Publikation
Bi-Level One-Shot Architecture Search for Probabilistic Time Series Forecasting
Jonas Seng; Fabian Kalter; Zhongjie Yu; Fabrizio Ventola; Kristian Kersting
In: Katharina Eggensperger; Roman Garnett; Joaquin Vanschoren; Marius Lindauer; Jacob R. Gardner (Hrsg.). International Conference on Automated Machine Learning, 9-12 September 2024, Sorbonne Université, Paris, France. International Conference on Machine Learning (ICML), Pages 10/1-20, Proceedings of Machine Learning Research, Vol. 256, PMLR, 2024.
Zusammenfassung
Time series forecasting is ubiquitous in many disciplines. A recent hybrid architecture
named predictive Whittle networks (PWNs) tackles this task by employing two distinct
modules, a tractable probabilistic model and a neural forecaster, with the former guiding the
latter by providing likelihoods about predictions during training. Although PWNs achieve
state-of-the-art accuracy, finding the optimal type of probabilistic model and neural fore-
caster (macro-architecture search) and the architecture of each module (micro-architecture
search) of such hybrid models remains difficult and time-consuming. Current one-shot
neural architecture search (NAS) methods approach this challenge by focusing on either
the micro or the macro aspect, overlooking mutual impact, and could attain the overall
optimization only sequentially. To overcome these limitations, we introduce a bi-level one-
shot NAS method that optimizes such hybrid architectures simultaneously, leveraging the
relationships between the micro and the macro architectural levels. We empirically demon-
strate that the hybrid architectures found by our method outperform human-designed and
overparameterized ones on various challenging datasets. Furthermore, we unveil insights
into underlying connections between architectural choices and temporal features.
