Josef Moudřík's Home Page

I try to make useful things that help people. Have a look on what I am doing Right Now!

I am a freelancing ML/AI developer by trade. I am a father, singer, ukulele fan and I play improvisational theatre. I like to go for a run and enjoy a nice game of Go (1-dan).

E-mail: josef@moudrik.cz | CV | LinkedIn | GitHub | Now Page     Old: Blog | Teaching / Výuka

Projects / Stuff

• www.krokyproklima.cz – a Czech-only website with information about steps each one of us can do to fight Global Warming. Let's focus on helping where the impact of our lives is the biggest. Insulate Your home, do not focus on plastic drinking straws!

• EarMuscle Training & ToneFinder – two musical training apps I've been working on recently. Wanna improve your musical improv skills, or be able to hear more?

• AI Horoscope – fun micro app, exploring what it takes to use GPT-3 in production (consistently good results, etc).

• GoStyle – is a project I co-founded with Petr Baudiš. Its aim is to study the possibilities of computer analysis of Go game records, utilising sophisticated machine-learning algorithms. This is interesting for a number of reasons: computer Go is a hard and challenging problem, there exists large databases of games, Go has a lot of tradidional domain knowledge, which can be cleverly utilised, and most importantly, possible applications are numerous.

  To prove the findings of the research, I realised the ideas in publicly available Web App, which helps Go players by pinpointing some issues in their games and giving personalized recommendations (such as professional games to review in order to improve, or Go literature of suitable level). See the Project Website or my Master Thesis to learn more.

• IGGSC 2015 – Second International Go Game Science Conference, which I co-organized.
• Anki decks for music – trying out how spaced repetition works for music.
• Orange hacks – extensions for the Orange data mining toolbox.

Publications

• Moudřík, Neruda: Evolving Non-linear Stacking Ensembles for Prediction of Go Player Attributes. Proceedings of IEEE SSCI, 2015. arXiv pre-print
• Moudřík, Baudiš, Neruda: Evaluating Go Game Records for Prediction of Player Attributes. Proceedings of IEEE CIG, 2015; arXiv pre-print
• Moudřík Josef: Meta-learning methods for analyzing Go playing trends. Master thesis, Charles University in Prague, Faculty of Mathematics and Physics, 2013.
• Moudřík, Baudiš: Evaluating Go Game Records for Prediction of Player Attributes. Presentation given at the European Go Congress Scientific Conference, 2013.
• Baudiš, Moudřík: On Move Pattern Trends in a Large Go Games Corpus, 2012.
• Moudřík Josef: V1 model with a realistic distribution of functionally different neuron classes. Bachelor thesis, Charles University in Prague, Faculty of Mathematics and Physics, 2009.
• Moudřík, Machálek, Antolík, Brom: Výpočetní model primární zrakové kůry: preference složitých tvarů a realistický vznik komplexních buněk. Proceedings of Kognicia a umely zivot IX (Czech/Slovak conference), 2009.