Using a Double Deep Q Network to Learn a Game with Random States

Image by author

Lock N’ Roll was created in 2009 by Armor Games. It is a game for those with advanced levels of intelligence, the chosen few who are gifted with massive IQs and a disposition toward mathematical and probabilistic thinking. The average gamer couldn’t possibly grasp such sophistication and elevated amusement.

Or, maybe that’s just what I tell myself to justify the fact that a surprising few of my friends have even heard of it, one of my favorite time wasters. It is a game that requires placing different color dice (red, yellow, green, or blue, numbered 1–4) on a 4x4 grid…


Sometimes you deliver a Cadillac when a minivan would have sufficed

Photo by Clay Banks on Unsplash

There is such a thing as the right tool for the job. When you or your organization have access to expensive visualization software, such as Tableau, Power BI, or Domo, every problem looks like a nail and your program is the hammer. If you are a data analyst for such an organization, you will inevitably spend countless hours on one of these dashboards and imbed every conceivable answer to any question a senior decision-maker of the company could have, just to be greeted with an email a week later containing some request that makes it obvious that no one is…


Create user-friendly code and widgets to access and compare forecast results

Photo by Markus Spiske on Unsplash

You have a couple of series to forecast out 24 months, let’s say between 50 and 150. You don’t have time to spend to find the perfect model for each one, so you need some kind of generalizable approach. You could write a loop and apply a safe, albeit, not very dynamic method, such as simple exponential smoothing, moving average, or others. But you want something a little bit more precise to bring your accuracy up without an enormous time investment.

Over the past couple of years, I’ve confronted this same problem on a few occasions. I developed a Python…


Forecasting New Housing Starts with Holt Winters Exponential Smoothing, SARIMA, and SARIMAX(13)

Photo by Eric Muhr on Unsplash

When it comes to forecasting, there is (unfortunately) no one-size-fits-all solution. There are, however, ways to automate certain aspects of the process and still obtain good results.

I have spent the better part of the last year developing a forecasting module in Python that is as close to one-size-fits-all as I’ve seen. I’ve used it to produce accurate and timely forecasts that are easy to manipulate and compare. It incorporates time series and machine learning models from several Python libraries, and even a few unique to R integrated through the rpy2 library. …

Michael Keith

Data Scientist for the Utah Department of Health. My specialties mainly lie in forecasting time series, but I like to dabble in a little bit of everything.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store