Frank Blaauw frbl

## gist:cae79f708cae75a20614
cat index.fdb_latexmk | grep .sty | awk -F' ' '{print $1}' | awk -F'"' '{print $2}' | xargs -J % cp % ~/test

## keybase.md

      
              1 file
            
          
              0 forks
            
          
              0 comments
            
          
              0 stars
            
          
                frbl
                / keybase.md
            
            
              Created
              November 16, 2016 22:13
            
              
                keybase.md
              
          
    Keybase proof

I hereby claim:

I am frbl on github.
I am frbl (https://keybase.io/frbl) on keybase.
I have a public key whose fingerprint is 8756 D347 8013 8489 9210  FDD0 1BC7 5A3E 743F 90E1

To claim this, I am signing this object:

  
## sample_runif.R
data <- c(1,2,1,3,2,1,2,3)
intervals <- c(0.1, 0.3, 0.6,1)
runif(data, intervals[data], intervals[data+1])

# Test if the output is correct:
check <- all(unlist(lapply(1:length(data), function(i) {res[i] >= intervals[data[i]] & res[i] <= intervals[data[i] + 1]})))
expect_true(check)

## notebook.sh
#!/usr/bin/env sh
lib_dir=~/.R/lib/notebooks
mkdir -p $lib_dir
`sleep 3; open http://127.0.0.1:8888` &
docker run --rm -it \
  -p 127.0.0.1:8888:8888 \
  -v `pwd`:/home/jovyan/work \
  -v $lib_dir:/usr/local/spark-2.1.0-bin-hadoop2.7/R/lib \
  frbl/docker-notebook \
  start-notebook.sh \

## data_export.html
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8">
  <title></title>
  <script src="https://code.jquery.com/jquery-3.2.1.min.js" integrity="sha256-hwg4gsxgFZhOsEEamdOYGBf13FyQuiTwlAQgxVSNgt4=" crossorigin="anonymous"></script>
</head>
<body>
  <ul id='data'>
  </ul>

## bounded_logistic_regression.R
#!/usr/bin/env Rscript

## Constrained logistic regression
## In this gist we create a regression for which the predicted probabilities are contstrained. That is, they can not be
## less than a minimum of delta, or a maxiumum of 1 - delta.

## Based on:
## - https://stats.idre.ucla.edu/r/dae/logit-regression/
## - https://github.com/achambaz/tsml.cara.rct/blob/2b2aa282d4a11c601b37cacb368b67d03f6e8fc9/R/misc.R#L440
## - https://github.com/achambaz/tsml.cara.rct/blob/fd426c2a6f91b692d379b4765d9900151d09daa6/R/targetGstar.R#L1

## glm_logistic_treatment.R
Osample_p_full <- read.csv('http://frbl.eu/files/gist_ac602828fedba735b4d1e4cb16472cbe.csv')
formula <- Delta ~ W + A
glm_pred <- glm(formula = formula(formula), data = Osample_p_full, family= binomial())
print(glm_pred)
# Call:  glm(formula = formula(formula), family = binomial(), data = Osample_p_full)
#
# Coefficients:
# (Intercept)            A            W
#      18.378      -19.241        0.379

## difference_between_rbind_rbindlist.R
library(microbenchmark)
library(data.table)

rbindlist_time <- function() {
  data <- data.table(a=c(1),b=c(2))
  for (item in 1:10000) {
    data <- rbindlist(data, data.table(a=c(1),b=c(2)))
  }
}

## xgboost_failing_update.R
set.seed(12345)
library('data.table')
library('xgboost')

## Create simulation function
give_me_data <- function(nobs, delta = 0.05) {
  X_mat <- data.table(A = rnorm(nobs, 0, 1), B = rnorm(nobs, 0, 1))
  probs <- pmax(as.numeric(X_mat$A > 0) - 2*delta, 0) + delta
  X_mat <- as.matrix(X_mat)
  Y_vals <- rbinom(nobs, 1, probs)

## eduroam
[connection]
id=eduroam
uuid=<THIS_CAN_BE_AUTOGENERATED>
type=wifi
permissions=user:<PC_USERNAME>:;

[wifi]
mac-address=<MACADDRESS_OF_YOUR_WLAN_CARD>
mac-address-blacklist=
mode=infrastructure
	data <- c(1,2,1,3,2,1,2,3)
	intervals <- c(0.1, 0.3, 0.6,1)
	runif(data, intervals[data], intervals[data+1])

	# Test if the output is correct:
	check <- all(unlist(lapply(1:length(data), function(i) {res[i] >= intervals[data[i]] & res[i] <= intervals[data[i] + 1]})))
	expect_true(check)
	#!/usr/bin/env sh
	lib_dir=~/.R/lib/notebooks
	mkdir -p $lib_dir
	`sleep 3; open http://127.0.0.1:8888` &
	docker run --rm -it \
	-p 127.0.0.1:8888:8888 \
	-v `pwd`:/home/jovyan/work \
	-v $lib_dir:/usr/local/spark-2.1.0-bin-hadoop2.7/R/lib \
	frbl/docker-notebook \
	start-notebook.sh \
	<!DOCTYPE html>
	<html lang="en">
	<head>
	<meta charset="UTF-8">
	<title></title>
	<script src="https://code.jquery.com/jquery-3.2.1.min.js" integrity="sha256-hwg4gsxgFZhOsEEamdOYGBf13FyQuiTwlAQgxVSNgt4=" crossorigin="anonymous"></script>
	</head>
	<body>
	<ul id='data'>
	</ul>
	#!/usr/bin/env Rscript

	## Constrained logistic regression
	## In this gist we create a regression for which the predicted probabilities are contstrained. That is, they can not be
	## less than a minimum of delta, or a maxiumum of 1 - delta.

	## Based on:
	## - https://stats.idre.ucla.edu/r/dae/logit-regression/
	## - https://github.com/achambaz/tsml.cara.rct/blob/2b2aa282d4a11c601b37cacb368b67d03f6e8fc9/R/misc.R#L440
	## - https://github.com/achambaz/tsml.cara.rct/blob/fd426c2a6f91b692d379b4765d9900151d09daa6/R/targetGstar.R#L1
	Osample_p_full <- read.csv('http://frbl.eu/files/gist_ac602828fedba735b4d1e4cb16472cbe.csv')
	formula <- Delta ~ W + A
	glm_pred <- glm(formula = formula(formula), data = Osample_p_full, family= binomial())
	print(glm_pred)
	# Call: glm(formula = formula(formula), family = binomial(), data = Osample_p_full)
	#
	# Coefficients:
	# (Intercept) A W
	# 18.378 -19.241 0.379
	library(microbenchmark)
	library(data.table)

	rbindlist_time <- function() {
	data <- data.table(a=c(1),b=c(2))
	for (item in 1:10000) {
	data <- rbindlist(data, data.table(a=c(1),b=c(2)))
	}
	}
	set.seed(12345)
	library('data.table')
	library('xgboost')

	## Create simulation function
	give_me_data <- function(nobs, delta = 0.05) {
	X_mat <- data.table(A = rnorm(nobs, 0, 1), B = rnorm(nobs, 0, 1))
	probs <- pmax(as.numeric(X_mat$A > 0) - 2*delta, 0) + delta
	X_mat <- as.matrix(X_mat)
	Y_vals <- rbinom(nobs, 1, probs)
	[connection]
	id=eduroam
	uuid=<THIS_CAN_BE_AUTOGENERATED>
	type=wifi
	permissions=user:<PC_USERNAME>:;

	[wifi]
	mac-address=<MACADDRESS_OF_YOUR_WLAN_CARD>
	mac-address-blacklist=
	mode=infrastructure