Machine Learning handwritten notes, Study notes of Machine Learning

Download Machine Learning handwritten notes and more Study notes Machine Learning in PDF only on Docsity!

Sp thor A Det seas Bn pct te : rdividua@k best : “PROpEEe. Fach tah Dest PISS hie cept, LOC hypothe sis Ch). YTS —catled as ome ANatl hypothe Sigs ee oe Tes an meeatk Starter, aig om é (6 sno -Tetettionshitp hetoeean bao — einige : events. @& DAeenative. hngpathesis. ae aba toes 9 Tt6 a divect comtiadiction ace ae sbiseotesie, which zmeans +f One of -the -o0_ hypotheses is tne. then the athee ene -must be -fatse. . 2 SS oe 3 CoSt Auction or fo s5 function => is | ____* Fi cost function is a pazacmetes that deteermines had : eo | eee welt & MIL «model _pesforams For given_dotases. é [eee ee (Cost-fenction_catcutokes datrezence hetieesa tne — ma. + expected value cond puedicted ac a —_.. : ca ob aS ou Shale eo amine eee ee ee Thee ann aspen eterna aeegesSaeE ha — aren ite. Poreear ecco | a lt beeen xX Cinpute) cand Co! fee ie *"The. satain genta eagle SMeaegesee a+ eninge ea ee eo : a Bis find the aptimat solution coc PE the value of cast -Fircon.—— @) Local. (nox emdany 2S Ve raidie OF Fur erg the ChAL OVA c se | * Main oblectue Gf usting gradient de 16 COMI Ze the. cost poime SCE altgorith ry ful asig iLes ; éaiOmn that’s wohy_ tt 16 also Crilled aS thesative : ~oplimizerrion 1 atgorithamn. ' * Tapes of gradient: descent algorith-n ee 1) Batch gvadierrt descent 2) Mini batch gradient. descent 5) Stochostic gradicmt descent 1 1 Gradient descent for Lineas segression == Dyn = prrtiat derivative Of cost FP W.-M q De = pasttiat decivative of cost fat LAD or hh Neem Mos, tohexe, X= independent vaziable me + t ona) | Y= dependent Wasiable. ) “5 : 1 4 [pki BCcost fu?) a) es ie GYi= = Yipred O70 Damn i r e i eee Dyn = 4 ~2_h Cy; bee Corn st rete) q r *. Hm Po= ost Am) = 4 ia De = tag ee Nena, : , c= c=lDe peuc cost ie is 2 We. exit repeats this. ipoeek sand very srmate Cideatly O- a iP e rire . ee Si ‘ex my an anne a Divide. all the va pucs py the Caluxzmmn by eh iL « aM a ca Vale ae 2 Pin — Tax ormatization = Sie Rey Sole Sire 2 ; if technique ve~Scates a feature. or observation — Sea VALU! with distaibution value betoeen O and 4... $$ __X veo = 2G — win) = EO 3) Stan dasdization => way - Z 2 oe hae © Et vescales a feature vatue Sa -thatkithas distribution with O mecen value and vasiance See ERG o's 100 85 ios eb es if x = Cj — Wmean Sahota 2 Aeve LIS 28 8 ET Baar Stomdetd Deviation CSD) —— * Bolynomi al regression ee © Tbe models the velaonship betaeen a dependent aay - Sa aee | CY)_ cand tdependent vatiable CO) as mth degree at ___palgnomiat.__—_— $s 2 OER ~- : ve Tenens wnadet ith some woadificatiora ee i ee He Sane ake some tendae dataset —— fee © srt a ee are. comuetted to Pierce Pie sn Sela, oe a SS 3 Ss ee ae ae |