An algorithm for discovering Lagrangians automatically from data

@article{Hills2015AnAF,
  title={An algorithm for discovering Lagrangians automatically from data},
  author={Daniel J. A. Hills and Adrian M. Gr{\"u}tter and Jonathan J. Hudson},
  journal={PeerJ Comput. Sci.},
  year={2015},
  volume={1},
  pages={e31}
}
An activity fundamental to science is building mathematical models. These models are used to both predict the results of future experiments and gain insight into the structure of the system under study. We present an algorithm that automates the model building process in a scientifically principled way. The algorithm can take observed trajectories from a wide variety of mechanical systems and, without any other prior knowledge or tuning of parameters, predict the future evolution of the system… 

Figures from this paper

Sparse Identification of Lagrangian for Nonlinear Dynamical Systems via Proximal Gradient Method

— Distilling physical laws autonomously from data has been of great interest in many scientific areas. The sparse identification of nonlinear dynamics (SINDy) and its variations have been developed to

Bayesian Identification of Hamiltonian Dynamics from Symplectic Data

TLDR
This approach uses an approximate marginal Markov Chain Monte Carlo algorithm to directly discover a system Hamiltonian from data and demonstrates that the resulting learning problem yields dynamical systems that have improved accuracy and reduced predictive uncertainty compared to existing state-of-the-art approaches.

Control-Oriented Learning of Lagrangian and Hamiltonian Systems

TLDR
A method based on quadratic programming for learning control-oriented models of physical systems for which limited data from only one trajectory is available is proposed, taking advantage of the principle of least action from physics to approximate either the Lagrangian or the Hamiltonian of the system from the data.

Deriving equations from sensor data using dimensional function synthesis

TLDR
A new method for deriving functions that model the relationship between multiple signals in a physical system, which applies to data streams where the dimensions of the signals are known and outperforms traditional regression and neural network models in inference accuracy.

Control-Oriented Learning on the Fly

TLDR
An algorithm is proposed that uses small perturbations in the control effort to learn system dynamics around the current system state, while ensuring that the system moves in a nearly optimal direction, and provide bounds for its suboptimality.

Optimal Control with Learning on the Fly: System with Unknown Drift

This paper derives an optimal control strategy for a simple stochastic dynamical system with constant drift and an additive control input. Motivated by the example of a physical system with an

Deriving Equations from Sensor Data Using Dimensional Function Synthesis

TLDR
Dimensional function synthesis implements a new method for deriving functions that model the relationship between multiple signals in a physical system and can generate models in less than 300 ms on average across all the physical systems evaluated.

Visual Physics: Discovering Physical Laws from Videos

TLDR
This paper teaches a machine to discover the laws of physics from video streams with no prior knowledge of physics, beyond a temporal stream of bounding boxes, and evaluates its ability to discover physical laws on videos of elementary physical phenomena.

Learning models of quantum systems from experiments

TLDR
This work proposes and demonstrates an approach to retrieving a Hamiltonian model from experiments, using unsupervised machine learning, and can gain further insight on the physics of quantum systems.

References

SHOWING 1-10 OF 45 REFERENCES

Discovering Dynamics

Distilling Free-Form Natural Laws from Experimental Data

TLDR
This work proposes a principle for the identification of nontriviality, and demonstrated this approach by automatically searching motion-tracking data captured from various physical systems, ranging from simple harmonic oscillators to chaotic double-pendula, and discovered Hamiltonians, Lagrangians, and other laws of geometric and momentum conservation.

Perspectives on system identification

  • L. Ljung
  • Computer Science
    Annu. Rev. Control.
  • 2010

Comment on the article "Distilling free-form natural laws from experimental data"

TLDR
This work demystifies how they are able to find Hamiltonians and special classes of Lagrangians from data, and implicitly incorporates Hamilton's equations of motions and Newton's second law.

Rediscovering Physics with BACON.3

TLDR
BACON.3 has shown its generality by rediscovering versions of the Ideal gas law, Kepler's third law, Coulomb's law, Ohm'slaw, and Galileo's laws for the pendulum and constant acceleration.

The use of a genetic algorithm to optimize the functional form of a multi-dimensional polynomial fit to experimental data

TLDR
The genetic algorithm method is compared with a particular genetic programming approach and it is shown that equally good results can be achieved using this simpler technique.

Genetic Programming Theory and Practice V (Genetic and Evolutionary Computation)

TLDR
The text explores the synergy between theory and practice, producing a comprehensive view of the state of the art in GP application.

Evolutionary Computation

  • A. Freitas
  • Computer Science
    Encyclopedia of Machine Learning
  • 2010
TLDR
This chapter addresses the integration of knowledge discovery in databases (KDD) and evolutionary algorithms (EAs) and suggests that this principle should be followed in other EA applications.