Customizing Sequence Generation with Multi-Task Dynamical Systems
This work addresses a limitation in sequence generation for real-world applications, offering customization capabilities like style transfer and interpolation, but it appears incremental as it builds on existing dynamical system frameworks.
The paper tackles the problem of dynamical system models lacking adaptability to context by introducing hierarchical multi-task dynamical systems (MTDSs) that enable user control over sequence generation through a latent code, resulting in improved predictions and avoidance of long-term performance degradation.
Dynamical system models (including RNNs) often lack the ability to adapt the sequence generation or prediction to a given context, limiting their real-world application. In this paper we show that hierarchical multi-task dynamical systems (MTDSs) provide direct user control over sequence generation, via use of a latent code $\mathbf{z}$ that specifies the customization to the individual data sequence. This enables style transfer, interpolation and morphing within generated sequences. We show the MTDS can improve predictions via latent code interpolation, and avoid the long-term performance degradation of standard RNN approaches.