Here is an interview with Zach Goldberg, who is the product manager of Google Prediction API, the next generation machine learning analytics-as-an-api service state of the art cloud computing model building browser app. Ajay- Describe your journey in science and technology from high school to your current job at Google.
Zach- First, thanks so much for the opportunity to do this interview Ajay! My personal journey started in college where I worked at a startup named Invite Media. From there I transferred to the Associate Product Manager (APM) program at Google. The APM program is a two year rotational program. I did my first year working in display advertising. After that I rotated to work on the Prediction API.
Ajay- How does the Google Prediction API help an average business analytics customer who is already using enterprise software , servers to generate his business forecasts. How does Google Prediction API fit in or complement other APIs in the Google API suite.
Zach- The Google Prediction API is a cloud based machine learning API. We offer the ability for anybody to sign up and within a few minutes have their data uploaded to the cloud, a model built and an API to make predictions from anywhere. Traditionally the task of implementing predictive analytics inside an application required a fair amount of domain knowledge; you had to know a fair bit about machine learning to make it work. With the Google Prediction API you only need to know how to use an online REST API to get started.
Ajay- What are the additional use cases of Google Prediction API that you think traditional enterprise software in business analytics ignore, or are not so strong on. What use cases would you suggest NOT using Google Prediction API for an enterprise.
Zach- We are living in a world that is changing rapidly thanks to technology. Storing, accessing, and managing information is much easier and more affordable than it was even a few years ago. That creates exciting opportunities for companies, and we hope the Prediction API will help them derive value from their data.
The Prediction API focuses on providing predictive solutions to two types of problems: regression and classification. Businesses facing problems where there is sufficient data to describe an underlying pattern in either of these two areas can expect to derive value from using the Prediction API.
Ajay- What are your separate incentives to teach about Google APIs to academic or researchers in universities globally.
Google thrives on academic curiosity. While we do significant in-house research and engineering, we also maintain strong relations with leading academic institutions world-wide pursuing research in areas of common interest. As part of our mission to build the most advanced and usable methods for information access, we support university research, technological innovation and the teaching and learning experience through a variety of programs.
Ajay- What is the biggest challenge you face while communicating about Google Prediction API to traditional users of enterprise software.
Zach- Businesses often expect that implementing predictive analytics is going to be very expensive and require a lot of resources. Many have already begun investing heavily in this area. Quite often we’re faced with surprise, and even skepticism, when they see the simplicity of the Google Prediction API. We work really hard to provide a very powerful solution and take care of the complexity of building high quality models behind the scenes so businesses can focus more on building their business and less on machine learning.
To join the preview group, go to the APIs Console and click the Prediction API slider to “ON,” and then sign up for a Google Storage account.
For the past several months, I have been member of a semi-public beta test/group/forum – that is headed by Travis Green of the Google Prediction API Team (not the hockey player). Basically in helping the Google guys more feedback on the feature list for model building via cloud computing. I couldn’t talk about it much , because it was all NDA hush hush.
Anyways- as of today the version 1.2 of Google Prediction API has been launched. What does this do to the ordinary Joe Modeler? Well it helps gives your models -thats right your plain vanilla logistic regression,arima, arimax, models an added ensemble option of using Google’s Machine Learning Continue reading “Google releases V1.2 of Google Prediction API”
A workshop on using R for Predictive Modeling, by the Director, Non Clinical Stats, Pfizer. Interesting Bay Area Event- part of next edition of Predictive Analytics World
Sunday, March 13, 2011 in San Francisco
R for Predictive Modeling:
A Hands-On Introduction
Intended Audience: Practitioners who wish to learn how to execute on predictive analytics by way of the R language; anyone who wants “to turn ideas into software, quickly and faithfully.”
Knowledge Level: Either hands-on experience with predictive modeling (without R) or hands-on familiarity with any programming language (other than R) is sufficient background and preparation to participate in this workshop.
This one-day session provides a hands-on introduction to R, the well-known open-source platform for data analysis. Real examples are employed in order to methodically expose attendees to best practices driving R and its rich set of predictive modeling packages, providing hands-on experience and know-how. R is compared to other data analysis platforms, and common pitfalls in using R are addressed.
The instructor, a leading R developer and the creator of CARET, a core R package that streamlines the process for creating predictive models, will guide attendees on hands-on execution with R, covering:
A working knowledge of the R system
The strengths and limitations of the R language
Preparing data with R, including splitting, resampling and variable creation
Developing predictive models with R, including decision trees, support vector machines and ensemble methods
Visualization: Exploratory Data Analysis (EDA), and tools that persuade
Evaluating predictive models, including viewing lift curves, variable importance and avoiding overfitting
Hardware: Bring Your Own Laptop
Each workshop participant is required to bring their own laptop running Windows or OS X. The software used during this training program, R, is free and readily available for download.
Attendees receive an electronic copy of the course materials and related R code at the conclusion of the workshop.
Ubuntu has a slight glitch plus workaround for installing the RCurl package on which the Google Prediction API is dependent- you need to first install this Ubuntu package for RCurl to install libcurl4-gnutls-dev
Once you install that using Synaptic,
Simply start R
2) Install Packages rjson and Rcurl using install.packages and choosing CRAN
Once you type in the basic syntax, the first time it will ask for your Google Credentials (email and password)
It then starts showing you time elapsed for training.
Now you can disconnect and go off (actually I got disconnected by accident before coming back in a say 5 minutes so this is the part where I think this is what happened is why it happened, dont blame me, test it for yourself) –
and when you come back (hopefully before token expires) you can see status of your request (see below)
Loading required package: bitops
> my.model <- PredictionApiTrain(data="gs://numtraindata/training_data")
The request for training has sent, now trying to check if training is completed
Training on numtraindata/training_data: time:2.09 seconds
Training on numtraindata/training_data: time:7.00 seconds
Note I changed the format from the URL where my data is located- simply go to your Google Storage Manager and right click on the file name for link address ( https://sandbox.google.com/storage/numtraindata/training_data.csv)
to gs://numtraindata/training_data (that kind of helps in any syntax error)
## Load googlepredictionapi and dependent libraries
## Make a training call to the Prediction API against data in the Google Storage.
## Replace MYBUCKET and MYDATA with your data.
my.model <- PredictionApiTrain(data="gs://MYBUCKET/MYDATA")
## Alternatively, make a training call against training data stored locally as a CSV file.
## Replace MYPATH and MYFILE with your data.
my.model <- PredictionApiTrain(data="MYPATH/MYFILE.csv")
At the time of writing my data was still getting trained, so I will keep you posted on what happens.
The new Google Ad Planner is really nice-seems better than old Adwords interface, though needs a UI redesign before it can complete with the clean cut slice and dice of Facebook Ad Planner.
It’s the interface, stupid that makes an Iphone sell more than the Symbian even with 90% functionality. Same reasons why Google Storage is okay but Google Prediction API gets slower liftoff than Amazon Console (now with FREE instances) – though the R interface to Prediction API sure helps.
Prediction API is a terrific tool dying for oxygen out there (and will end up like Wave- I hope not)
Sometimes you need artists as well as engineers to design query tools, G Men- and guess the Double Click anti trust rumours have quietened down enough because why the heck did double click interface integration take so loooong.
( and btw why cant Google just get into the multi billion dashboard business if they can manage ALL the data IN THE INTERNET ——they sure can do it for specific companies- – but wait-
they are probably waiting for AsterData to stop sucking thumbs ,chanting on MapReduceSQL, MapReduce SQL nursery rhymes and start inventing NEW STUFF again (or atleast creating two product brands from nCluster (when you and I were in school together giggle)
Btw the time Google make up their mind to enter BI or wait for Aster to finish- IBM would have gulped and burped all there it is- and thats the way that market rolls.
Back to Ad s and Mad Men.
Here are some screenshots-of the new Google Ad Planner-
I found it useful to review traffic for third party websites (even better than Google Trends) and thats a definite plus over Facebooks closed dormitory world of ads.
Here is a short list of resources and material I put together as starting points for R and Cloud Computing It’s a bit messy but overall should serve quite comprehensively.
Cloud computing is a commonly used expression to imply a generational change in computing from desktop-servers to remote and massive computing connections,shared computers, enabled by high bandwidth across the internet.
As per the National Institute of Standards and Technology Definition,
Cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.
The paper “Rweb: Web-based Statistical Analysis”, providing a detailed explanation of the different versions of Rweb and an overview of how Rweb works, was published in the Journal of Statistical Software (http://www.jstatsoft.org/v04/i01/).
Rcgi is a CGI WWW interface to R by MJ Ray. It had the ability to use “embedded code”: you could mix user input and code, allowing the HTMLauthor to do anything from load in data sets to enter most of the commands for users without writing CGI scripts. Graphical output was possible in PostScript or GIF formats and the executed code was presented to the user for revision. However, it is not clear if the project is still active.
Currently, a modified version of Rcgi by Mai Zhou (actually, two versions: one with (bitmap) graphics and one without) as well as the original code are available from http://www.ms.uky.edu/~statweb/.
David Firth has written CGIwithR, an R add-on package available from CRAN. It provides some simple extensions to R to facilitate running R scripts through the CGI interface to a web server, and allows submission of data using both GET and POST methods. It is easily installed using Apache under Linux and in principle should run on any platform that supports R and a web server provided that the installer has the necessary security permissions. David’s paper “CGIwithR: Facilities for Processing Web Forms Using R” was published in the Journal of Statistical Software (http://www.jstatsoft.org/v08/i10/). The package is now maintained by Duncan Temple Lang and has a web page athttp://www.omegahat.org/CGIwithR/.
Jeff Horner is working on the R/Apache Integration Project which embeds the R interpreter inside Apache 2 (and beyond). A tutorial and presentation are available from the project web page at http://biostat.mc.vanderbilt.edu/twiki/bin/view/Main/RApacheProject.
Rserve is a project actively developed by Simon Urbanek. It implements a TCP/IP server which allows other programs to use facilities of R. Clients are available from the web site for Java and C++ (and could be written for other languages that support TCP/IP sockets).
OpenStatServer is being developed by a team lead by Greg Warnes; it aims “to provide clean access to computational modules defined in a variety of computational environments (R, SAS, Matlab, etc) via a single well-defined client interface” and to turn computational services into web services.
Two projects use PHP to provide a web interface to R. R_PHP_Online by Steve Chen (though it is unclear if this project is still active) is somewhat similar to the above Rcgi and Rweb. R-php is actively developed by Alfredo Pontillo and Angelo Mineo and provides both a web interface to R and a set of pre-specified analyses that need no R code input.
webbioc is “an integrated web interface for doing microarray analysis using several of the Bioconductor packages” and is designed to be installed at local sites as a shared computing resource.
Rwui is a web application to create user-friendly web interfaces for R scripts. All code for the web interface is created automatically. There is no need for the user to do any extra scripting or learn any new scripting techniques. Rwui can also be found at http://rwui.cryst.bbk.ac.uk.
Finally, the R.rsp package by Henrik Bengtsson introduces “R Server Pages”. Analogous to Java Server Pages, an R server page is typically HTMLwith embedded R code that gets evaluated when the page is requested. The package includes an internal cross-platform HTTP server implemented in Tcl, so provides a good framework for including web-based user interfaces in packages. The approach is similar to the use of the brew package withRapache with the advantage of cross-platform support and easy installation.
Remote access to R/Bioconductor on EBI’s 64-bit Linux Cluster
Start the workbench by downloading the package for your operating system (Macintosh or Windows), or via Java Web Start, and you will get access to an instance of R running on one of EBI’s powerful machines. You can install additional packages, upload your own data, work with graphics and collaborate with colleagues, all as if you are running R locally, but unlimited by your machine’s memory, processor or data storage capacity.
Most up-to-date R version built for multicore CPUs
Access to all Bioconductor packages
Access to our computing infrastructure
Fast access to data stored in EBI’s repositories (e.g., public microarray data in ArrayExpress)
Amazon’s EC2 is a type of cloud that provides on demand computing infrastructures called an Amazon Machine Images or AMIs. In general, these types of cloud provide several benefits:
Simple and convenient to use. An AMI contains your applications, libraries, data and all associated configuration settings. You simply access it. You don’t need to configure it. This applies not only to applications like R, but also can include any third-party data that you require.
On-demand availability. AMIs are available over the Internet whenever you need them. You can configure the AMIs yourself without involving the service provider. You don’t need to order any hardware and set it up.
Elastic access. With elastic access, you can rapidly provision and access the additional resources you need. Again, no human intervention from the service provider is required. This type of elastic capacity can be used to handle surge requirements when you might need many machines for a short time in order to complete a computation.
Pay per use. The cost of 1 AMI for 100 hours and 100 AMI for 1 hour is the same. With pay per use pricing, which is sometimes called utility pricing, you simply pay for the resources that you use.
#This example requires you had previously created a bucket named data_language on your Google Storage and you had uploaded a CSV file named language_id.txt (your data) into this bucket – see for details
Elastic-R is a new portal built using the Biocep-R platform. It enables statisticians, computational scientists, financial analysts, educators and students to use cloud resources seamlessly; to work with R engines and use their full capabilities from within simple browsers; to collaborate, share and reuse functions, algorithms, user interfaces, R sessions, servers; and to perform elastic distributed computing with any number of virtual machines to solve computationally intensive problems.
Also see Karim Chine’s http://biocep-distrib.r-forge.r-project.org/
R for Salesforce.com
At the point of writing this, there seem to be zero R based apps on Salesforce.com This could be a big opportunity for developers as both Apex and R have similar structures Developers could write free code in R and charge for their translated version in Apex on Salesforce.com
Personal Note-Mentioning SAS in an email to a R list is a big no-no in terms of getting a response and love. Same for being careless about which R help list to email (like R devel or R packages or R help)