This page now contains a list of my creations. See here for an archive of worthy things on internet I have thought worth collecting.
My academic creations
- Most up-to-date collection of publications can be found at my Google Scholar profile (2018 – ).
- My M.Sc. thesis (“X-ray computed tomography of pipe sections by discrete tomography and total variation minimization”) is available at University of Helsinki library internet archive (pdf). (2018).
- I am still quite proud of my B.Sc. thesis on simulated annealing (2016).
- PDF link (in Finnish).
- Thesis LaTeX source: https://github.com/aa-m-sa/kandi-tutkielma
- MATLAB (Octave) / Python code for numerical experiments: https://github.com/aa-m-sa/kandi-matlab
- Selected presentations:
- Kurkistus tilastollisesti oppivien koneiden eläintarhaan (2020, in Finnish).
- I had the honor to talk about algorithms and statistical machine learning at the IV installment of the public science outreach seminar “Studia Varjomafia”, organized by Tutka ry (University of Turku Science Fiction Club).
- PDF slides, LaTeX source, computer code.
- Follow up about backpropagation detals (2021, in Finnish)
- Kurkistus tilastollisesti oppivien koneiden eläintarhaan (2020, in Finnish).
- Other school projects and works:
- Multilabel classification of Reuters newswires with deep learning, a course project (2017) together with Ioanna Bouri and Ari Siitonen. We used CNNs and Keras.
- Final report PDF slides.
- Project in probabilistic modelling (2017)
- Challenge: Recover unknown directed Bayesian graphical model (ie conditional independence) structure for 26 variables (letters A-Z) with three values (0,1,2), given data of 2500 instances generated by the said unknown network.
- Greedy hill climbing algorithm with BDeu scores worked quite well.
- Source code (Python/C++): https://github.com/aa-m-sa/probmodels-project
- During my undergraduate studies, I built and programmed a Lego Mindstorms robot to play Tic-Tac-Toe against human opponent. It felt very futuristic then. (2014).
- Java source code: https://github.com/aa-m-sa/massive-ironman
- Final report (in Finnish, PDF): https://raw.githubusercontent.com/aa-m-sa/massive-ironman/master/docs/loppuraportti/loppuraportti.pdf
- Pictures or didn’t happen: https://github.com/aa-m-sa/massive-ironman/tree/master/docs/loppuraportti/kuvat
- Videos or didn’t happen: https://drive.google.com/file/d/0B1_61yv8HENsbF9YTGVja3RDVHc/view
- Multilabel classification of Reuters newswires with deep learning, a course project (2017) together with Ioanna Bouri and Ari Siitonen. We used CNNs and Keras.
My other writings, internet presence and hobbies
The contact page above lists some social media channels. In addition to that, I am active with my own name also elsewhere:
- Check out the old blog archived under the kapsi domain: https://aqsalose.kapsi.fi/blog/
- I was quite active at bloggin 2016-2018.
- I am proud of the tag system, because I believe it is useful. Unfortunately eventually it became nightmare to manage because of ill-thought hacks I sued to bootstrap it.
- My Metaculus profile: https://www.metaculus.com/accounts/profile/117552/
- Metaculus is an internet platform where users compete making probabilistic predictions (forecasts) about various questions for glory and internet points. It is an excellent hobby for numerically minded folk. See also their FAQ.
- For wider context, Philip Tetlock has researched the accuracy and and talked about usefulness of making careful, quantified and benchmarked forecasts. I recommend his book Superforecasting (GoodReads, Wikipedia). If you are not into books, here is a discussion with Tetlock on Edge (2012, with Daniel Kahneman, video + transcript) and discussion on J. Horgan’s blog at Scientific American (2015, transcript).
This page is actively updated, previous change on 2021-03-10 (moved some items to Things on the internet page).