Encoding Temporal Statistical-space Priors via Augmented Representation. (arXiv:2401.16808v1 [cs.LG])

Modeling time series data remains a pervasive issue as the temporal dimension
is inherent to numerous domains. Despite significant strides in time series
forecasting, high noise-to-signal ratio, non-normality, non-stationarity, and
lack of data continue challenging practitioners. In response, we leverage a
simple representation augmentation technique to overcome these challenges. Our
augmented representation acts as a statistical-space prior encoded at each time
step. In response, we name our method Statistical-space Augmented
Representation (SSAR). The underlying high-dimensional data-generating process
inspires our representation …

arxiv augmentation challenges cs.lg data domains encoding forecasting issue modeling noise normality representation series signal simple space statistical temporal time series time series forecasting via