Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
DecoderHiddenStates -> QuantileForecasts
Map model hidden states to a predefined array of target percentiles using a single joint projection layer.
Problem it solves
Point forecasting does not quantify uncertainty, and training separate models for each individual quantile is computationally prohibitive.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.