Abstract
A universal rule-based self-learning approach using deep reinforcement learning (DRL) is proposed for the first time to solve nonlinear ordinary differential equations and partial differential equations. The solver consists of a deep neural network-structured actor that outputs candidate solutions, and a critic derived only from physical rules (governing equations and boundary and initial conditions). Solutions in discretized time are treated as multiple tasks sharing the same governing equation, and the current step parameters provide an ideal initialization for the next owing to the temporal continuity of the solutions, which shows a transfer learning characteristic and indicates that the DRL solver has captured the intrinsic nature of the equation. The approach is verified through solving the Schrödinger, Navier–Stokes, Burgers’, Van der Pol, and Lorenz equations and an equation of motion. The results indicate that the approach gives solutions with high accuracy, and the solution process promises to get faster.
Similar content being viewed by others
Abbreviations
- DRL:
-
Deep reinforcement learning
- ODE:
-
Ordinary differential equation
- PDE:
-
Partial differential equation
- AGI:
-
Artificial general intelligence
- DNN:
-
Deep neural network
- MDP:
-
Markov decision process
- SDOF:
-
Single degree of freedom
- MDOF:
-
Multi-degree of freedom
- \( {\mathbf{x}} \) :
-
Spatial coordinate
- \( t \) :
-
Time coordinate
- \( u \) :
-
Solution of the ODE or PDE
- \( u_{t} \) :
-
Derivative of \( u \) with respect to time
- \( \vartheta \) :
-
Parameters of ODE or PDE
- \( s \) :
-
State
- \( {{\boldsymbol{\mathcal{S}}}} \) :
-
Domain of the state
- \( a \) :
-
Action
- \( {{\boldsymbol{\mathcal{A}}}} \) :
-
Domain of the action
- \( \theta \) :
-
Trainable parameters of the policy network
- \( \pi_{\theta } \left( {a|s} \right) \) :
-
Probabilistic policy of the action given the state
- \( \mu_{\theta } \) :
-
Mean value of the action determined by the probabilistic policy of the action given the state
- \( \sigma_{\theta } \) :
-
Standard deviation of the action determined by the probabilistic policy of the action given the state
- \( r\left( {s,a} \right) \) :
-
Imbalance of an ODE or PDE
- \( r_{Eq} \) :
-
Imbalance of governing equation
- \( r_{B} \) :
-
Imbalance of boundary conditions
- \( r_{I} \) :
-
Imbalance of initial conditions
- \( J \) :
-
Loss function of policy network
- \( {\rm E}_{{\pi_{\theta } }} \) :
-
Expectation calculated on probabilistic policy
- \( \hat{u} \) :
-
Candidate solution sampled from policy
- \( N \) :
-
Batch size of sampling states
References
Cole JD (1951) On a quasi-linear parabolic equation occurring in aerodynamics. Q Appl Math 9(3):225–236
Soliman AA (2006) The modified extended tanh-function method for solving Burgers-type equations. Physica A 361(2):394–404
Feit M, Fleck J Jr, Steiger A (1982) Solution of the Schrödinger equation by a spectral method. J Comput Phys 47(3):412–433
Wang J X, Kurth-Nelson Z, Tirumala D, Soyer H, Leibo J Z, Munos R, Blundell C, Kumaran D, Botvinick M (2016) Learning to reinforcement learn. arXiv preprint arXiv:1611.05763
Finn C, Abbeel P, Levine S (2017) Model-agnostic meta-learning for fast adaptation of deep networks. arXiv preprint arXiv:1703.03400
LeCun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521(7553):436–444
Sutton RS, Barto AG (1998) Introduction to reinforcement learning, vol 135. MIT press, Cambridge
Raissi M, Yazdani A, Karniadakis G E (2018) Hidden fluid mechanics: a Navier–Stokes informed deep learning framework for assimilating flow visualization data. arXiv preprint arXiv:1808.04327
Raissi M, Wang Z, Triantafyllou MS, Karniadakis GE (2019) Deep learning of vortex-induced vibrations. J Fluid Mech 861:119–137
Lagaris IE, Likas A, Fotiadis DI (1998) Artificial neural networks for solving ordinary and partial differential equations. IEEE Trans Neural Netw 9(5):987–1000
Raissi M, Perdikaris P, Karniadakis GE (2019) Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys 378:686–707
Han J, Jentzen A, EW (2017) Overcoming the curse of dimensionality: solving high-dimensional partial differential equations using deep learning. arXiv preprint arXiv:1707.02568
Mills K, Spanner M, Tamblyn I (2017) Deep learning and the Schrödinger equation. Phys Rev A 96(4):042113
Han EW, Jentzen A (2017) Deep learning-based numerical methods for high-dimensional parabolic partial differential equations and backward stochastic differential equations. Commun Math Stat 5(4):349–380
Khoo Y, Lu J, Ying L (2017) Solving parametric PDE problems with artificial neural networks. arXiv preprint arXiv:1707.03351
Bostanabad R, Zhang Y, Li X, Kearney T, Brinson LC, Apley DW, Liu WK, Chen W (2018) Computational microstructure characterization and reconstruction: review of the state-of-the-art techniques. Prog Mater Sci 95:1–41
Li X, Yang Z, Brinson L C, Choudhary A, Agrawal A, Chen W (2018). A deep adversarial learning methodology for designing microstructural material systems. Paper presented at the ASME 2018 international design engineering technical conferences and computers and information in engineering conference
Mnih V, Kavukcuoglu K, Silver D, Graves A, Antonoglou I, Wierstra D, Riedmiller M (2013) Playing atari with deep reinforcement learning. arXiv preprint arXiv:1312.5602
Silver D, Huang A, Maddison CJ, Guez A, Sifre L, Van Den Driessche G, Schrittwieser J, Antonoglou I, Panneershelvam V, Lanctot M (2016) Mastering the game of Go with deep neural networks and tree search. Nature 529(7587):484
Wang K, Sun W (2019) Meta-modeling game for deriving theory-consistent, microstructure-based traction–separation laws via deep reinforcement learning. Comput Methods Appl Mech Eng 346:216–241
Wang K, Sun W, Du Q (2019) A cooperative game for automated learning of elasto-plasticity knowledge graphs and models with AI-guided experimentation. arXiv preprint arXiv:1903.04307
Jin X, Cheng P, Chen W-L, Li H (2018) Prediction model of velocity field around circular cylinder over various Reynolds numbers by fusion convolutional neural networks based on pressure on the cylinder. Phys Fluids 30(4):047105
Li S, Laima S, Li H (2018) Data-driven modeling of vortex-induced vibration of a long-span suspension bridge using decision tree learning and support vector regression. J Wind Eng Ind Aerodyn 172:196–211
Silver D, Lever G, Heess N, Degris T, Wierstra D, Riedmiller M (2014) Deterministic policy gradient algorithms. In: ICML
Kingma DP, Ba J (2014) Adam: a method for stochastic optimization. arXiv preprint arXiv:1412.6980
Goodfellow I, Bengio Y, Courville A (2016) Deep learning. MIT press, Cambridge
Bergstra J, Bengio Y (2012) Random search for hyper-parameter optimization. J Mach Learn Res 13:281–305
Snoek J, Larochelle H, Adams R P (2012). Practical bayesian optimization of machine learning algorithms. Paper presented at the advances in neural information processing systems
Guckenheimer J, Holmes P (2013) Nonlinear oscillations, dynamical systems, and bifurcations of vector fields, vol 42. Springer, Belrin
Tsatsos M (2006) Theoretical and Numerical study of the Van der Pol equation. Doctoral desertation, Aristotle University of Thessaloniki, vol 4, p 6
Chopra AK (2007) Dynamics of structures: theory and applications to earthquake engineering. Prentice-Hall, Upper Saddle River
Hopf E (1950) The partial differential equation ut + uux = μxx. Commun Pure Appl Math 3(3):201–230
Skeel RD, Berzins M (1990) A method for the spatial discretization of parabolic equations in one space variable. SIAM J Sci Stat Comput 11(1):1–32
Schrödinger E (1926) An undulatory theory of the mechanics of atoms and molecules. Phys Rev 28(6):1049
Cox SM, Matthews PC (2002) Exponential time differencing for stiff systems. J Comput Phys 176(2):430–455
Driscoll TA, Hale N, Trefethen LN (2014) Chebfun guide. Pafnuty Publications, Oxford
Bessa M, Bostanabad R, Liu Z, Hu A, Apley DW, Brinson C, Chen W, Liu WK (2017) A framework for data-driven analysis of materials under uncertainty: countering the curse of dimensionality. Comput Methods Appl Mech Eng 320:633–667
Acknowledgements
This research was funded by the National Natural Sciences Foundation of China (NSFC) (Grant Nos. U1711265 and 51638007).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Wei, S., Jin, X. & Li, H. General solutions for nonlinear differential equations: a rule-based self-learning approach using deep reinforcement learning. Comput Mech 64, 1361–1374 (2019). https://doi.org/10.1007/s00466-019-01715-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00466-019-01715-1