of raining data, §Isdefned as ie HS) = ~P, 108, P,~Py 108, Py sxhere, ps defined a8 the proportion ofthe raining dt Tne postive examples, nds definedas the proprio that Jnciades negative exaple. ‘The entropy ofS is zero when all the examples ae postive, or svhon ll the examples are negative. g The entropy reaches ss masium value of when era halo tho eaanples are pti andl are nea ‘Te information gin ofa pariularfextre teu how cls that featur represents the entire target untan andan leeh tage the feature that ives the highest infrmaton ain rchoeen tozsra toa question. Giniindex: {Gini indoxia a metre to measure haw fen arandomlychonea flement would be incorrectly identi 4A Temeans an attribute with omer Gini index shoud be etre “i Theformla forthe calculation ofthe ofthe vind: Ginter ='1-5 af PART-3 mont probable fiestas” te oven paras of parameter ‘tae yi tt adn Gained totes ute a Sti decom: foe I eats eo ia ee sero to sotisty the condition that by, is maximus, i, [fees FREER Dace earning with compete deta derde mde ‘Naive Bayes model, continuous model. Write short. ae ea Cea aay aeimum -kioo parr eran 4 Continwous model. Discrete model: learning the paramere gun den aaanigeeeis ire 1 Ro- gee 4 Mhepuesoetersof thls mod are thereat andthe standard devi 2 MPtbe observed valuesbes, ... Then tho lglikelbeod is BE teas pert = Noe IR togo)- EM Eating he derivatives to sre we obs, “osind the main sith respect to 0 and st the eal a a: LAS iy tao ‘hemasimim ile valu ofthe aa tbe semper aad te Sid value of th aodand aviators he equa ‘porafthe sample variance Now ender linear gaussian del with ne eotaunus prt tnd eontinvous eld has gnsandstribtion whe mean depelioeriy ote rl tend ioe sianard deviations e 7 To ean the cnditona distin PO, we cn simi “raion ined, emt Poin= get en te paranstrs are 0,0, ado, The ai te tenn of 2} par =a onder just te parameters 0 an tat otcens andy bcooae ear at aga cbmerved rewardsta ler nt, Anoplizal policy isa pai tht goad and what is bo, the ett ble inthe fel of obi, he ‘eomplex enougt 0 einforsement learingiall about making decisions esuentaly. Wecansay that ‘he output depends on the ‘areofthe current ipa and the next Input depends on the outpet af the previous Tn sporvaedlearaog, he desi mae on the ti Inport put em th sat Th reinforcement learning, decision is dependent, So we ive labels to sequences of ependent devinons Te sopervied learning Oe decisions are of echothersolabelsare gente ocho ‘ample et ost PART-1 e a Pattern Recognition: Introduction, Design — ‘Principles of Patern Recognition Set GRIT] Write a short note on pattern recognition, Gees 1 Paiternsesoniin isthe process of recognising patter by aig ‘machinelearsing algrith. 2 Patter recognition ene defined a the classification of database ea ‘knowledge already gainod or on sttitial infor satlens and/or their representation & Ine pattem recognition application, the raw data is processed sd converted into a frm that is amenable fora machine tose ‘4 Bxamples : Speech reoogition, speaker identification, Maltiniis ‘Document Recognition (MDR), atomatie mediesl diagnos. 5 Pattern recognition involves : 4 Glassifeation : In classification, an appropriate class at! assigned to a patton basod on an abetraction that x geasraad sng ase of reining pattern or demain koowledge. Classis ‘suse in supervise learning 4% Chesteriugst generate pantition ofthe data which helpsdesson ‘making, the spedfc decision making aetvity of interest U (Chustering sued in an unsupervised learning. 5°53] Write down the epplications of pattern recogni “answer Applications of pattern recognition: A. Image: Segmentation and analysis: Patten econ 5: ose to gue human recopation intelligence to machine Which "eguired in mage processing ion extrvetd ice r EE a SAC # I Se 7 pute vslon Pelton rogues ne computer vision isttdio caren Camps rom ven imago'ies cane oat nen opt ga kd ert Tatler recogsition SpProsch se he en stintereatin of oper ptern anc geek 4. Sasi patternrecogtonsinpicenes nee ‘ypesot seismic analysis model sod sein diferent, padar signal lasifiation/analpse Tater 4 Refeingmeliods ae we In eros apne een ‘nosificationslike AP mine detection andidentigectng’ ee = Speech recognition:Speteh rosin 4 PPepeech recognition which tries to av ‘i clama level ef densi el tat edn various athe ld the problens of cg 8 Tacgr ais eh os word pave. 4 Finger print fdentification: 4. The ingorprint recognition technique ina donna ecosagyin {he biometric markes, von iA number of recognition methods have boo wed to perform fingerprint matching oto hick pater rset pps inwidely sod. [EER] Discuss the of ification approach of pattern ATT BT Marka) secognition. ‘Two basic approaches of pattern recognition: 1. Statistical approach (decision theoretic approach) : 4 ‘Thoatatistical approach based onthe us of desisionfunsionst0 lassify objects. A docision funetion maps pattern veetrs Xin dessin regions of De ‘The problem ean bo stated os flows: Given objeets o= fo oy 0 letencho ave kobserrable Attributes and relations expressible ns ever, Ve (Wy By 0 > Determine i saubeot afm

You might also like