dhimmel

Verifying that +dhimmel is my Bitcoin username. You can send me #bitcoin here: https://onename.io/dhimmel

dhimmel

library(ROCR)
library(caTools)

VariableThresholdMetrics <- function(score, status) {
  #' Evaluate the performance of predictions for a binary outcome.
  #' 
  #' @param score A vector of predictions for which performance should be evalauted.
  #' @param status A vector of the actual outcome: \code{0} (for negatives) and \code{1} (for positives).
  #' @return A list.

dhimmel

dhimmel

dhimmel

dhimmel

# Compare four methods for computing the R-squared (R2, coefficient of determination)
# with wieghted observations for a linear regression model in R.
# This work was written by Daniel Himmelstein (@dhimmel) with guidance
# from Alex Pankov (@a-pankov). It is released as CC0 (public domain).

get_r2_cor <- function(y, y_pred, w) {
  # Calculate R2 using the correlation coefficient method
  xy = cbind(y, y_pred)
  return(boot::corr(d=xy, w=w) ^ 2)
}

dhimmel

Keybase proof

I hereby claim:

• I am dhimmel on github.
• I am dhimmel (https://keybase.io/dhimmel) on keybase.
• I have a public key whose fingerprint is 8CA9 B3A5 A4B4 D442 A285 867B 03A8 B7F8 47EE A45D

To claim this, I am signing this object:

dhimmel

Degree-Preserving Edge-Swap

Background

We designed a hetnet for drug repurposing that contains 50 thousand nodes (of 10 labels) and 3 million relationships (of 27 types). And we’ve chosen neo4j for handling network storage and interaction.

dhimmel

dhimmel

Drug repurposing by hetnet relationship prediction: a new hope

---

A long time ago in a galaxy far, far away…​.