winvector / zmpdswr Goto Github PK

On page 36 of the book, code refers to "ORIGINRANDGROUP". Dataset is "ORIGRANDGROUP"

Nice to meet you. Who are you

Hi, sorry for bothering you. I have a small question regarding to the function for calculating the squared distance between two vectors on Chapter 8.

In the book the function is :

sqr_edist <- function(x, y) {
sum((x-y)^2)
}

but don't you need to square root it after the calculation. so it'll be like :

sqr_edist <- function(x, y) {
sqrt( sum( (x-y)^2 ) )
}

Thanks for helping out.

Hi, I'm working through the book and I have an issue with the following example: 00070_example_5.14_of_section_5.2.5.R

When running outer(1:n,1:n,function(a,b) d[a,'cluster']) I get the following error Error in dim(robj) <- c(dX, dY) : dims [product 529] do not match the length of object [1].

I'm not using the original dataset but my data has the same format.

Re-writing the code this way fixes the issue

n <- dim(d)[[1]]
pairs <- merge(1:n,1:n,all=TRUE)
names(pairs) <- c("a", "b")
pairs$ca <- d[pairs$a,'cluster'][[1]]
pairs$cb <- d[pairs$b,'cluster'][[1]]
pairs$dist <- sqrt((d[pairs$a,'x']-d[pairs$b,'x'])^2 + (d[pairs$a,'y']-d[pairs$b,'y'])^2)[[1]]

dcast(pairs,ca~cb,value.var='dist',mean)

Here,in chapter 6,there is a function I think something wrong.
while "pPosWna <- (naTab/sum(naTab))[pos] "can't get the stat that we want,
We really want the result is positive proportions,but the results are the negative ones.
have a try ,hope your answers,thanks.

mkPredC <- function(outCol,varCol,appCol) { # Note: 1
pPos <- sum(outCol==pos)/length(outCol) # Note: 2
naTab <- table(as.factor(outCol[is.na(varCol)]))
pPosWna <- (naTab/sum(naTab))[pos] # Note: 3
vTab <- table(as.factor(outCol),varCol)
pPosWv <- (vTab[pos,]+1.0e-3*pPos)/(colSums(vTab)+1.0e-3) # Note: 4
pred <- pPosWv[appCol] # Note: 5
pred[is.na(appCol)] <- pPosWna # Note: 6
pred[is.na(pred)] <- pPos # Note: 7
pred # Note: 8
}

Hi! Working my way through the book (love it so far!) and I had a question about how the F1 score is calculated in the accuracyMeasures function (

In Listing 9.1, it says that f1 is the product of precision and recall, which is how it's implemented in the code. Looking elsewhere, I see F1 defined as the harmonic mean or weighted average of precision and recall:
https://en.wikipedia.org/wiki/F1_score
http://scikit-learn.org/stable/modules/generated/sklearn.metrics.f1_score.html

I didn't find anything in the errata, so I wanted to make sure that this was deliberate. Are there different definitions for the F1 score, or should I use a weighted average for F1 score in my code?