This work employs Neural ODEs equipped with a curated collection of equilibrium points to perform classification tasks. The planted attractors serve as indicators for target classes, while the velocity field shapes the dynamical landscape to direct inputs toward their corresponding destinations.
- Neural ODEs are utilized with a curated set of equilibrium points for classification.
- Planted attractors act as indicators for the target classes.
- The velocity field leverages universal approximation capabilities to shape the dynamical landscape.
- This process defines basins of attraction, directing each input initial condition to its destination.