Download Machine Learning Notes and more Study notes Engineering in PDF only on Docsity!
Date: / / a _Mochine learning C Pane / E I —— = Gece { q Me f Unsupervtied mL. : cleaiheshion chalitiag Oman — Recommdale. =A (ogi th’ > kemcanp | KNN ° DGScan (onkend Memory OT « hverorcht'c Le ea Nav bye Gc (lean RE bord bane NGO Nincabtonel Redu ae GBM j T l SVM PA tse [CA SD tute ANN Enel ing ahi calboost, Tee Supe ‘se ML —> Je woit| stot with Regrairon —y [incor rt ! Linsar Rearaion i Eel Muth'p le pe ly nemtal glesteo linc or Rearest'm Regrasion (mL R) Ca Reqrasi'on (tbe) = when for ule we hou. only favawa fcotue theo yl! caee | pockeqe 2LPA 3 LpA GLA, Raster of inter Regression : | Equa hon of clratgh ling Yomute find valuc of om and ¢ Po (2, 4) 2 2h VP Ty) eer eto £ Ns slope = Way = 6-4 Ur —X, Cee NE By Se A = 9 5 7 ! slope pecans wilh every change in Value of % how couch edt }}ece away) Jatecceph 26 Py (24 ¥2) —> (4, 9) A) Tasee Gale Qe A) te G Bo pr C= 6 Lnfecoce Mnclilveaabous line eq” ly funchon thot Kelolk: ¥% ord ar 7 = for_gisen value o| x we can find Cosrexd0nd jhe Value of- y what if tue aie two Or More thon two indeped cat variabl« y a at: dhea th ir calles multiple linea fegrey ian pa ied 7 GAR GENRES pre _ i Et PNG ala PaO eT ED rie 7 [ (3) J Now find the lore: funchon fous [0-8 eye) (eed Coens] a Mpdiiter hagas QO ele fe a (ein ole quer. porthot pos ikve and negokve, } S veluc moy nel Cancel each other ww low volte ——» High Accuracy a high [oi value ——» low Accuracy v We can non rave dhe mod! by Jom pphmizehion technique Called ca grecl cent hae where Jan ae : I thecokvely Rak Wwe Yall bul poromche cm, ¢) for Lwhith a will give minimum ow 4tunchon. MWlette doy “global minimum 16 gredive} descent” we Con use grocers Meee far linear J for oph me Soko Pct ty determining beet porometes ogaielc Ate i Ae The erclid ee set ored. ik Bes en Cy \ and eared, porame le ref 3 trode! Qty phimumM Op himr2 < q / \Date: 7 [| { iN {thir will chep down unk rebel?bwn fo Y a : pa é global minimum by ikeohve mY | (re cers . [ ' t i 7 hh SS : a __poromel ¢, Rordom — Gleba) \ a 4 ath‘a) Myay mum _Nalue Cm3(s) la Fe inn 1 ake pital Ing ‘ Hae. < [mm feyern) |= LOm We Us idvw Bie \c/-\e, + L De b= b- Ldb im- slope ret Ini Mog U) = uecakk ae ccent ote. bita4,. IL = learning oem | i mognrtud of chenge thof- you . . t wort in porom tte during iteroh’on ‘La Pi perkel derevokive of lou function w-r-+ m > pork cre Vob've of let Ain chiA. (a cx} ¢ A Sol.) i, x ae rN Meewiemanches) 2 / 1 = (y; vf )* =i Pe. am om Aa 1 Ss === gee? fle Oat WB ar urs 2G (Yin yi) | q re st Clos funthon). 9 / | <= Gant )") Ls C. a prc, Rin ites Sie AN, “ es ? ; ial ce , od: | oy Page ONS oO Fineaime aye.s/0n d are five Main astumphion Mbilincor re on between ‘input _and Ou put. 1 No roulkicolinearity i! y mali aft Renidual ! ee ohicr by. Giess ton A a BA cs Me dell (eawarl relahan Gelween ‘individual Feoduve cand dorget (output) could he poukye pr acgohve corel ob'on ey No \ -/s I ee ath 9 : re \ ay a Son ae lincay VU Nine ow Y Non-lineav(x J plicobl, +p multiple eae pale gen aa tel Vo | m0) this * Scaler plot \ (feoture ANctorg ct) | feoture Ne tog cb) : Ag Mutticollineant hy .: ik mean there ure jhowld hb, indenendent ox should nok ove any torrelohon among themrdues the rable © mulki linear Regression Model for 80 we deo va Ahin mo Or im geod] - (& Saab gs sueohet ) ges NI q ) bets dy check multtcal linecorly 1 h— i vif (Voniance InAlotion factor) eae ilat ie ground | thee feoturt donb hove th. on | i and i} ib ve ¢ or more than thot Hue that | perbcular feghure hove roth ‘coll incorhy fasue and need do remove it Gnuther method te bo find out correlohon Peete a al{ Ay 9 kde Q-@ plot AD} dh. 2 fe otures ( H
a Date: / / = Page a i [indo Set Alien Sea. dl df Po Nee ZL Meare | fig & Re mul tas va we oe eee es Tipe We 4 r= a I —_—_ + | uoder file oveh Huy | SE) buVannnce Hah Vona‘a re Z (| siodeih thing | t | low) Way We, y VES : rr ge ene. | i | se E Ce : Ree i | undeeti tring — Paes 
