Residual Neural Networks for the Prediction of Planetary Collision Outcomes
arXiv:2210.04248 · doi:10.1093/mnras/stac2933
Abstract
Fast and accurate treatment of collisions in the context of modern N-body planet formation simulations remains a challenging task due to inherently complex collision processes. We aim to tackle this problem with machine learning (ML), in particular via residual neural networks. Our model is motivated by the underlying physical processes of the data-generating process and allows for flexible prediction of post-collision states. We demonstrate that our model outperforms commonly used collision handling methods such as perfect inelastic merging and feed-forward neural networks in both prediction accuracy and out-of-distribution generalization. Our model outperforms the current state of the art in 20/24 experiments. We provide a dataset that consists of 10164 Smooth Particle Hydrodynamics (SPH) simulations of pairwise planetary collisions. The dataset is specifically suited for ML research to improve computational aspects for collision treatment and for studying planetary collisions in general. We formulate the ML task as a multi-task regression problem, allowing simple, yet efficient training of ML models for collision treatment in an end-to-end manner. Our models can be easily integrated into existing N-body frameworks and can be used within our chosen parameter space of initial conditions, i.e. where similar-sized collisions during late-stage terrestrial planet formation typically occur.
13 pages, 7 figures, 7 tables
References in corpus (22)
- Array Programming with NumPy
- Water Delivery and Giant Impacts in the 'Grand Tack' Scenario
- Collisional Stripping and Disruption of Super-Earths
- On Neural Differential Equations
- A giant impact as the likely origin of different twins in the Kepler-107 exoplanet system
- Predicting the long-term stability of compact multiplanet systems
- The early instability scenario: terrestrial planet formation during the giant planet instability, and the effect of collisional fragmentation
- A Collisional Origin to Earth's Non-chondritic Composition?
- Habitable Planet Formation in Binary-Planetary Systems
- Collisional stripping of planetary crusts
- Realistic On-the-fly Outcomes of Planetary Collisions II: Bringing Machine Learning to N-body Simulations
- Numerically Predicted Indirect Signatures of Terrestrial Planet Formation
- The Effect of Planets Beyond the Ice Line on the Accretion of Volatiles by Habitable-Zone Rocky Planets
- Transfer, loss and physical processing of water in hit-and-run collisions of planetary embryos
- Realistic collisional water transport during terrestrial planet formation: Self-consistent modeling by an N-body--SPH hybrid code
- The Effect of Inefficient Accretion on Planetary Differentiation
- Deep learning for surrogate modelling of 2D mantle convection
- Fragmentation of colliding planetesimals with water content
- Building Terrestrial Planets: Why results of perfect-merging simulations are not quantitatively reliable approximations to accurate modeling of terrestrial planet formation
- Protoplanet Collisions: Statistical Properties of Ejecta
- Water delivery to dry protoplanets by hit-and-run collisions
- On the formation of terrestrial planets between two massive planets: The case of 55 Cancri