Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Statistical Decision Making Supervised Learning: Using a training set to design classifier – Using a separate test set for accuracy Unsupervised Learning: clustering Parametric decision making: probability density function is known for each class, not the parameters (mean, variance) – must be estimated. Pattern Recognition1
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recogntion2
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion3
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion4
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion5
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion6
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Bayesian decision making refers to choosing the most likely class, given the value of the feature(s) P(x/C) is the conditional probability of obtaining feature x given that the sample is from class C P(C/x) = P(C) P(x/C) P(x) Example: What is the probability that a person has a cold (C) given that he or she has a fever (f) P(C) =0.01, P(f)=0.02, P(f/C)=0.04 P(C/f) = P(C) P(f/C) = (0.01)(0.4) = 0.2 P(f) 0.02 Pattern Recgntion7
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion8
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion9
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion10
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion11
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion12
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion13
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion14
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recogntion15
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Likelihood Ratio between class C i and C i R = P(C i /x) = P(C i ) P(x/C i ) P(C j /x) P(C j ) P(x/C j ) Likelihood Ratio between class A and B R = P(A /x) = P(A) P(x/A) P(B/x) P(B) P(x/B) If R>1 – select class A If R<1 – select class B Pattern Recgntion16
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Example: Detecting the HIV virus using the ELISA test H – patient has HIV virus H’ – patient does not have HIV virus Pos – patient tests positive Neg – patient tests negative Let: P(H)=0.15 P(H’)=0.85 P(Pos/H) = 0.95 and P(Pos/H’)=0.02 Pattern Recgntion17
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Bayes’ Theorem P(H/Pos) = P(H) P(Pos/H), P(H) P(Pos/H)+P(H’) P(Pos/H’) = (0.15)(0.95) = 0,893 (0.15)(0.95) + (0.85)(0.02) P(H/Pos)>0.5 Likelihood Ratio R = P(H) P(Pos/H) = (0.15)(0.95) = P(H’) P(Pos/H’) (0.85)(0.02) R>1 Pattern Recgntion18
Technological Educational Institute Of Crete Department Of Applied Informatics and Multimedia Intelligent Systems Laboratory Pattern Recgntion19