Hi there, I'm @BaerVervergaert a mathematician, data scientist and writer. I'm interested in applying and developing machine learning code, as well as developing simulation code for historical geopolitical simulations. For more details, see below. Apart from my professional and educational interests I'm also a great fan of history, literature and the arts. On the occasion I muck about with pictures trying to create some bedazzling or haunting images through custom filters written in python. And I try my hand at writing poems and prose when I find the inspiration for it.
Concerning data science of a particular interest to me are online learning alogirthms. These algorithms, under the right conditions, can update the model to accomodate for new data. All of the modern gradient descent learners, that I know of, are an example of online learning. What I think is fascinating about these models is playing with the idea of convergence of its various parts. Usually, there is an interpretation of the formula that drives the model updates and by playing about with which parts need to converge or not we can attempt better learners. Per example, say we have a linear model for a time series. After having created the initial model we could update by a) refitting the model given every new data point or b) update using gradient descent. The a) method has the advantage of converging quickly, so if the data was relatively stationary (for example, the discharge of rivers) then we would quickly converge to a satisfying solution. On the other hand if the data changes over time (for example, market changes) then we would prefer method b) because it can more quickly adapt to the changes and 'forget' the older less significant data points. We might even prefer a mixture of the two methods.
I have studied mathematics for nearly ten years and it is near and dear to my heart. Simulatinously, I grew disillusioned with mathematics and have become critical of some of its ideas and the attitude of many of its practicioners. Mathematics claims absolute truth, and rightfully so; it was build that way. However, that is also its weakness. So much of modern mathematics is build upon classical logic which it presumes to be absolute. It is powerful in the cases of absolute truths, such as physics. On the other hand, it fails to give insight many situations outside the absolutes. Think of the field of history: it naturally deals with complicated situations, possible datings, trustworthiness of sources, interpretations. This should be a gold mine for possibility theorists, to study how we can best quantify the possibility or necessity of a hypothesis being true. Yet, possibility theory remains underdeveloped. Meanwhile, theoretical mathematics keep developing mathematics to better understand mathematics. Mathematics, is and remains the study of turning knowledge into more knowledge, but it will be better for it if it focusses on more types of knowledge than the absolutes.
Historical simulations have been a bit of an obsession of mine. We live in a wonderful world. You can dive into history and come across such immense moments of drama, commedy and tragedy: the Punic wars where Hannibal swears a blood oath never to be a friend of Rome, the army of Liechtenstein marching to war with 80 men and returning home with 81 because they made an italian friend along the way, and the youth lost during the first world war, gloriously they marched to the front and bitterly they wept as the war machine crunched through their comrades and friends. As a statistician I'm fascinated by how such a diverse range of emotions and stories can be brought forth by randomness, as a mathematician I'm interested in the structure behind the stories, as a data scientist I'm interested in how I can model those settings, and as a writer I'm thoroughly thrilled to see what the other three can come up with.
If you found this page and want to contact me. You can hit me up at bgmj (dot) ververgaert (at) gmail (dot) com, or try my LinkedIn at https://www.linkedin.com/in/baer-ververgaert-699500162/