Thoughts and experiments in data science
A practical guide to creating guardrails before AI use creates risk. - What’s your plan? Your organization may not have formally adopted AI, but your employees are probably already using it. It may be a personal AI account, a writing assistant inside office software, or an AI feature built into a vendor platform. That creates risk even when the organization has no...
Why it's important for a company to balance technology, vision, sales, and product. - A colleague recently shared a video entitled “Escaping the Build Trap”[1]. This video was the keynote address in the 2018 Productized conference. In the keynote address, Melissa Perri discussed how to build organizations that thrive off of product management. The talk is in part based on her book, Escaping the...
What happens when you try to run dictionary comprehension in the Python debugger (Pdb)? - This is primarily a note to remind myself of this bug the next time I encounter it. This issue is seen on Python version 3.10.6. I’m a big fan of using the Python and Ipython debuggers, pdb and ipdb respectively, when trying to debug code. Recently, I was debugging a...
Enhancing Particle Swarm Optimization with Local Optimization - The purpose of this post is to introduce a new two-step optimizing technique that utilizes Particle Swarm Optimization (PSO) and local numerical optimization. The code for this optimization method can be found here and is a work in progress. In real-world applications such as hyper-parameter tuning of machine learning models,...
Exploring Bayesian Optimization - Intent In this blog, post I will cover different approaches for gradient-free optimization. The main focus will be on Bayesian Optimization and in order to understand Bayesian Optimization, we will cover Gaussian processes. In full transparency, my original intent for the post was to show for a real-world example that...
Modeling human behavior with physical laws - Introduction Where is the intersection of social-economic behavior and mathematics? Is it possible to use fundamental equations to model population behavior? A while back a colleague, Kevin James, and I created an interactive report detailing a hypothetical forced migration scenario in Yemen. That report gave a high-level graphic overview of...
Investigating uncertainty in model predictions with practical examples - Introduction For every model prediction we make there is an underlying uncertainty in that prediction. One of the key aspects of operationalizing AI models, is to understand the uncertainties of the predictions. If you are going to make a decision or take an action based on a model that could...
Investigating the use of Dynamic Time Warping to test correlation between time series - Introduction This is a cautionary tale of using Dynamic Time Warping (DTW). We often see examples of successful applications of algorithms and models; however, this is an example when things do not go according to plan. Although we all enjoy talking about our success, I think we should take more...
An interactive report showing a hypothetical forced migration event in Yemen. This work was done in collaboration with Kevin James and links out to our Visual Perspective website -
Tracking changes in Iceland's vegetation through time. This work was done in collaboration with Kevin James and links out to our Visual Perspective website -
Using satellite data to observe Iceland's seasonal vegetation. This work was done in collaboration with Kevin James and links out to our Visual Perspective website -