The #IDA2023 runner-up Frontier Prize was awarded to the paper "LEMON: Alternative Sampling for More Faithful Explanation through Local Surrogate Models" by @imDecode, @pratik_gajane, Joost Jorritsma, Jarke van Wijk, and @pechenizkiy, of @TUeindhoven. Congratulations!
De explosief toenemende populariteit van zelflerende computermodellen kent ook een gevaar, als ze mee mogen beslissen over of iemand een lening krijgt of wanneer ze voorspellingen moeten doen over iemands gezondheid. https://t.co/V6Q41PPiB2
Our recent work (with @imDecode@DaphneMiedema@pechenizkiy) might shed some light on this "disagreement problem" - at least for feature importance techniques like SHAP and LIME: https://t.co/wCns8pzURc
SHAP and LIME are often put under the same umbrella 1/n
Zelflerende computermodellen kunnen waardevol zijn voor bijvoorbeeld spraakherkenning, fraudedetectie en het inschatten van medische risico’s. Om data-experts daarbij te helpen, ontwikkelde promovendus Dennis Collaris interactieve visualisatietools. https://t.co/oP3JvszY6O
"Visualization for Explainable AI", great talk by @imDecode from @TUeindhoven at @NEC Labs Europe. To get an idea of the concepts discussed at thetalk and interactively play with them checkout https://t.co/YXgruEFsWn Discussion in the thread👇 1/5 #dataviz#ml#ai#xai
TIL about sklearn-pmml-model, a python library that allows you to export PMML files to @scikit_learn models. Looks like it could be *very* useful to increase reproducibility of research results, especially across different programming languages!
https://t.co/EF7YxWtjgf
@kyriakidestella Hi Stella, congratulations on your new job! Can you please legalize Canadian Intense? It’s the best recipe for less cancer in the EU. Walther from @rivm already has the machine. Thanks! #saverickbrandsteder#zondagmetlubach