In many machine learning modeling techniques, the goal is to find model parameters which produce the least error in predicting outcomes given a particular set of data on which the model can be trained. In Bayesian modeling, the underlying philosophy changes slightly, and the goal becomes to find model parameters which are most likely to explain the observed data. The reason “Bayesian” is attached to this underlying thought is that one can ask the mathematical question “Given the data, which parameters are most likely to explain the outcome?”
Tags: bayesian, modelingThu, Jan 10, 2019
Here at Open Data Group, we’re big proponents of delivering software and models as containerized microservices. In fact, it’s a core part of our value proposition! Because we find ourselves doing this a lot, our team has standardized around a systematic approach using existing tooling to allow us to rapidly prototype and develop production-grade containerized applications. In this blog post, I’ll talk about three parts to this approach, specifically focusing on applications implemented in Go.
Tags: docker, swagger, golangThu, Dec 20, 2018
If I could predict the future, some aspects of business would become much easier. For instance, if I owned a store and could somehow divine what my next week’s sales would look like, I could make sure that my inventory was perfectly matched to the coming demand. Similarly, if I could predict the movements of the stock market, I could use this information to make advantageous trades. Sadly, I can’t see into the future. So the natural question becomes: what are the best predictions I can make about the future given the information that is available to me today?