Video of “Automating Complex Data Analysis” Presentation to the Bay Area SAS Users Group

 

This is an edited video of my presentation on “Automating Complex Data Analysis” to the Bay Area SAS Users Group (BASAS) on August 31, 2017 at Building 42, Genentech in South San Francisco, CA.

The demonstration of the Analyst in a Box prototype starts at 14:10 (14 minutes, 10 seconds). The demo is a video screen capture with high quality audio.

Unfortunately there was some background noise from a party in the adjacent room starting about 12:20 until 14:10 although my voice is understandable.

Updated slides for the presentation are available at: https://goo.gl/Gohw87

You can find out more about the Bay Area SAS Users Group at http://www.basas.com/

Abstract:

Complex data analysis attempts to solve problems with one or more inputs and one or more outputs related by complex mathematical rules, usually a sequence of two or more non-linear functions applied iteratively to the inputs and intermediate computed values. A prominent example is determining the causes and possible treatments for poorly understood diseases such as heart disease, cancer, and autism spectrum disorders where multiple genetic and environmental factors may contribute to the disease and the disease has multiple symptoms and metrics, e.g. blood pressure, heart rate, and heart rate variability.

Another example are macroeconomic models predicting employment levels, inflation, economic growth, foreign exchange rates and other key economic variables for investment decisions, both public and private, from inputs such as government spending, budget deficits, national debt, population growth, immigration, and many other factors.

A third example is speech recognition where a complex non-linear function somehow maps from a simple sequence of audio measurements — the microphone sound pressure levels — to a simple sequence of recognized words: “I’m sorry Dave. I can’t do that.”

State-of-the-art complex data analysis is labor intensive, time consuming, and error prone — requiring highly skilled analysts, often Ph.D.’s or other highly educated professionals, using tools with large libraries of built-in statistical and data analytical methods and tests: SAS, MATLAB, the R statistical programming language and similar tools. Results often take months or even years to produce, are often difficult to reproduce, difficult to present convincingly to non-specialists, difficult to audit for regulatory compliance and investor due diligence, and sometimes simply wrong, especially where the data involves human subjects or human society.

A widely cited report from the McKinsey management consulting firm suggests that the United States may face a shortage of 140,000 to 190,000 such human analysts by 2018: http://www.mckinsey.com/business-functions/digital-mckinsey/our-insights/big-data-the-next-frontier-for-innovation.

This talk discusses the current state-of-the-art in attempts to automate complex data analysis. It discusses widely used tools such as SAS and MATLAB and their current limitations. It discusses what the automation of complex data analysis may look like in the future, possible methods of automating complex data analysis, and problems and pitfalls of automating complex data analysis. The talk will include a demonstration of a prototype system for automating complex data analysis including automated generation of SAS analysis code.

(C) 2017 John F. McGowan, Ph.D.

About the author

John F. McGowan, Ph.D. solves problems using mathematics and mathematical software, including developing gesture recognition for touch devices, video compression and speech recognition technologies. He has extensive experience developing software in C, C++, MATLAB, Python, Visual Basic and many other programming languages. He has been a Visiting Scholar at HP Labs developing computer vision algorithms and software for mobile devices. He has worked as a contractor at NASA Ames Research Center involved in the research and development of image and video processing algorithms and technology. He has published articles on the origin and evolution of life, the exploration of Mars (anticipating the discovery of methane on Mars), and cheap access to space. He has a Ph.D. in physics from the University of Illinois at Urbana-Champaign and a B.S. in physics from the California Institute of Technology (Caltech).

 

Machine Learning at Google Event

I attended a “Machine Learning at Google” event at the Google Quad 3 building off Ellis in Mountain View last night (August 23, 2017).  This seemed to be mostly a recruiting event for some or all of Google’s high profile Machine Learning/Deep Learning groups, notably the team responsible for TensorFlow.

Token Good Looking Woman Opens Event
Woman Opens Event

I had no trouble finding the registration table when I arrived and getting my badge.  All the presentations seemed to run on time or nearly on time.  There was free food, a cute bag with Google gewgaws, and plenty of seating (about 280 seats with attendance about 240 I thought).

The event invitation that I received was rather vague and it did not become clear this was a recruiting event until well into the event.  It had the alluring title:

An Exclusive Invite | Machine Learning @ Google

Ooh, exclusive!  Aren’t I special!  Along with 240 other attendees as it turned out.  🙂

Andrew Zaldivar (see below) explicitly called it a recruiting event in the Q&A panel at the end.  It would have been good to know this as I am not looking for a job at Google. That does not mean the event wasn’t interesting to me for other reasons, but Google and other companies should be up front about this.

Although I think the speakers were on a low platform, they weren’t up high enough to see that well, even though I was in the front.  This was particularly true of Jasmine Hsu who was short.  I managed to get one picture of her not fully or mostly obscured by someone’s head.  Probably a higher platform for the presenters would have helped.

A good looking woman who seemed to be some sort of public relations or marketing person opened the event at 6:30 PM.  She went through all the usual event housekeeping and played a slick Madison Avenue style video on the coming wonders of machine learning.  Then she introduced the keynote speaker Ravi Kumar.

Ravi Kumar Keynote
Ravi Kumar Keynote

Ravi was followed by a series of “lightning talks” on machine learning and deep learning at Google by Sandeep Tata, Heng-Tze Cheng, Ian Goodfellow, James Kunz, Jasmine Hsu, and Andrew Zaldivar.

The presentations tended to blur together.  The typical machine learning/deep learning presentation is an extremely complex model that has been fitted to a very large data set.  Giant companies like Google and Facebook have huge proprietary data sets that few others can match.  The presenters tend to be very confident and assert major advances over past methods and often to match or exceed human performance.  It is often impossible to evaluate these claims without access to both the huge data sets and vast computing power.  People who try to duplicate the reported dramatic results  with more modest resources often report failure.

The presentations often avoid the goodness-of-fit statistics, robustness, and overfitting issues that experts in mathematical modeling worry about with such complex models.  A very complex model such as a polynomial with thousands of terms can always fit a data set but it will usually fail to extrapolate outside the data set correctly.  Polynomials, for example, always blow up to plus or minus infinity as the largest power term dominates.

In fact one Google presenter mentioned a “training-server skew” problem where the field data would frequently fail to match the training data  used for the model.  If I understood his comments, this seemed to occur almost every time supposedly for different reasons for each model.  This sounded a lot like the frequent failure of complex models to extrapolate to new data correctly.

Ravi Kumar’s keynote presentation appeared to be a maximum likelihood estimation (MLE) of a complex model of repeat consumption by users: how often, for example, a user will replay the same song or YouTube video.  MLE is not a robust estimation method and it is vulnerable to outliers in the data, almost a given in real data, yet there seemed to be no discussion of this issue in the presentation.

Often when researchers and practitioners from other fields that make heavy use of mathematical modeling such as statistics or physics bring up these issues, the machine learning/deep learning folks either circle the wagons and deny the issues or assert dismissively that they have the issues under control.  Move on, nothing to see here.

Sandeep Tata
Sandeep Tata
Hang Tze
Hang Tze
Ian Goodfellow on Deep Learning Research at Google
Ian Goodfellow on Deep Learning Research at Google
Jasmine Hsu on Robotics and Computer Vision
Jasmine Hsu on Robotics and Computer Vision
James Kunz
James Kunz
Andrew Zaldivar on SPAM Fighting with Machine Learning
Andrew Zaldivar on SPAM Fighting with Machine Learning

Andrew Zaldivar introduced the Q&A panel for which he acted as moderator.  Instead of having audience members take the microphone and ask their questions uncensored as many events do, he read out questions supposedly submitted by e-mail or social media.

Andrew Zaldivar Introduces the Panel
Andrew Zaldivar Introduces the Panel
Q and A Panel
Q and A Panel

The Q&A panel was followed by a reception from 8-9 PM to “meet the speakers.”  It was difficult to see how this would work with about thirty (30) audience members for each presenter.  I did not stay for the reception.

Conclusion

I found the presentations interesting but they did not go into most of the deeper technical questions such as goodness-of-fit, robustness, and overfitting that I would have liked to hear.  I feel Google should have been clearer about the purpose of the event up front.

(C) 2017 John F. McGowan, Ph.D.

About the author

John F. McGowan, Ph.D. solves problems using mathematics and mathematical software, including developing gesture recognition for touch devices, video compression and speech recognition technologies. He has extensive experience developing software in C, C++, MATLAB, Python, Visual Basic and many other programming languages. He has been a Visiting Scholar at HP Labs developing computer vision algorithms and software for mobile devices. He has worked as a contractor at NASA Ames Research Center involved in the research and development of image and video processing algorithms and technology. He has published articles on the origin and evolution of life, the exploration of Mars (anticipating the discovery of methane on Mars), and cheap access to space. He has a Ph.D. in physics from the University of Illinois at Urbana-Champaign and a B.S. in physics from the California Institute of Technology (Caltech).