(last edited 28.12.2022) You can find me on mastodon (@vielaugen@sigmoid.social, @Vielauge@muenchen.social), kaggle (Vielauge), git (Vielauge) and at home (Holger).

Who am I

Once upon a time (1975) I was born in Augsburg, Germany. I studied statistics (diploma) in Munich and lived there for a while. I married and had two children. My wife died 2019, after 2 years struggling with cancer. I am single parenting with my two sons born 2006 and 2010. I am working at an clinical research organization over 20 years now. In the moment I work only two days a week. I have a girl friend living 12 km away with two younger children. My main interests are: My family, data science and climate catastrophe.

Statistics, data science, crises, kaggle and fast.ai

Studies

While studying statistics I was working at Stablab. I was very excited about the versatility in this job. I helped other students of different fields (e.g economics, veternarians, fairy tales, wind speeds and others) with their statistics for their doctoral and other thesises. My own thesis was for BMW, regression analyzes in the context of crash tests.

Kaggle

After completing my diploma, my interest in new statistical methods was less pronounced. I stuck in my field and work. Some regressions with variable selection and a lot of descriptive statistics for the pharmaceutical industry. Everything programmed with SAS.

This changed 2011. At this time I found kaggle.com and was amazed about predictive modelling. Suddenly I was interested in other statistical methods since they outperformed my own regressions. State of the art was in these days was random forests. An outstanding performer was Jeremy Howard. I am still a very big fan of him, such an inspiring person! Big thanks to him, if he will ever read this. I don’t think so.

Over the next view years I did some kaggling with R and with very minor success. With two young childs, a job to do, a wife to love/deal (you can choose), time was very limited. At kaggle they began with Jupyter notebooks. The workflow was different to my own workstyle and it threw me somehow out! Unbelivable in restrospective, cause I am using Jupyter a lot now.

Next episode was deep learning and I guess this was the most dissatisfying phase. I had the wrong GPU and the language of the experts was so IT screwed. It was more about software/hardware installation than statistics. I gave up for another few years!

Interlude: Cancer and other crisis

No details to cancer here, but this was a stopper for my data science ambitions! The priorities were on my wife, the children and life must go on. After she died I got a lot of help from my parents and mother in law. With upcoming corona and isolation the help ended again very soon. Now things have a bit normalized. The multiple crisis (corona, ukrainian war, climate catastrophe) are there, but my personal crisis is almost over, having more time again. With a small widower’s pension and 16h regular work, two days a week, I have enough money to make ends meet and have enough time to focus again more on my main interests.

FASTAI

Luckily I followed Jeremy and even more lucky he founded FastAI! His book “Deep learning for Coders with fastai & PyTorch” is lying beside my keyboard. His book and his courses are an great inspiration. I started with the third edition. Although I was very motivated, I stuck again. The main reasons were: I am a statistician, but not a real coder/programmer. I have experience wit SAS and some R, but no experience with python. And the real horror is setting up IT (ubuntu, git, paperspace, hugging face….). The how to’s and tutorials are very good, but sometimes there is an unexpected error. And then well -nothing. Other priorities do the rest!

FASTAI 5th edition

I hope this time everything is different! Again motivated. 5th edition uses more Kaggle. This is a plus for me. Git, Paperspace was already there. Still struggling more with installation issues than with statistical concepts. Now I run python on Kaggle, or with visual studio code on windows. Many setup difficulties with Git, paperspace, Hugging face and my own windows ubuntu. Hope I can keep on jumping or crawling over these hurdles and invest more energy in the real thing (predictive modelling, random forest, deep learning, graphics and so on)

This blog is part of my ambitions and I am still burning for progress but with pauses (due to lack of energy and/or buddies to keep the spirit high)

Final words

Thank you for reading me. Next steps and plans hopefully in regular posts. Kind regards.

Vielauge aka Holger