RT @innixma: 🚀 AutoGluon 1.2 is here! Key highlights: 1. Tabular & TimeSeries: 70% win-rate over v1.1. 2. New Features: TabPFNMix foundatio…

@innixma Unsolicited excitement! Congrats on a feature full release, happy to have the default AutoML tool to recommend in almost every situation <3

The things a random `-inf` will drive a man to... Anyways, guess it's Python and we can just overwrite this method anyways... :shrug: Bonus fun:

We all know Wikipedia is the golden source of truth, so what do they say? Seems either (low, high) or [low, high] is fine, but the pdf function they give lands on [low, high]

@willmcgugan Just did this today! I had a function operating doing some matrix calculations which I knew could be simplified but not how, asked claude to rewrite it (it was incorrect) and provide test cases to assert equivalent behaviour and then twiddled Claude's answer until it was correct!

RT @neeratyoy: In #Vienna for #ICML2024 to present our latest work! With @_herilalaina_, @StAdriaensen, @karibbov, @Edberg_Wardman, and @Fr…

RT @julien_siems: Excited to announce the 1st Workshop on In-Context Learning (ICL) at ICML. #ICML2024 #ICL

@willmcgugan @deprecated Can Report that the open source pyright does the same, it's been great and @deprecated has become my default

@bdivsalar Heyo, unfortunately I'm not really a knowledgeable maintainer of the repo or it's internals. I just added some CI so that one our dependant libraries could install from pip without issue. I'd recommend the GitHub issues and possibly bumping it if you got no response!

RT @CHvarfner: #ICLR2024 Spotlight! Happy to announce that our paper, "A General Framework for User-Guided Bayesian Optimization", has bee…

RT @LuisOala: starting the year inspired with great people in a great place @dagstuhl we made good progress of bringing infrastructure an…

@hbouammar @andrewgwils Might depend on interpretation but you could view an ensemble as a bayesian model. SMAC is just an ensemble of decision trees at the end of the day it can be fairly effective. I'm not too picky on nomenclature though and there's likely reasons it's not "bayesian" in a full sense

Note to anyone attending a conference: update your CV!!! Every event requires a CV submission?

@willmcgugan Looks like 2014 was a good year for weekends

Not to say class hierarchies are objectively good or bad, it depends. Other approaches you could consider: * A Union of types (classes small, you don't plan to extend) * Parameterized methods (if only small functional changes) * Dependancy Injection (great for algorithmic code)