Data Skeptic

When dealing with dynamic systems that are potentially undergoing constant change, its helpful to describe what "state" they are in.  In many applications the manner in which the state changes from one to another is not completely predictable, thus, there is uncertainty over how it transitions from state to state.  Further, in many applications, one cannot directly observe the true state, and thus we describe such situations as partially observable state spaces.  This episode explores what this means and why it is important in the context of chess, poker, and the mood of Yoshi the lilac crowned amazon parrot.

My guest this week is Anh Nguyen, a PhD student at the University of Wyoming working in the Evolving AI lab. The episode discusses the paper Deep Neural Networks are Easily Fooled [pdf] by Anh Nguyen, Jason Yosinski, and Jeff Clune. It describes a process for creating images that a trained deep neural network will mis-classify. If you have a deep neural network that has been trained to recognize certain types of objects in images, these "fooling" images can be constructed in a way which the network will mis-classify them. To a human observer, these fooling images often have no resemblance whatsoever to the assigned label. Previous work had shown that some images which appear to be unrecognizable white noise images to us can fool a deep neural network. This paper extends the result showing abstract images of shapes and colors, many of which have form (just not the one the network thinks) can also trick the network.

This episode introduces a high level discussion on the topic of Data Provenance, with more MINI episodes to follow to get into specific topics. Thanks to listener Sara L who wrote in to point out the Data Skeptic Podcast has focused alot about using data to be skeptical, but not necessarily being skeptical of data. Data Provenance is the concept of knowing the full origin of your dataset. Where did it come from? Who collected it? How as it collected? Does it combine independent sources or one singular source? What are the error bounds on the way it was measured? These are just some of the questions one should ask to understand their data. After all, if the antecedent of an argument is built on dubious grounds, the consequent of the argument is equally dubious. For a more technical discussion than what we get into in this mini epiosode, I recommend A Survey of Data Provenance Techniques by authors Simmhan, Plale, and Gannon.

I had the change to speak with well known Sharon Hill (@idoubtit) for the first episode of 2015. We discuss a number of interesting topics including the contributions Doubtful News makes to getting scientific and skeptical information ranked highly in search results, sink holes, why earthquakes are hard to predict, and data collection about paranormal groups via the internet.

In this quick holiday episode, we touch on how one would approach modeling the statistical distribution over the probability of belief in Santa Claus given age.

Economist Peter Backus joins me in this episode to discuss a few interesting topics. You may recall Linhda and I previously discussed his paper "The Girlfriend Equation" on a recent mini-episode. We start by touching base on this fun paper and get a follow up on where Peter stands years after writing w.r.t. a successful romantic union. Additionally, we delve in to some fascinating economics topics. We touch on questions of the role models, for better or for worse, played a role in the ~2008 economic crash, statistics in economics and the difficulty of measurement, and some insightful discussion about the economics charities. Peter encourages listeners to be open to giving money to charities that are good at fundraising, and his arguement is a (for me) suprisingly insightful logic. Lastly, we have a teaser of some of Peter's upcoming work using unconventional data sources. For his benevolent recommendation, Peter recommended the book The Conquest of Happiness by Bertrand Russell, and for his self-serving recom

Love and Data is the continued theme in this mini-episode as we discuss the game theory example of The Battle of the Sexes. In this textbook example, a couple must strategize about how to spend their Friday night. One partner prefers football games while the other partner prefers to attend the opera. Yet, each person would rather be at their non-preferred location so long as they are still with their spouse. So where should they decide to go?

Can algorithms help you find love? Many happy couples successfully brought together via online dating websites show us that data science can help you find love. I'm joined this week by Thomas Levi, Senior Data Scientist at Plenty of Fish, to discuss some of his work which helps people find one another as efficiently as possible. Matchmaking is a truly non-trivial problem, and one that's dynamically changing all the time as new users join and leave the "pool of fish". This episode explores the aspects of what makes this a tough problem and some of the ways POF has been successfully using data science to solve it, and continues to try to innovate with new techniques like interest matching. For his benevolent references, Thomas suggests readers check out All of Statistics as well as the caretlibrary for R. And for a self serving recommendation, follow him on twitter (@tslevi) or connect withThomas Levi on Linkedin.

Economist Peter Backus put forward "The Girlfriend Equation" while working on his PhD - a probabilistic model attempting to estimate the likelihood of him finding a girlfriend. In this mini episode we explore the soundness of his model and also share some stories about how Linhda and Kyle met.

I'm joined this week by Alex Boklin to explore the topic of magical thinking especially in the context of Rhonda Byrne's "The Secret", and the similarities it bears to The Global Consciousness Project (GCP). The GCP puts forward the hypothesis that random number generators elicit statistically significant changes as a result of major world events.

What is randomness? How can we determine if some results are randomly generated or not? Why are random numbers important to us in our everyday life? These topics and more are discussed in this mini-episode on random numbers. Many readers will be vaguely familar with the idea of "X number of monkeys banging on Y number of typewriters for Z number of years" - the idea being that such a setup would produce random sequences of letters. The origin of this idea was the mathemetician Borel who was interested in whether or not 1,000,000 monkeys working for 10 hours per day might eventually reproduce the works of shakespeare. We explore this topic and provide some further details in the show notes which you can find over at dataskeptic.com

This week's episode explores the possibilities of extracting novel insights from the many great social web APIs available. Matthew Russell's Mining the Social Web is a fantastic exploration of the tools and methods, and we explore a few related topics. One helpful feature of the book is it's use of a Vagrant virtual machine. Using it, readers can easily reproduce the examples from the book, and there's a short video available that will walk you through setting up the Mining the Social Web virtual machine. The book also has an accompanying github repository which can be found here. A quote from Matthew that particularly reasonates for me was "The first commandment of Data Science is to 'Know thy data'." Take a listen for a little more context around this sage advice. In addition to the book, we also discuss some of the work done by Digital Reasoning where Matthew serves as CTO. One of their products we spend some time discussing is Synthesys, a service that processes unstructured data and delivers knowled

This episode explores the basis of why we can trust encryption.  Suprisingly, a discussion of looking up a word in the dictionary (binary search) and efficiently going wine tasting (the travelling salesman problem) help introduce computational complexity as well as the P ?= NP question, which is paramount to the trustworthiness RSA encryption. With a high level foundation of computational theory, we talk about NP problems, and why prime factorization is a difficult problem, thus making it a great basis for the RSA encryption algorithm, which most of the internet uses to encrypt data.  Unlike the encryption scheme Ray Romano used in "Everybody Loves Raymond", RSA has nice theoretical foundations. It should be noted that although this episode gives good reason to trust that properly encrypted data, based on well choosen public/private keys where the private key is not compromised, is safe.  However, having safe encryption doesn't necessarily mean that the Internet is secure.  Topics like Man in the Middle attac

Jeff Stanton joins me in this episode to discuss his book An Introduction to Data Science, and some of the unique challenges and issues faced by someone doing applied data science. A challenge to any data scientist is making sure they have a good input data set and apply any necessary data munging steps before their analysis. We cover some good advise for how to approach such problems.

The t-test is this week's mini-episode topic. The t-test is a statistical testing procedure used to determine if the mean of two datasets differs by a statistically significant amount. We discuss how a wine manufacturer might apply a t-test to determine if the sweetness, acidity, or some other property of two separate grape vines might differ in a statistically meaningful way.

This week I'm joined by Karl Mamer to discuss the data behind three well known urban legends. Did a large blackout in New York and surrounding areas result in a baby boom nine months later? Do subliminal messages affect our behavior? Is placing beer alongside diapers a recipe for generating more revenue than these products in separate locations? Listen as Karl and I explore these claims.

The Data Skeptic Podcast is launching a contest- not one of chance, but one of skill. Listeners are encouraged to put their data science skills to good use, or if all else fails, guess! The contest works as follows. Below is some data about the cumulative number of downloads the podcast has achieved on a few given dates. Your job is to predict the date and time at which the podcast will recieve download number 27,182. Why this arbitrary number? It's as good as any other arbitrary number! Use whatever means you want to formulate a prediction. Once you have it, wait until that time and then post a review of the Data Skeptic Podcast on iTunes. You don't even have to leave a good review! The review which is posted closest to the actual time at which this download occurs will win a free copy of Matthew Russell's "Mining the Social Web" courtesy of the Data Skeptic Podcast. "Price is Right" rules are in play - the winner is the person that posts their review closest to the actual time without going over. More infor

A discussion about conducting US presidential election polls helps frame a converation about selection bias.

Commute times and BBQ invites help frame a discussion about the statistical concept of confidence intervals.

A discussion about getting ready in the morning, negotiating a used car purchase, and selecting the best AirBnB place to stay at help frame a conversation about the decision theoretic principal known as the Value of Information equation.