Abstract:When solving Hamiltonian systems using numerical integrators, preserving the symplectic structure may be crucial for many problems. At the same time, solving chaotic or stiff problems requires integrators to approximate the trajectories with extreme precision. So, integrating Hamilton's equations to a level of scientific reliability such that the answer can be used for scientific interpretation, may be computationally expensive. However, a neural network can be a viable alternative to numerical integrators, offering high-fidelity solutions orders of magnitudes faster.To understand whether it is also important to preserve the symplecticity when neural networks are used, we analyze three well-known neural network architectures that are including the symplectic structure inside the neural network's topology. Between these neural network architectures many similarities can be found. This allows us to formulate a new, generalized framework for these architectures. In the generalized framework Symplectic Recurrent Neural Networks, SympNets and HénonNets are included as special cases. Additionally, this new framework enables us to find novel neural network topologies by transitioning between the established ones.We compare new Generalized Hamiltonian Neural Networks (GHNNs) against the already established SympNets, HénonNets and physics-unaware multilayer perceptrons. This comparison is performed with data for a pendulum, a double pendulum and a gravitational 3-body problem. In order to achieve a fair comparison, the hyperparameters of the different neural networks are chosen such that the prediction speeds of all four architectures are the same during inference. A special focus lies on the capability of the neural networks to generalize outside the training data. The GHNNs outperform all other neural network architectures for the problems considered.

The Hamilton Neural-Network Model

Computer simulation investigation on the Hamilton neural network model

Cayley neural network model: to recognize 256-level color patterns

A Quantum Hopfield Neural Network Model and Image Recognition

The three-dimensional rotation neural network

A new class of Hopfield neural network with double memristive synapses and its DSP implementation

A new mechanical approach to handle generalized Hopfield neural networks

Chaos and firing patterns in a discrete fractional Hopfield neural network model

Neural Hamilton: Can A.I. Understand Hamiltonian Mechanics?

Andysis of Discrete Hopfield Neural Network

Cascade Tri-Neuron Hopfield Neural Network: Dynamical Analysis and Analog Circuit Implementation

An FHN-HR Neuron Network Coupled With a Novel Locally Active Memristor and Its DSP Implementation

Stability and Periodicity of Discrete Hopfield Neural Networks with Column Arbitrary-Magnitude-dominant Weight Matrix

Memristive Bi-Neuron Hopfield Neural Network with Coexisting Symmetric Behaviors

Hybrid Impulsive and Switching Hopfield Neural Networks with State-Dependent Impulses

Nonergodic delocalized paramagnetic states in quantum neural networks

A Generalized Framework of Neural Networks for Hamiltonian Systems

Complex Dynamics of a Delayed Discrete Neural Network of Two Nonidentical Neurons.

Symplectic Learning for Hamiltonian Neural Networks

Memristor Synapse-Driven Simplified Hopfield Neural Network: Hidden Dynamics, Attractor Control, and Circuit Implementation

Dynamics analysis and image encryption application of Hopfield neural network with a novel multistable and highly tunable memristor