Scribd
Upload
45 views

MLE Math

Uploaded by

Marufa Kamal Labonno
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
45 views

MLE Math

Uploaded by

Marufa Kamal Labonno
Copyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 16
Classifier based on Maximum Likelihood (ML) Event (MLE) and Maximum-A-Posterior Probability (MAP) 9: (Math 1.1) Animal scientists have found that the probability of finding small ears in Cats and Dogs are 0.8 and 0.1, respectively. Suppose An animal is observed with Large Ears. (i) What is the probability that the observed animal is a dog? (ii) To which class the test animal will be classified by the MLE classifier - CAT or DOG? Cats and Dogs = Suppose we have these conditional probability mass functions for cats and dogs = P(small ears | dog) = 0.1, P(large ears | dog) = 0.9 = P(small ears | cat) = 0.8, P(large ears | cat) = 0.2 = Observe an animal with large ears likelihood = P (feature | class) ... caited likelihood of that = Dog or a cat? particular feature in the given class = Makes sense to say dog because probability of observing large ears in a dog is much larger than probability of observing large ears in a cat ketinood) ® Prilarge ears | dog] = 0.9 > 0.2= Prilarge ears | cat] = 0.2 0 ===5 Classity as DOG" (Solved) = Gore idéa’ ‘Of MLE “classifier: Choose the event of largest probability, ie. maximum likelihood event (here, the events are "being Dog" and "being Cat" and the Maximum Likelihood event is "being Dog” we Ocul) 2 P (apo) POD) mapa ancien A Pe Poster lo: ’ - ir Sad o etal) = ae @ He) pz Peature Om Q: (Math 1.2) fish experts have found that the length of Salmon & Bass fishes follow Gaussian distribution (i.e. Normal distribution) with mean of $ and 10 inches, respectively and vprince of 1 and 4 inch’, respectively. A fish is observed with length of 7 inch. Explain how an (i) MLE and (ii) MAP(Bayes) classifiers would classify it - Salmon or Bass? (iii) What will be the classification decision by them if (a) Salmon & Bass are equally likely (b) Salmon is twice as likely as Bass? * Salmon’ length has distribution N(5,1) —zpianiurdasnaion) Marnie * Sea bass’s length has distribution N(10,4) rsvecives: = Respected fish expert says that Fee Probability = Recall if rv. is Mei it's density isgersiy p(l)=—— mo < resem the 'minus'in the = Thus class Sppattional defisitie: class conditional’ ia (07 Al saimoni=-—€ * 7 Hlbasd=— ia? zs Likelihood function = Thus class conditional densities are (us¥ (L107 All salmon)=—7—6 pl |ba: as)= ae ze fixed fixed = Fix length, let fish class vary. Then we get likelihood function|(it is not density and not probability mass function) ‘This is called lass conditional probability 7 esr deanity value: Ton? 2? if class= salmon pl |class)= 1 (10? fixed » 8 if class= bass (Ze, likelihood of finding the as fixed length of 'T'in the class ‘salmon’ or 'bass') * ML (maximum likelihood) Classifier We would like to choose salmon if Prilength=7 | salmor|> Prilength=7 | bass] However, since length is a continuous r.v., Prilength=7 | salmon = Prilength= 7 | bass|=0 Instead, we choose class which maximizes likelihood P| j=! -3F ol je “(-10F J | salmon) =——e ? I |bass = e 74 | RS I ML classifier: for an observed I: 2a basg< pll/salmon) ?p{I/bass) in words:itp(! | salmon) > p(l| bass), classify as salmon, else classify as bass > salmon (, (Math 13) Find the Deion Boundary betveen th Salon an Bas clases sed on the ent, when aw pro kmoiede aval Decision Boundary 0.4 0.3) 0.2 O14 1G: What i decision Doundary? The Decision Boundary betveen Stass fand clase i the ef of All Points where both the cases Ihave equal likelihood value (MLE classifier), or equal posterior probability value (Bayes classifier) or equal discrminant function alue Gor the Minimum error rate Tassifier) classify B ‘classify as sea\pass — salmon —— bass on Rayon chassihie®) GRE Be ianineh arate oe steel

You might also like