Packages

This talk will present packages to:

• Scale computing: What to do when code is slow?
• Perform non-standard visualizations, as in, beyond ggplot2.
• Train deep learning models, to classify images and beyond.
• Interoperate with other languages, like Python.

But first, lets talk about the lifecycle of a package…

Lifecycle

![](https://img.shields.io/badge/lifecycle-maturing-orange.svg)

SpaceX

A private aerospace manufacturer and space transportation services company.

Boston Robotics

An engineering and robotics design company founded in 1992 as a spin-off from the Massachusetts Institute of Technology.

Polar Bear

A polar bear exploring technological advancements applicable to maritime transportation.

What to do when code is slow?

hpiR::seattle_sales %>% lm(sale_price ~ area + beds + age, .)

# Sample
hpiR::seattle_sales %>% dplyr::sample_n(1000) %>% lm(sale_price ~ area + beds + age, .)

# Profile
profvis::profvis(
  hpiR::seattle_sales %>% lm(sale_price ~ area + beds + age, .)
)

# Scale Up
cloudml::cloudml_train("train.R")

# Scale Out
copy_to(sc, hpiR::seattle_sales) %>%
  sparklyr::ml_linear_regression(mpg ~ wt + cyl)

profvis

Visualize R profiling data.

install.packages("profvis")

library(ggplot2)

profvis::profvis({
  g <- ggplot(diamonds, aes(carat, price)) + geom_point(size = 1, alpha = 0.2)
  print(g)
})

cloudml

R interface to Google Cloud Machine Learning Engine.

install.packages("cloudml")
library(cloudml)
gcloud_install()
gcloud_init()

# train.R
saveRDS(lm(sale_price ~ area + beds + age, hpiR::seattle_sales), "model.rds")

cloudml_train("train.R")
job_collect("cloudml_2019_04_18_003956121")
readRDS("runs/cloudml_2019_04_18_003956121/model.rds")

Call:
lm(formula = sale_price ~ area + beds + age, data = hpiR::seattle_sales)

Coefficients:
(Intercept)         area         beds          age  
   166678.1      -1579.2     146588.3        216.7  

sparklyr

R Interface to Apache Spark.

spark_install()                                      # Install Apache Spark
sc <- spark_connect(master = "local")                # Connect to Spark cluster

houses <- copy_to(sc, hpiR::seattle_sales)           # Copy data to Spark
dplyr::summarize(houses, n = n())                    # Count records with dplyr

ml_linear_regression(houses, sale_price ~ area + beds + age)

Formula: sale_price ~ area + beds + age

Coefficients:
(Intercept)        area        beds         age 
166678.1295  -1579.1792 146588.2559    216.7374 

r2d3

R Interface to D3 Visualizations.

r2d3::r2d3(c(0.3, 0.6, 0.8, 0.95, 0.40, 0.20), "
var barHeight = Math.ceil(height / data.length);
svg.selectAll('rect').data(data).enter().append('rect')
    .attr('width', function(d) { return d * width; })
    .attr('height', barHeight)
    .attr('y', function(d, i) { return i * barHeight; })
    .attr('fill', 'steelblue');
")

nomnoml

The sassy diagram renderer.

nomnoml

The sassy diagram renderer.

devtools::install_github("javierluraschi/nomnoml")

nomnoml::nomnoml("[Import] -> [Understand]
                  [Understand | [Wrangle] -> [Visualize]
                                [Visualize] -> [Model]
                                [Model] -> [Wrangle]]
                  [Understand] -> [Communicate]")

voxels

devtools::install_github("javierluraschi/voxels")

voxels::voxels(matrix(c(
    1, 0, 1, 0, 1, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1,
    1, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1,
    1, 1, 1, 0, 1, 1, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1,
    1, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1,
    1, 0, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1
), nrow = 5, byrow = T))

How does one classifies images?

Many more applications: Speech recognition, natural language processing, self-driving cars, etc. See arXiv:1803.01164.

TensorFlow

install.package("tensorflow")

library(tensorflow)
add <- tf$add(tf$constant(1), tf$constant(1))

sess <- tf$Session()
tf$summary$FileWriter(file.path(".", "graph"), sess$graph)
sess$run(add)

[1] 2

tensorboard("graph")

keras

Deep Learning for humans.

library(keras)
mnist <- dataset_mnist()
x_train <- array_reshape(mnist$train$x, c(nrow(mnist$train$x), 784)) / 255
x_test <- array_reshape(mnist$test$x, c(nrow(mnist$test$x), 784)) / 255
y_train <- to_categorical(mnist$train$y, 10)
y_test <- to_categorical(mnist$test$y, 10)

model <- keras_model_sequential() 
model %>% 
  layer_dense(units = 256, activation = 'relu', input_shape = c(784)) %>% 
  layer_dropout(rate = 0.4) %>% 
  layer_dense(units = 128, activation = 'relu') %>%
  layer_dropout(rate = 0.3) %>%
  layer_dense(units = 10, activation = 'softmax')

model %>% compile(loss = 'categorical_crossentropy', optimizer = optimizer_rmsprop(), metrics = c('accuracy'))
model %>% fit(x_train, y_train, epochs = 3, batch_size = 128, validation_split = 0.2)
model %>% evaluate(x_test, y_test)

$acc
[1] 0.9681

pixels

Tools for Working with Pixels in R

install.packages("pixels")

library(pixels)
get_pixels()

kerasjs

Run Keras models in the browser, with GPU support using WebGL.

devtools::install_github("rstudio/kerasjs")

library(kerasjs)
kerasjs_convert("path-to-model.hdf5")

Show HN: Embedding Deep Learning models into your Slides using R

kartsim

Kart simulation for model training

devtools::install_github("javierluraschi/kartsim")

library(kartsim)
kartsim_play(circuit = 3)

How do I work with Python projects?

• reticulate: Call Python code from R.
• feather: Save and load data from R and Python.
• arrow: Share data in memory from R and other languages.

Feather

A lightweight binary columnar data store designed for maximum speed, based on Arrow’s memory layout.

library(feather)                                     # Import feather package

write_feather(mtcars, "cars.feather")                # Write feather file in R
read_feather("cars.feather")                         # Read feather file in R

import feather                                       # Import feather package

df = feather.read_dataframe("cars.feather")          # Read feather file in Python
feather.write_dataframe(df, "cars.feather")          # Write feather file in Python

Arrow

Apache Arrow is a cross-language development platform for in-memory data.

devtools::install_github("apache/arrow", subdir = "r", ref = "apache-arrow-0.13.0")

The R arrow package supports feather, parquet, streams, and more.

library(arrow)                                       # Import arrow package

read_feather("cars.feather")                         # Can still read feather file
read_parquet("cars.parquet")                         # Can also read parquet files

write_arrow(mtcars, raw())                           # Can efficiently serialize

Thanks!

• github: github.com/javierluraschi
• twitter: twitter.com/javierluraschi
• linkedin: linkedin.com/in/javierluraschi