Download Machine learning notes and more Study notes Software Development in PDF only on Docsity!
Date: / | \ Machine learning C Pave J N T. T Cmo n) Uniupevsed mL. | clouthe chon Cluska 19g a pam Recomndaly: (ogirh'c ° kemeang pees | ( KNN ° DGScan Conkerd Memory ea, OT < hverorcht j 9 me ‘cron eb SS Una bye Vier ews RE based bare) NGe Diowcabenell Rediveh as GBM j T 1 SVM PA tone [A Svd Hut PAIN. Exod ing =i calboost, adabeos | z upemied ML —> wort stot witty Reqraiton —y lincoy # f Lincor Regraion | ‘ Mut hale ps ly nomial grasion incon Regression Regression (MLR) il Iineor Regrusi'on (64R) ¢= whe for wall au 2p ete only One fot 1 then if al packeqe ZLPA RES Mecca. Raster of linter Regression Cquahion of chargh line Yemute And value of on and ¢ el Ps Co, 4) —> xv, Ms ala Paar) Nome. Mn ; i n i : : | = f a 3 4 OS Me slope = Vay = 6-4 Mo—X, QineQ mn. 2 i ef eo ® ilo me pecans with every chiang 10 Value of Beinn eh eli'}fece lay laterceph ae Py (a2) —» (yy, 2) Aes rnc C geo(ay te Se gar Ce=26 Wien erce ee at esnboyar tine eq” iy acy pete tcloke X ord + =r 7 m for_gieen valuc of X We Can ind tosreayd0nd jhe Value of a whak if we have dio or proce then wo indeped cat Variable q =| then th ir called muti pl « linecy Legrewsi'an PES, eae, 7 aren At Fs wil, nea bh £04 7 plod ual TELL ERA (3p) bit 7 al ex shh » Now find he loce:funchon fous :[lo-8 ecru) EUR) Af ae at C2b—20)* a > 4Adtqtlster+36 - AO cle 4 es a re teoson qo toke squers: Jothol pos ikve and negohve._] ‘ Veluc_ moy nek cancel each othy de vw lon Gone —~ High Accuracy — high [ou value ——~ low: Accuracy [We can feoprewe the model by Jom pphmizehion technique | called QA * gred éenl ae wher Pepecit the prote Herokvaly Baki Wwe gel bal poromile Cm, ¢) for | | Hi: an will give Paint ioue tunehion. eel a || 1 ‘als bal minimum ie grodive} devecol” arenes eum | yuck thal [nner funch'en 9 The Patch | = PA hich anweld C cor predict / a. ; ar ee C baeeinst fa ace (3 {tho will Step down onkil War 6 te } . Bah i" j “global Minimum by eave q L. x ] Seay Sa / ey Je : | = ie poremel ¢ ‘i Rondom Globa) oe! a jnth'a) Mini mum value Cmx(a) ; ps! a In mo chine learning an [rte [re \ LD wenia.r, bap ole ac. Lert L Dc 1S Bed) Bhan slope > nite es ye | Og Picccent ee ce - learning rate mb mognetud of chonge thot you a wart in porom cer during tprohioa 4 + parked dlerevokve of Jou funchionw:r- + m is denvohve of fous functhion Wr: | (Cues teh e) mn s aa, | now tt aes SST a Serres oy . Kn bso a ll ae Da hos: funthon) . 2 ope, 7) ee Xe . Beare } eek —— re PL ay Bee tS Cy, — , NI 9 ae | Page — Umpuons o BOOMER §1c1100 } ‘a4 five main assumphion (p Blinc or — brhorsh pnput_and oul put. Jo. teulkicolincarity i == -" onal iby af Dori diva } = ashtes by . 's to cotrelahon of €rroy = Bereihotl di... | cna relakon . Pemenbk tad Eee Feadure ane dorget (output) fee lalea poukve OL negakve Gocclenon lincav Rerw Noon (aear(x)_ lonplreoabl, +y multiple dae publ con ta _tael ie Netels to check thy - ne plot (Leoture Aiclogd) feoturc Ne tog cb). : sphinn ok. m Mutbicolhincort ly .: ik reanlotvere Fecture should hb, independent or should nok urrclohon amang By nit lied whol the pablur° iat Linco Regression Model for 30 we dea om good . eS nee as alii i lon to check miibtcalliaepertky oe | =~ | 1) | vit (Voniance InAloh'on Facto) Sic | il at ic around | then feoturt dont hove the iusue cond i} a} v1 ¢ or more than thot Hue that | perbicul ar feqturt hove pout ‘coll incorhy fas ue and need fp remove it a} | anuther method te h find out corrcohon behaecn al) th: : fe otura C Heolmap ) fgsutm ph'on 3l- Normality af Rend ual il oe Hhok fondo dnchee Clam teldsgatelict nell] nlof or goph lawl 2S salvar lived! Domo I | dithibuhan— means Mmarsimum ervor theutd Ground mtan =0 | how to chee ye: kernel deasihy ediqrabon( kele ) or Q-Q plot ce ma | =€ ras \ § ca | Pe ind v6 : | [ Me cs oo 8 / \ cM ean = a a é 3 | kde Q- @ plok ) | (Therohcal Quanh'k ) 41 Assumphion Al: Homascedastivity * wi —__——_ Some ScaHered Lipee ef u Jpreod of _testdual shoud he equelly | or voitor! ly Jealterecs if A feral Wee Calle of eee chuly whics ty nob desirable . q | Date: / / | te | an Non - linear Equation = Poly norniaf Reg re ssio, ws tnow 1 Qouckon of llireantp = mx+C | and Pawahion of, Stoople (in eo'e Regretston | “y : fo + Aix land fics mul ee ftacar Lquakion ee fo + Ai % 1 + Bo % + fig %q te « WE Bp%n | I t ai Have Lr A clelitee kit Paralun rel evan Nala is ae ia wwhak if data 1: ‘Do Lane ? i iy | | | | | oe e ee such (cenan'o we exckack i | ee to. polyncons ol Ee rene Dees . i ee o} tonal Wome len ayn * e ; s Pri fro Ctsiing shaqe eee Z| eo Sle s } | fay we “ool ty Abele: poly noms al of d
kt} one Cont Hn ir telekon bchween Vy candl coe lfu' en) iy JK [ine or me 43 2 ab 2Hlyrese. ae fi, 146 be ey = mex Sra nt 4 tyeest (29) ‘@ Anal a oe # = + b = -o-segs#, a Ts lincas tquabon ae J | Sawld| plug value. , ia : als math. lis pak iy ae AIX - 0-83. ll [x | tot ES for wu y Nack 4 mt ec 4i(H)-0-83 = 3-94 3-8 | y+ jt NCES a ee ppeani(g-o-e + {604 | 16. Sur : ele api-OLs (y,%) : depeder| Yoanabl, n_— Indepederl No abies Jempor} stahmade|- api aa $m #teoding dolo rom ¢sV odt = pa: neoee Gao ain. csv 1} tt dehining the Varrobles os Ab Le] dole O 2 Pe Ae doliclt) (ee) H_adding th wo pS Sc add constant (eo) # Performming reqvaasy fal Nn peel OND Cree OAS (yo bE it price} Jummory Be ee pant (_ruull- Juma th) ea eee a 1 4, i ——_——— oe te Date: / / Page Ls ee wee) ——_Y a aot <~— Nowe odd perally 4 x as befovy sO uonnrenalye redutrac th ich wo) add he) this term Ay ty id fox Hh fini term Bea Fel Corer _{ lies huynce spotorncdey | : Reiduaf —* Pm > slope. | Now we will coleutats [ocr for both |itne for * =) | with eqe © i mmee [inc 1b posing (2-3- o-q-les)> + FESS both Roane: ($-4- a-F-1-QE 4+ (0-4) . perfecHy Hoo! 's why Lt dem i De werk = (OP fine & 0-4) > \) 2-03 JL ot! (r-n~ = Bist j Lee we ge thing s1'g.0i hicon} ie ee Was lout for the nud Itne As our mod! tan see thy change ik will releck and is ro del althoug ete dive ha d O(Curciu DK a. fyacs roe ie atgns rcumbl y reduce ane b las pocreasce “ My tare Called Le Ante | if we ep mote than one tonut ots penal hy - aaa od Dy ( mittm, 2) a An A (misimetme) Ver Wet Are eloing iquar (*) at} fe yim we Callid ae L192 Nom oy {ala regul avyach'gn “ | Lasso Rearession et Rea daal ena, J I dats it pins ie lp to reduce ouerfitting- In Qidae qa we have veen— 2 Ae my fee Net ae = Wea yp) fea Ne 2 Ss eee rtt (eto ee a” ew) A) tre frdaer ir eth —~ Losin te ful andthe r wanzhon af Rvelae nN en ew) SNe z xa oR ee ea eS +iw.ll In ridge regression for ny voluco} A thee vere a] Deu, 1om« Gale. hoy dee ek of in pup feature buf in lave 2 yuu conkreusly intra valve afl Aras testa'n 7 _paio| Gps will xen lun For som o} Cocttvenesd which ar al Lenpatlon| yo hea we unknowing doing . Aclechon amd it 3 odvenl oq Tee lao Whee you aie ere high oltmenitcm al cla Land dur feahune ore nol al we shoud pref | Lang aur Ridge ! Page } > ( Date: // é i € Z 4 : Fe. g Vo A Variance ~< | a fig a. i 24 cui = g ; 8 Teer. lies — i “il $=. —¢— 4 oder filly Ove hy Baca) eeViamancee HH val Vom'one rg(8 | Seedelitihg aig a sb: | es Te! aed | | ! | undesti Hing | | | EE Aah High bios (os [as C { | = a i Vortance Tradeott: Lil ty important to Bada dead pred( chon e1rore¢ Lins aad Variance) when tbc comes to accuraty |_Lin cone pk PA gen'fher J gllethhe i frodioft between model's» ability do minimize ie a Vorionce which j, rebbinest +p I heal _srluhon -fr Sdeckng c eral of tequl oxzekin | Conlon ts _ pre Pts voderstonding a} these error: woutd help | al ave id the ovclti ting and undepitting d e doleue ukile choining ol grithon. Jd yn ML sere on el Sioa Qeing hugh i ia ive large err ta bras'ning ou laut Cy A pi whey ae, We ads a retorm mend ral Thal i fting Lovie OA oder pHing. hop nen the hynatheais ic is Lionple ax _linem y T rg h Li aul ae 4 data pire inltaoe Wie tad Lpread >| ous sala af Colle d Wirenivaron lle Sneed ne mode} Lath high Uam'on ce Kas ‘oO neh tompl te ht Mangala ‘to ae act able fo fe ar the tel f | pe dl ieee al ry Li ae hrgh erry totes alo a {4 Hibs } y } ap qo ak D f dale. oul Us ‘& hing Th. beosinta hate 1p) Curve ord Lula h order hy pith exis heel apt luhton atthe evar Hee eels tha ae ! 
