Posted on Categories Administrativia, Exciting Techniques, Practical Data Science, Pragmatic Data Science, Pragmatic Machine Learning, StatisticsTags , , , ,

Big cdata News

I have some big news about our R package cdata. We have greatly improved the calling interface and Nina Zumel has just written the definitive introduction to cdata.

cdata is our general coordinatized data tool. It is what powers the deep learning performance graph (here demonstrated with R and Keras) that I announced a while ago.

KerasPlot

However, cdata is much more than that.

Continue reading Big cdata News

Posted on Categories data science, Pragmatic Data Science, Pragmatic Machine Learning, StatisticsTags , , , , 1 Comment on Announcing rquery

Announcing rquery

We are excited to announce the rquery R package.

rquery is Win-Vector LLC‘s currently in development big data query tool for R.

rquery supplies set of operators inspired by Edgar F. Codd‘s relational algebra (updated to reflect lessons learned from working with R, SQL, and dplyr at big data scale in production).

Continue reading Announcing rquery

Posted on Categories Exciting Techniques, Pragmatic Data Science, Pragmatic Machine Learning, Statistics, TutorialsTags , , , 2 Comments on Plotting Deep Learning Model Performance Trajectories

Plotting Deep Learning Model Performance Trajectories

I am excited to share a new deep learning model performance trajectory graph.

Here is an example produced based on Keras in R using ggplot2:

Unknown Continue reading Plotting Deep Learning Model Performance Trajectories

Posted on Categories data science, Exciting Techniques, Pragmatic Data Science, Pragmatic Machine Learning, Programming, Statistics, TutorialsTags , , , ,

How to Greatly Speed Up Your Spark Queries

For some time we have been teaching R users "when working with wide tables on Spark or on databases: narrow to the columns you really want to work with early in your analysis."

The idea behind the advice is: working with fewer columns makes for quicker queries.


speed

photo: Jacques Henri Lartigue 1912

The issue arises because wide tables (200 to 1000 columns) are quite common in big-data analytics projects. Often these are "denormalized marts" that are used to drive many different projects. For any one project only a small subset of the columns may be relevant in a calculation.

Continue reading How to Greatly Speed Up Your Spark Queries

Posted on Categories Administrativia, Opinion, Pragmatic Data Science, Pragmatic Machine Learning, Programming, StatisticsTags , , , , , , , , 4 Comments on Getting started with seplyr

Getting started with seplyr

A big “thank you!!!” to Microsoft for hosting our new introduction to seplyr. If you are working R and big data I think the seplyr package can be a valuable tool.


Safety
Continue reading Getting started with seplyr

Posted on Categories Coding, data science, Exciting Techniques, Pragmatic Data Science, Pragmatic Machine Learning, Statistics, TutorialsTags , , , , , , , , 1 Comment on Win-Vector LLC announces new “big data in R” tools

Win-Vector LLC announces new “big data in R” tools

Win-Vector LLC is proud to introduce two important new tool families (with documentation) in the 0.5.0 version of seplyr (also now available on CRAN):

  • partition_mutate_se() / partition_mutate_qt(): these are query planners/optimizers that work over dplyr::mutate() assignments. When using big-data systems through R (such as PostgreSQL or Apache Spark) these planners can make your code faster and sequence steps to avoid critical issues (the complementary problems of too long in-mutate dependence chains, of too many mutate steps, and incidental bugs; all explained in the linked tutorials).
  • if_else_device(): provides a dplyr::mutate() based simulation of per-row conditional blocks (including conditional assignment). This allows powerful imperative code (such as often seen in porting from SAS) to be directly and legibly translated into performant dplyr::mutate() data flow code that works on Spark (via Sparklyr) and databases.


Blacksmith working

Image by Jeff Kubina from Columbia, Maryland – [1], CC BY-SA 2.0, Link

Continue reading Win-Vector LLC announces new “big data in R” tools

Posted on Categories Coding, Pragmatic Data Science, Pragmatic Machine Learning, StatisticsTags , , , , , 3 Comments on Vectorized Block ifelse in R

Vectorized Block ifelse in R

Win-Vector LLC has been working on porting some significant large scale production systems from SAS to R.

From this experience we want to share how to simulate, in R with Apache Spark (via Sparklyr), a nifty SAS feature: the vectorized “block if(){}else{}” structure. Continue reading Vectorized Block ifelse in R

Posted on Categories Pragmatic Data Science, Pragmatic Machine Learning, Programming, Statistics, TutorialsTags , , , , , , ,

Data Wrangling at Scale

Just wrote a new R article: “Data Wrangling at Scale” (using Dirk Eddelbuettel’s tint template).

Fd

Please check it out.

Posted on Categories Exciting Techniques, Practical Data Science, Pragmatic Data Science, Pragmatic Machine Learning, Statistics, TutorialsTags , , , 2 Comments on Partial Pooling for Lower Variance Variable Encoding

Partial Pooling for Lower Variance Variable Encoding


Terraces
Banaue rice terraces. Photo: Jon Rawlinson

In a previous article, we showed the use of partial pooling, or hierarchical/multilevel models, for level coding high-cardinality categorical variables in vtreat. In this article, we will discuss a little more about the how and why of partial pooling in R.

We will use the lme4 package to fit the hierarchical models. The acronym “lme” stands for “linear mixed-effects” models: models that combine so-called “fixed effects” and “random effects” in a single (generalized) linear model. The lme4 documentation uses the random/fixed effects terminology, but we are going to follow Gelman and Hill, and avoid the use of the terms “fixed” and “random” effects.

The varying coefficients [corresponding to the levels of a categorical variable] in a multilevel model are sometimes called random effects, a term that refers to the randomness in the probability model for the group-level coefficients….

The term fixed effects is used in contrast to random effects – but not in a consistent way! … Because of the conflicting definitions and advice, we will avoid the terms “fixed” and “random” entirely, and focus on the description of the model itself…

– Gelman and Hill 2007, Chapter 11.4

We will also restrict ourselves to the case that vtreat considers: partially pooled estimates of conditional group expectations, with no other predictors considered.

Continue reading Partial Pooling for Lower Variance Variable Encoding

Posted on Categories data science, Exciting Techniques, Practical Data Science, Pragmatic Data Science, Pragmatic Machine Learning, Statistics, TutorialsTags , , ,

Custom Level Coding in vtreat

One of the services that the R package vtreat provides is level coding (what we sometimes call impact coding): converting the levels of a categorical variable to a meaningful and concise single numeric variable, rather than coding them as indicator variables (AKA "one-hot encoding"). Level coding can be computationally and statistically preferable to one-hot encoding for variables that have an extremely large number of possible levels.

Speed

Level coding is like measurement: it summarizes categories of individuals into useful numbers. Source: USGS

By default, vtreat level codes to the difference between the conditional means and the grand mean (catN variables) when the outcome is numeric, and to the difference between the conditional log-likelihood and global log-likelihood of the target class (catB variables) when the outcome is categorical. These aren’t the only possible level codings. For example, the ranger package can encode categorical variables as ordinals, sorted by the conditional expectations/means. While this is not a completely faithful encoding for all possible models (it is not completely faithful for linear or logistic regression, for example), it is often invertible for tree-based methods, and has the advantage of keeping the original levels distinct, which impact coding may not. That is, two levels with the same conditional expectation would be conflated by vtreat‘s coding. This often isn’t a problem — but sometimes, it may be.

So the data scientist may want to use a level coding different from what vtreat defaults to. In this article, we will demonstrate how to implement custom level encoders in vtreat. We assume you are familiar with the basics of vtreat: the types of derived variables, how to create and apply a treatment plan, etc.

Continue reading Custom Level Coding in vtreat