Variational Hyper RNN for Sequence Modeling
This work addresses sequence modeling for data with complex dynamics, but appears incremental as it builds on existing probabilistic and recurrent methods.
The authors tackled the problem of modeling high variability in time series data by proposing a probabilistic sequence model that uses temporal latent variables to dynamically modify decoder and recurrent weights, achieving efficacy on synthetic and real-world data with large variations and regime shifts.
In this work, we propose a novel probabilistic sequence model that excels at capturing high variability in time series data, both across sequences and within an individual sequence. Our method uses temporal latent variables to capture information about the underlying data pattern and dynamically decodes the latent information into modifications of weights of the base decoder and recurrent model. The efficacy of the proposed method is demonstrated on a range of synthetic and real-world sequential data that exhibit large scale variations, regime shifts, and complex dynamics.