Skip to content

Comment on TimesFM: Time Series Foundation Model for time-series forecastingparent

Comments

Even if it was for multivariate time series, the model would first need to infer what machine are we talking about, then its working conditions, and only then make a reasonable forecast based on an hypothesis of its dynamics. I don’t know, seems pretty hard.

Indeed. An issue I ran into over and over while doing research for semiconductor manufacturing.

My complaint was more illustrative than earnest.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.