Scribd
Upload
9 views31 pages

AIML Unit-1

all the best

Uploaded by

chandukaranam64
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
9 views31 pages

AIML Unit-1

all the best

Uploaded by

chandukaranam64
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/ 31
Unit =| Ay siftctal ntelligerces ahi ts a bunch ol cornet Dene nice mate machines “to “tink (O17) behave lhe burner Called Ay “Types ob pa AX can be dtuided fnto Uloxtann “Uypes, Anere ane pata —woo ee © Baned on capopiltties ® Based on punatoratiy i \ Based On copatlity + 4n thts, wle have 3 wpe \ Noviow AL te > “ie 8th nome oh novnow AL > uleah M1: Narrow pt Connol peroren Spend tre fold oy Uterfilations « 4 99) “tained binly ye Apecllf ‘Lok: i : nes ce > Cxornpless one plorinq ches, & commer t | Duper AI +- > ge tsa type Oo} inreltioence Eeneval_ M1 +: > & io & tyre Could pehorm ony tor -eotlh ¢ pickera Bue pe { hb { burnon- —» de takes ot off .etports oh itetigence tinge “to | develop. upintch “Courd Pexyorrn om: {ass with ajpictency thove thon humans. ia one “tasty human . (9) rip = ae “hun, 2 Based “on fyunettonatty "Reactive Mochines => Pegetlue Metines ane ripst basse types oh ostifjtol Sntetigence : = Peoetive Morhines clo not tore memories by past exxponiences or [juterre ie ~~ “these machines Otoves only cunnmt ilua —s Dann Example ane fjoegte Alpha Js. a & Limited Memniy =. = Lintted refnny machines can Stove past . eacperience and memmies: —> This + machines can tue “Examples Cae tb ana sme 3 = of Mind s- . “Teoy off Mind Can understand “the human emotions”, human's peelin 5. > Tis pe of: Machines ane not Chveleped . 4 Ah - Auwarnenen §- . > Aak- Awareness fs the future of evita intellfgence —> These machines eoftt have none Kinciedge Companed td humans Machine Leaning yg Machine aa ts a grouting teh iL tures — Pant later : @ gt ta a Subset hor AT 3 gt unes historical data yer future pre dfetions: uy Application» ane imoge eage eptech secognition- ‘ : > ML tNap intiodiced fy ‘Anthun “Aamuel’ in 1954. . bh tacuing oh Mit input past “Totoig ]M Ba data lalgoyithm 2 i] Featuses :- mens ML unes dato to detect Ucoifow» Potten & dunes part data 3 4 is Aimilon to data Mint cleals tpith huge data. q Ogep Leaning | Deep tening ’s a Aubpfeld of Mt that deals with al vfthms- 2 4n deep Leow We Use deep Pr netuoort to train a model. Need of aleep at farectss of Auge amount of data & Patform COrnplex algorthens 3s. pub Performance “tuith: login amount of abté * Fettune “extraction ~ Prplfeations :- : io 1 Concer detection 2. ~ Autonomous driving Cons 3. ~Tranplatio u Muate i Oy | @ 3 [piewion |——>(P) our wy hw) —> Other narre 73 deep speural networka- cK _ “Tyee of Machine leaning ystems we have 3 types oh Mt ' Aupeuised Machine leaning ¢- ' 4 fs a type op Machine TT & dn this Leonid machines ane -trained by using “water “Labelled training data- 3. “The labeled, data means Some Input dato fs Already ‘tagged ulfth correct * oe 4 The atm ofr upentoed leanniog.” ‘fs t0 Mop the input uantable (x) tuitty output NE: Loring of Auperufsed teanning Data — Trcin data +s prediction — ofp a Oo ts Tralnfng — en > if -the ha f pe have jo Aides ¢ gy Aides ane egual ft ts Aquane > fp the Abpe have “tinee Sides FL 5 ‘trtangt > ip te ‘Abope fave Aix des Ht *5 hexagon. 4 Aruperulsed — leamiry ieee 4 ) ion - Keegrecoton °5 wwsed ther -t is a velatienship blww inpat and oops Uaniobl J a Te gmat ase s- Linear Regreeston Non ~ linean Kegrecdion 3 _ Pagpemia tem 2 Classiations- class)jeatfon “Cosi tue used then Output vantable ‘5 ca eyoretd lite Yes] No, Male/Female - ; > Closefiation alge ane te 1 Random’ froreat 2 Deciston tree 3: SVM Feduagtng t= ' Can predict the output on bans op past data: & te can hae fdea about classes of object Pisaduantag cot: | Not = Auitable for complex tanks 2 No correct output fh “test cata and tratnin pat fh te mtg ! data ane diferent is “Time Compleatty Application’ 2 (Health cane feetat|, Store: 3+ tleather Ops & Uns netupenvised Machine leaning ' Unsuperutsed Lenitning » 75 ‘type. of ML fh Ulhich — models. ane -trained usin “unlabeled dato. 2. Unlabelled data, meons «le have tpt data but no correspond) output data. 3. The aim of tunsuperLised “Leow ig MB Tepresent the dataset fn patapessal format: pel ff Unaupersiised. (eoocg Unlabelled data —» fntrepretation —> agettnn a prssteotny => Otut put pes 4 Uni perufoed Machine teaming 7 Clean clustering fs a method %h quping -the objects fnto Clusters. > ~The objects talbfeh one in Aame eluster have i Aame properties: => Tk objects tubich ane mn Aippeent Clusters , ILLS have difperent peopotts. er Linsupera sel aig method. +S Usedewsty fin d the relationships blo ‘Uaniobl 2 Association: . Sociati®ns- — An* assoctation ule #5 an Te lowe dataset: : > "-€Cxarmple #5. Matct bastiet Arn Ae ob Linsupervised leaangs- = means cluster Yy. pcA 2. Hrevanchical clustering DrpSinguian value 3: KNNT ” de ompesitien peortgas ot thy ised lena. le compared te Auper e. Performs complex a Ceaanney2 = ! The outtpat maybe. less Qrruracy ®t chest have aoncapendiy cup Applications ' : eal wotore a. -tealth case 3: Biometite artendence be Pn | [3 fe elnforcement (eaaoy™ | day Reinforcement am tle can dtreay | determine ouput ‘ le an FReinpoteement leita te clot ude. label, data. Freinoreerent leaning #3 an rH i fateracts uatth ~the enbtwrment and act 6n tt * ¢nuftnoment Dies % vat kewar" y actions att “\y agent Terns sed in Relofyncernent tart Agnto:- An entity that can perceive and ack on enutvonment : —Enuitonment (9 i “The Place Ulhere aent 15 ded and actions are ee Place. | Key features a) Ketofprcerent pt Agent: 2g not frvstructed about envionment 8% — Bad 07 hfl and ~traf piece : | ; \ 2 ght 8 gM Bo pte CLEA CU aceocting ~to output « Ae. ey ee Clemens op ke - «Gg I polfey a fewara ae 3: Uatue-foretor U-— Mode! of; enutronment: A Igoxitners ie - g - Leaning 2 Deep g neural networ'c Aippticatfonss- l- fepbotica 9. Business 3: ‘Game plagfnc Main challengt> ob Mb ty, | poor Quatiey of data . Non- 7 epresenntotie neon) data 3 ouerpitting Ae Underpfting . 5° Coroplextt 6 Lack Ob Abilled veaounces 4 Aryelevant pentenes ! 8 Data biar | Algorithms in Au euised leayning ; ‘Decision —tvee algoutim : \ Dectsion tyee fs a Auperitised leaming technique used hov both al ‘ond: Reg reesion — techniques : ; vs fs pant preferred for astiiog al on problems: 3° A ig a -tree Atuctuned classifier iE Cintdins intemal nodes and leah nodes: 1° & 4nternal nodes , presents, peters off data Dyonches fees the: decision Tales and leah node represents vaitpat: Atuctune pa decision TEES decision node Tr ss | ee eel —) Nee eci4t0 (3) crm \ be 01 t > e =e ode) | Terminologies op decision tree ot a 9 Kioo Node ; Root mode: is from cOhere decision ~tree = Atants —> ft divide fnto two or more homogenous As t Leaf, node :- leah Nodes ane the fiat output nodes — . > “Wee cannot. be Aegreled after ett ing: Api ep ’g the panes o ducting decision oot node into Aub nodes: , Aub tee - py tree formed vbg aphitting ‘the “tree: oo - . panos Prantn Ps paneess *h versting the unwanted — branches prem “tree : Parent node: —Hhje ook nade ofy the akciston tree called ponent node child Node :~ ~The other node» eu woot necle called child ntdes | — ft Aire to Understand 12 : ve Useful po Aolving dectsfon problems Sess requirement of data at Disaauantages t ta fs Cormplex contains no: of ae & hove an ouerpftting fScue. ck = 3: For more labels -the Computation inoeases Pgthon implementation = : ' “Data _Ppwcessig : Fitted dataset +o alanithm 3 Pretetg teat Yesult, , 4. “Test | {jor ryesatt 5° UYfsualize the test Det vesult- CRY P; Sailcay | Spon - 24 [ VOM - Zoo00 Aes ™ Tpkntce rho e NeQy home (dealing ignite: teeeming is a machine 40 Dupewised learning ' Random ‘pa or vot ea con Of wea Pr bor costes % To eee Se 5 topes OM UE eratinbble leoneting ° 9, RE te Gotta oh corntih4 muiple | | classifiers “4p Avie comple? problem ) a, Random orest fs Qa classifier that \ Contains = NOH deefaton teen aNd takes caer 10 trnproue GEER of; datoset- 5 “We renter noroh trees Fig axceroet} and prevents ane cuerpitiing, 4 ny in {prot leads 4 Working of. Algorithm = Rata Taint \ ecision tree \ Test cet (Prediction shete one W00 assumptions Assumptions i \ + te Ahould Have Actwol Yaluer wn {ygatue oly dataset Ao we con predict accurate ryepnltS gather ON Quesset! result 2 Prediction» frorm each, tree oud fat!

You might also like