Steven Braun

122 Followers
134 Following
81 Posts
PhD Student working on #tractable #probabilistic modeling at the AIML Lab, TU Darmstadt with Kristian Kersting
Websitehttps://www.steven-braun.com
GitHubhttps://www.github.com/braun-steven
Research Grouphttps://www.aiml.informatik.tu-darmstadt.de

RT @[email protected]

Since it's impossible to search for past tweets and it's that time of the year again - here is our brief "How to ML Paper" guide again: https://docs.google.com/document/d/16R1E2ExKUCP5SlXWHr-KzbVDx9DBUclra-EbU8IB-iE/edit#heading=h.16t67gkeu9dx

Good luck with #ICML2023 and remember there is always another deadline!

🐦🔗: https://twitter.com/j_foerst/status/1617670799552569345

How to ML Paper - A brief Guide

How to ML Paper - A brief Guide Feel free to comment / share and happy paper writing! Also, please see caveats* below. If you like this, why not follow How to ML on Twitter and share the advice/love? Canonical ML Paper Structure Abstract (TL;DR of paper): X: What are we trying to do and why is i...

Google Docs

I'm happy to share that our paper "Bayesian Structure Scores for Probabilistic Circuits" has been accepted at #aistats23! This is joint work with Yang Yang (MSc @mastodon.world, incoming PhD @leuvenai and Gennaro Gala.

The main contribution of the paper is to propose a new (actually old) way to learn the structure of #ProbabilisticCircuits. We take lessons from classical structure learning in Bayesian networks which optimizes some structure score, where a principled choice is a Bayesian score.

This term, my group is teaching a Master-level lecture course on *Numerics of Machine Learning*. Naturally, from the perspective of #probabilisticnumerics. Featuring established results and new insights, the course is taught primarily by the PhD students - the people closest to the cutting edge.

We are making the course public as material for interested students and lecturers.

Today we're releasing Lectures 1 (Intro) and 2-4, which cover linear algebra. Here are links, and what to expect:

I repeatedly get notifications of Elon Musk Twitter posts on my lockscreen even though I’m not even following him and already selected “not interested” at some point.

Is Musk enforcing the visibility of his posts on all Twitter users?

toot toot!
Just getting started here. Start following us for updates.

This leads to famous (in terms of followers etc.) people being recommended with a higher likelihood. E.g. on Twitter, I don't even follow Elon Musk, but I see every single post of his on my timeline. What I'm getting at is: While Mastodon should stick with not recommending posts based on the popularity of the author, it might still be a good idea to recommend, or rather score, posts based on how relevant it is to the user.

6/6

Option 2) is often waived off by Mastodon users, saying that they enjoy the algorithm absence in comparison to Twitter. While I generally agree with this statement, I think this is the perspective on a recommendation algorithm that is based on likes and retweets and the popularity of the author and retweeter (is that even a word?).

5/6

There are two solutions I see here: 1) create a second, community-specific Mastodon account. 2) Have some form of recommendation algorithm.

I think 1) is annoying and too much trouble for the end-user.

4/6

I know that many are using Twitter and Mastodon from a purely academic and professional perspective to stay up-to-date on relevant works and those could be annoyed (or see Mastodon as not well suited for their use-case) when people they follow also post and boost stuff from, to them, unrelated topics.

3/6

If I now start posting and boosting in these communities, since there is no interest-based content recommendation algorithm on Mastodon, people who follow me will see these in their timelines, no matter if they are interested in those particular topics or not.

2/6