set.seed(1)
source('my.cn.inference.r')
sensit <- rep(NA,nrow(data))
guess.imp <- sensit
for(i in 1:nrow(data)) {
    cat('sample=',i,'\n')
    evi <- data[i,]
    evi[3] <- NA
    query['lung']='true'
    eps.max <- 1
    eps.min <- 0
    res <- NULL
    cat('eps=')
    while(eps.max > eps.min +0.05 || is.null(res)) {
        eps <- eps.max/2 + eps.min/2
        cat(eps,' ')
        nets <- my.econtamin(net,e=eps)
        res.t <- my.cn.inference(nets$net.1,nets$net.2,query,evi,verbose=FALSE,n.sample=20)
        if(res.t$max.p > 0.5 && res.t$min.p < 0.5) eps.max <- eps
        else {
            eps.min <- eps
            res <- res.t
        }
    }
    cat('\n')
    sensit[i] <- eps

    if(res$min.p > 0.5) guess.imp[i] <- 'true'
    else guess.imp[i] <- 'false'
}
## check result is ok
sum(guess==guess.imp)
## sensitivity of correctly classified
mean(sensit[data[,'lung']==guess])
## sensitivity of incorrectly classified
mean(sensit[data[,'lung']!=guess])
