afb0b97df87090596ae7c503f60bb23f-AuthorFeedback.pdf
–Neural Information Processing Systems
The approach of Linderman 2017 is unfortunately not14 suitable for optimization with SGD since the Polya-Gamma distribution can not bereparametrised. Incontrast,SSMsmaintainuncertainty46 information and provide a principled way to ignore missing or anomalous data, i.e. by omitting the Bayes update47 (Kalman updatestep).
Neural Information Processing Systems
Feb-9-2026, 20:46:04 GMT
- Technology:
- Information Technology (0.75)