即使学过机器学习的人,对机器学习中的MLE(极大似然估计)、MAP(最大后验估计)以及贝叶斯估计(Bayesian)仍有可能一知半解。对于一个基础模型,通常都可以从这三个角度去建模,比如对于逻辑回归(logistics Regression)来说: MLE: logistics Regression MAP: Regularized Logistics Regression Bayesian: Bayesian Logistic Regression# ----------------------------- multiple multi-variate probabli: # acts=4 tunes=5 weather=7 how many??: # prob entries= 4x5x7=140 # params= (of classes)2x139(likeli 1 class)+ (2-1)prior=279 features: acts , tunes ,weahter #params= classesx Πi(values featuresj)-1
params to estimate likelihood: 2x(4-1)+2X(5-1)+2X(7-1) #classnumber#actvals tunes weather benefit of naive bayes: very fast learning and classifying.
|