<p align="center">
<img src="figs/Neuromancer.png" width="250">
</p>
# NeuroMANCER v1.5.2
[](https://pypi.org/project/neuromancer)
[](https://github.com/pnnl/neuromancer/blob/master/LICENSE.md)
[](https://pnnl.github.io/neuromancer/)
**Neural Modules with Adaptive Nonlinear Constraints and Efficient Regularizations (NeuroMANCER)**
is an open-source differentiable programming (DP) library for solving parametric constrained optimization problems,
physics-informed system identification, and parametric model-based optimal control.
NeuroMANCER is written in [PyTorch](https://pytorch.org/) and allows for systematic
integration of machine learning with scientific computing for creating end-to-end
differentiable models and algorithms embedded with prior knowledge and physics.
---
## Table of Contents
1. [Overview](#overview)
2. [Key Features](#key-features)
3. [What's New in v1.5.2](#whats-new)
4. [Getting Started](#getting-started)
5. [Tutorials](#domain-examples)
5. [Installation](#installation)
6. [Documentation and User Guides](#documentation-and-user-guides)
---
### Key Features
* **Learn To Model, Learn To Control, Learn To Optimize**: Our library is built to provide end users a multitude of tools to solve Learning To Optimize (L2O), Learning To Model (L2M), and Learning To Control (L2C) tasks. Tackle advanced constrained parametric optimization, model fluid dynamics using physics-informed neural networks, or learn how to control indoor air temperature in buildings to maximize building efficiency.
* **Symbolic programming** interface makes it very easy to define embed prior knowledge of physics, domain knowledge, and constraints into those learning paradigms.
* **Comprehensive Learning Tools**: Access a wide array of tutorials and example applications—from basic system identification to advanced predictive control—making it easy for users to learn and apply NeuroMANCER to real-world problems.
* **State-of-the-art methods**: NeuroMANCER is up-to-date with SOTA methods such as Kolgomorov-Arnold Networks (KANs) for function approximation, neural ordinary differential equations (NODEs) and sparse identification of non-linear dynamics (SINDy) for learning to model dynamical systems, and differentiable convex optimization layers for safety constraints in learning to optimize and learning to control.
## What's New in v1.5.2
### Load Forecasting Capabilities and Transformers
We expand our energy systems domain examples with load forecasting for buildings.
We showcase the use of time-series modelling and forecasting using the Short-Term Electricity Load Forecasting (Panama case study) dataset. We demonstrate forecasting capaibilities using a Transformer model, a new block added to our (neural) blocks.py, as well as other standard blocks. We also utilize historical weather data to assist in energy forecasting.
### Finite-Basis Kolgomorov Arnold Networks
Kolmogorov-Arnold networks (KANs) have attracted attention recently as an alternative to multilayer
perceptrons (MLPs) for scientific machine learning. However, KANs can be expensive to train, even
for relatively small networks. We have implemented an FBKAN block (FBPINNs),
for domain decomposition method for KANs that allows for several small
KANs to be trained in parallel to give accurate solutions for multiscale problems.
**New Colab Examples:**
> ⭐ [Load Forecasting](#energy-systems)
> ⭐ [Function Approximation with Kolgomorov-Arnold Networks ](#function-approximation)
## Getting Started
```
pip install neuromancer
```
Extensive set of tutorials can be found in the
[examples](https://github.com/pnnl/neuromancer/tree/master/examples) folder and the [Tutorials](#domain-examples) below.
Interactive notebook versions of examples are available on Google Colab!
Test out NeuroMANCER functionality before cloning the repository and setting up an
environment.
The notebooks below introduce the core abstractions of the NeuroMANCER library, in particular our symbolic programming interface and Node classes.
### Symbolic Variables, Nodes, Constraints, Objectives, and Systems Classes
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/tutorials/part_1_linear_regression.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 1: Linear regression in PyTorch vs NeuroMANCER.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/tutorials/part_2_variable.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 2: NeuroMANCER syntax tutorial: variables, constraints, and objectives.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/tutorials/part_3_node.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 3: NeuroMANCER syntax tutorial: modules, Node, and System class.
### PyTorch Lightning Integration
We have integrated PyTorch Lightning to streamline code, enable custom training logic, support GPU and multi-GPU setups, and handle large-scale, memory-intensive learning tasks.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/lightning_integration_examples/Part_1_lightning_basics_tutorial.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 1: Lightning Integration Basics.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/lightning_integration_examples/Part_2_lightning_advanced_and_gpu_tutorial.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 2: Lightning Advanced Features and Automatic GPU Support.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/lightning_integration_examples/other_examples/lightning_custom_training_example.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 4: Defining Custom Training Logic via Lightning Modularized Code.
### Example
Quick example for how to solve parametric constrained optimization problem using NeuroMANCER, leveraging our symbolic programming interface, Node and Variable, Blocks, SLiM library, and PenaltyLoss classes.
```python
# Neuromancer syntax example for constrained optimization
import neuromancer as nm
import torch
# define neural architecture
func = nm.modules.blocks.MLP(insize=1, outsize=2,
linear_map=nm.slim.maps['linear'],
nonlin=torch.nn.ReLU, hsizes=[80] * 4)
# wrap neural net into symbolic representation via the Node class: map(p) -> x
map = nm.system.Node(func, ['p'], ['x'], name='map')
# define decision variables
x = nm.constraint.variable("x")[:, [0]]
y = nm.constraint.variable("x")[:, [1]]
# problem parameters sampled in the dataset
p = nm.constraint.variable('p')
# define objective function
f = (1-x)**2 + (y-x**2)**2
obj = f.minimize(weight=1.0)
# define constraints
con_1 = 100.*(x >= y)
con_2 = 100.*(x**2+y**2 <= p**2)
# create penalty method-based loss function
loss = nm.loss.PenaltyLoss(objectives=[obj], constraints=[con_1, con_2])
# construct differentiable constrained optimization problem
problem = nm.problem.Problem(nodes=[map], loss=loss)
```
## Domain Examples
NeuroMANCER is built to tackle a variety of domain-specific modeling and control problems using its array of methods. Here we show how to model and control building energy systems, as well as apply load forecasting techniques.
For more in-depth coverage of our methods, please see our general [Tutorials](#tutorials-on-methods-for-modeling-optimization-and-control) section below.
### Energy Systems
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/NODE_building_dynamics.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Learning Building Thermal Dynamics using Neural ODEs
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/NODE_RC_networks.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Multi-zone Building Thermal Dynamics Resistance-Capacitance network with Neural ODEs
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/NODE_swing_equation.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Learning Swing Equation Dynamics using Neural ODEs
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/DPC_building_control.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Learning to Control Indoor Air Temperature in Buildings
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/HVAC_load_forecasting.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Energy Load Forecasting for the Air Handling System of an Office Building with MLP and CNN models
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/building_load_forecasting_Transformers.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Energy Load Forecasting for Building with Transformers Model
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/DPC_PSH.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Learning to Control Pumped-storage Hyrdoelectricity System
## Tutorials on Methods for Modeling, Optimization, and Control
### Learning to Optimize (L2O) Parametric Programming
Neuromancer allows you to formulate and solve a broad class of parametric optimization problems leveraging machine learning to learn the solutions to such problems. [More information on Parametric programming](https://github.com/pnnl/neuromancer/tree/develop/examples/parametric_programming)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_1_basics.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 1: Learning to solve a constrained optimization problem.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_2_pQP.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 2: Learning to solve a quadratically-constrained optimization problem.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_3_pNLP.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 3: Learning to solve a set of 2D constrained optimization problems.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_4_projectedGradient.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 4: Learning to solve a constrained optimization problem with the projected gradient.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_5_cvxpy_layers.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 5: Using Cvxpylayers for differentiable projection onto the polytopic feasible set.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_6_pQp_lopoCorrection.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a>
Part 6: Learning to optimize with metric learning for Operator Splitting layers.
### Learning to Control (L2C)
Neuromancer allows you to learn control policies for full spectrum of white/grey/black-box dynamical systems, subject to choice constraints and objective functions.
[More information on Differential Predictive Control](https://github.com/pnnl/neuromancer/tree/develop/examples/control)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_1_stabilize_linear_system.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 1: Learning to stabilize a linear dynamical system.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_2_stabilize_ODE.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 2: Learning to stabilize a nonlinear differential equation.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_3_ref_tracking_ODE.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 3: Learning to control a nonlinear differential equation.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_4_NODE_control.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 4: Learning neural ODE model and control policy for an unknown dynamical system.
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_5_neural_Lyapunov.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 5: Learning neural Lyapunov function for a nonlinear dynamical system.
### Function Approximation
Neuromancer is up-to-date with state-of-the-art methods. Here we showcase the powerful Kolgomorov-Arnold networks [More information on Kolgomorov-Arnold Networks](https://github.com/pnnl/neuromancer/tree/develop/examples/KANs)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/feature/fbkans/examples/KANs/p1_fbkan_vs_kan_noise_data_1d.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 1: A comparison of KANs and FBKANs in learning a 1D multiscale function with noise
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/feature/fbkans/examples/KANs/p2_fbkan_vs_kan_noise_data_2d.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 2: A comparison of KANs and FBKANs in learning a 2D multiscale function with noise
### Neural Operators
Neuromancer allows one to use machine learning, prior physics and domain knowledge, to construct mathematical and differentiabl models of dynamical systems given the measured observations of the system behavior.
[More information on System ID via Neural State Space Models and ODEs](https://github.com/pnnl/neuromancer/tree/develop/examples/ODEs)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_1_NODE.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 1: Neural Ordinary Differential Equations (NODEs)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_2_param_estim_ODE.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 2: Parameter estimation of ODE system
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_3_UDE.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 3: Universal Differential Equations (UDEs)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_4_nonauto_NODE.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 4: NODEs with exogenous inputs
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_5_nonauto_NSSM.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 5: Neural State Space Models (NSSMs) with exogenous inputs
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_6_NetworkODE.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 6: Data-driven modeling of resistance-capacitance (RC) network ODEs
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_7_DeepKoopman.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 7: Deep Koopman operator
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_8_nonauto_DeepKoopman.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 8: control-oriented Deep Koopman operator
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_9_SINDy.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 9: Sparse Identification of Nonlinear Dynamics (SINDy)
### Physics-Informed Neural Networks (PINNs)
Neuromancer's symbolic programming design is perfectly suited for solving PINNs. [More information on PINNs](https://github.com/pnnl/neuromancer/tree/develop/examples/PDEs)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_1_PINN_DiffusionEquation.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 1: Diffusion Equation
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_2_PINN_BurgersEquation.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 2: Burgers' Equation
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_3_PINN_BurgersEquation_inverse.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 3: Burgers' Equation w/ Parameter Estimation (Inverse Problem)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_4_PINN_LaplaceEquationSteadyState.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 4: Laplace's Equation (steady-state)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_5_Pendulum_Stacked.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 5: Damped Pendulum (stacked PINN)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_6_PINN_NavierStokesCavitySteady_KAN.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> Part 6: Navier-Stokes equation (lid-driven cavity flow, steady-state, KAN)
### Stochastic Differential Equations (SDEs)
Neuromancer has been integrated with TorchSDE to handle stochastic dynamical systems. [More information on SDEs](https://github.com/pnnl/neuromancer/tree/develop/examples/SDEs)
+ <a target="_blank" href="https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/SDEs/sde_walkthrough.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/></a> LatentSDEs: "System Identification" of Stochastic Processes using Neuromancer x TorchSDE
## Documentation and User Guides
The documentation for the library can be found [online](https://pnnl.github.io/neuromancer/).
There is also an [introduction video](https://www.youtube.com/watch?v=YkFKz-DgC98) covering
core features of the library.
For more information, including those for developers, please go to our [Developer and User Guide](USER_GUIDE.md)
## Installation
Simply run
```
pip install neuromancer
```
For manual installation please refer to [Installation Instructions](INSTALLATION.md)
## Community Information
We welcome contributions and feedback from the open-source community!
### Contributions, Discussions, and Issues
Please read the [Community Development Guidelines](https://github.com/pnnl/neuromancer/blob/master/CONTRIBUTING.md)
for further information on contributions, [discussions](https://github.com/pnnl/neuromancer/discussions), and [Issues](https://github.com/pnnl/neuromancer/issues).
### Release notes
See the [Release notes](https://github.com/pnnl/neuromancer/blob/master/RELEASE_NOTES.md) documenting new features.
### License
NeuroMANCER comes with [BSD license](https://en.wikipedia.org/wiki/BSD_licenses).
See the [license](https://github.com/pnnl/neuromancer/blob/master/LICENSE.md) for further details.
## Publications
+ [Jan Drgona, Aaron Tuor, Draguna Vrabie, Learning Constrained Parametric Differentiable Predictive Control Policies With Guarantees, IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2024](https://ieeexplore.ieee.org/abstract/document/10479163)
+ [Renukanandan Tumu, Wenceslao Shaw Cortez, Ján Drgoňa, Draguna L. Vrabie, Sonja Glavaski, Differentiable Predictive Control for Large-Scale Urban Road Networks, arXiv:2406.10433, 2024](https://arxiv.org/abs/2406.10433)
+ [Ethan King, James Kotary, Ferdinando Fioretto, Jan Drgona, Metric Learning to Accelerate Convergence of Operator Splitting Methods for Differentiable Parametric Programming, arXiv:2404.00882, 2024](https://arxiv.org/abs/2404.00882)
+ [James Koch, Madelyn Shapiro, Himanshu Sharma, Draguna Vrabie, Jan Drgona, Neural Differential Algebraic Equations, arXiv:2403.12938, 2024](https://arxiv.org/abs/2403.12938)
+ [Wenceslao Shaw Cortez, Jan Drgona, Draguna Vrabie, Mahantesh Halappanavar, A Robust, Efficient Predictive Safety Filter, arXiv:2311.08496, 2024](https://arxiv.org/abs/2311.08496)
+ [Shrirang Abhyankar, Jan Drgona, Andrew August, Elliott Skomski, Aaron Tuor, Neuro-physical dynamic load modeling using differentiable parametric optimization, 2023 IEEE Power & Energy Society General Meeting (PESGM), 2023](https://ieeexplore.ieee.org/abstract/document/10253098)
+ [James Koch, Zhao Chen, Aaron Tuor, Jan Drgona, Draguna Vrabie, Structural Inference of Networked Dynamical Systems with Universal Differential Equations, arXiv:2207.04962, (2022)](https://aps.arxiv.org/abs/2207.04962)
+ [Ján Drgoňa, Sayak Mukherjee, Aaron Tuor, Mahantesh Halappanavar, Draguna Vrabie, Learning Stochastic Parametric Differentiable Predictive Control Policies, IFAC ROCOND conference (2022)](https://www.sciencedirect.com/science/article/pii/S2405896322015877)
+ [Sayak Mukherjee, Ján Drgoňa, Aaron Tuor, Mahantesh Halappanavar, Draguna Vrabie, Neural Lyapunov Differentiable Predictive Control, IEEE Conference on Decision and Control Conference 2022](https://arxiv.org/abs/2205.10728)
+ [Wenceslao Shaw Cortez, Jan Drgona, Aaron Tuor, Mahantesh Halappanavar, Draguna Vrabie, Differentiable Predictive Control with Safety Guarantees: A Control Barrier Function Approach, IEEE Conference on Decision and Control Conference 2022](https://arxiv.org/abs/2208.02319)
+ [Ethan King, Jan Drgona, Aaron Tuor, Shrirang Abhyankar, Craig Bakker, Arnab Bhattacharya, Draguna Vrabie, Koopman-based Differentiable Predictive Control for the Dynamics-Aware Economic Dispatch Problem, 2022 American Control Conference (ACC)](https://ieeexplore.ieee.org/document/9867379)
+ [Drgoňa, J., Tuor, A. R., Chandan, V., & Vrabie, D. L., Physics-constrained deep learning of multi-zone building thermal dynamics. Energy and Buildings, 243, 110992, (2021)](https://www.sciencedirect.com/science/article/pii/S0378778821002760)
+ [E. Skomski, S. Vasisht, C. Wight, A. Tuor, J. Drgoňa and D. Vrabie, "Constrained Block Nonlinear Neural Dynamical Models," 2021 American Control Conference (ACC), 2021, pp. 3993-4000, doi: 10.23919/ACC50511.2021.9482930.](https://ieeexplore.ieee.org/document/9482930)
+ [Skomski, E., Drgoňa, J., & Tuor, A. (2021, May). Automating Discovery of Physics-Informed Neural State Space Models via Learning and Evolution. In Learning for Dynamics and Control (pp. 980-991). PMLR.](https://proceedings.mlr.press/v144/skomski21a.html)
+ [Drgoňa, J., Tuor, A., Skomski, E., Vasisht, S., & Vrabie, D. (2021). Deep Learning Explicit Differentiable Predictive Control Laws for Buildings. IFAC-PapersOnLine, 54(6), 14-19.](https://www.sciencedirect.com/science/article/pii/S2405896321012933)
+ [Tuor, A., Drgona, J., & Vrabie, D. (2020). Constrained neural ordinary differential equations with stability guarantees. arXiv preprint arXiv:2004.10883.](https://arxiv.org/abs/2004.10883)
+ [Drgona, Jan, et al. "Differentiable Predictive Control: An MPC Alternative for Unknown Nonlinear Systems using Constrained Deep Learning." Journal of Process Control Volume 116, August 2022, Pages 80-92](https://www.sciencedirect.com/science/article/pii/S0959152422000981)
+ [Drgona, J., Skomski, E., Vasisht, S., Tuor, A., & Vrabie, D. (2020). Dissipative Deep Neural Dynamical Systems, in IEEE Open Journal of Control Systems, vol. 1, pp. 100-112, 2022](https://ieeexplore.ieee.org/document/9809789)
+ [Drgona, J., Tuor, A., & Vrabie, D., Learning Constrained Adaptive Differentiable Predictive Control Policies With Guarantees, arXiv preprint arXiv:2004.11184, (2020)](https://arxiv.org/abs/2004.11184)
## Cite as
```yaml
@article{Neuromancer2023,
title={{NeuroMANCER: Neural Modules with Adaptive Nonlinear Constraints and Efficient Regularizations}},
author={Drgona, Jan and Tuor, Aaron and Koch, James and Shapiro, Madelyn and Jacob, Bruno and Vrabie, Draguna},
Url= {https://github.com/pnnl/neuromancer},
year={2023}
}
```
## Development team
**Active core developers**: [Jan Drgona](https://drgona.github.io/), [Rahul Birmiwal](https://www.linkedin.com/in/rahul-birmiwal009/), [Bruno Jacob](https://brunopjacob.github.io/)
**Notable contributors**: [Aaron Tuor](https://sw.cs.wwu.edu/~tuora/aarontuor/), Madelyn Shapiro, James Koch, Seth Briney, Bo Tang, Ethan King, Elliot Skomski, Zhao Chen, Christian Møldrup Legaard
**Scientific advisors**: Draguna Vrabie, Panos Stinis
Open-source contributions made by:
<a href="https://github.com/pnnl/neuromancer/graphs/contributors">
<img src="https://contrib.rocks/image?repo=pnnl/neuromancer" />
</a>
Made with [contrib.rocks](https://contrib.rocks).
## Acknowledgments
This research was partially supported by the Mathematics for Artificial Reasoning in Science (MARS) and Data Model Convergence (DMC) initiatives via the Laboratory Directed Research and Development (LDRD) investments at Pacific Northwest National Laboratory (PNNL), by the U.S. Department of Energy, through the Office of Advanced Scientific Computing Research's “Data-Driven Decision Control for Complex Systems (DnC2S)” project, and through the Energy Efficiency and Renewable Energy, Building Technologies Office under the “Dynamic decarbonization through autonomous physics-centric deep learning and optimization of building operations” and the “Advancing Market-Ready Building Energy Management by Cost-Effective Differentiable Predictive Control” projects. This project was also supported from the U.S. Department of Energy, Advanced Scientific Computing Research program, under the Uncertainty Quantification for Multifidelity Operator Learning (MOLUcQ) project (Project No. 81739). PNNL is a multi-program national laboratory operated for the U.S. Department of Energy (DOE) by Battelle Memorial Institute under Contract No. DE-AC05-76RL0-1830.
Raw data
{
"_id": null,
"home_page": null,
"name": "neuromancer",
"maintainer": null,
"docs_url": null,
"requires_python": "<3.11,>=3.9",
"maintainer_email": null,
"keywords": "Deep Learning, Pytorch, Linear Models, Dynamical Systems, Data-driven control",
"author": null,
"author_email": "Aaron Tuor <aaron.tuor@pnnl.gov>, Jan Drgona <jan.drgona@pnnl.gov>, James Koch <james.koch@pnnl.gov>, Madelyn Shapiro <madelyn.shapiro@pnnl.gov>, Rahul Birmiwal <rahul.birmiwal@pnnl.gov>, Draguna Vrabie <draguna.vrabie@pnnl.gov>, Bruno Jacob <bruno.jacbo@pnnl.gov>",
"download_url": "https://files.pythonhosted.org/packages/c9/5d/5eb20f67f8527ae34fbf14eb6840bafc58a8597371a2b6da78260c7d9f6a/neuromancer-1.5.2.tar.gz",
"platform": null,
"description": "\n<p align=\"center\">\n <img src=\"figs/Neuromancer.png\" width=\"250\"> \n</p>\n\n# NeuroMANCER v1.5.2\n\n[](https://pypi.org/project/neuromancer)\n[](https://github.com/pnnl/neuromancer/blob/master/LICENSE.md)\n[](https://pnnl.github.io/neuromancer/)\n\n\n**Neural Modules with Adaptive Nonlinear Constraints and Efficient Regularizations (NeuroMANCER)**\nis an open-source differentiable programming (DP) library for solving parametric constrained optimization problems, \nphysics-informed system identification, and parametric model-based optimal control.\nNeuroMANCER is written in [PyTorch](https://pytorch.org/) and allows for systematic \nintegration of machine learning with scientific computing for creating end-to-end \ndifferentiable models and algorithms embedded with prior knowledge and physics.\n\n---\n\n## Table of Contents\n1. [Overview](#overview)\n2. [Key Features](#key-features)\n3. [What's New in v1.5.2](#whats-new)\n4. [Getting Started](#getting-started)\n5. [Tutorials](#domain-examples)\n5. [Installation](#installation)\n6. [Documentation and User Guides](#documentation-and-user-guides)\n\n\n---\n\n### Key Features\n* **Learn To Model, Learn To Control, Learn To Optimize**: Our library is built to provide end users a multitude of tools to solve Learning To Optimize (L2O), Learning To Model (L2M), and Learning To Control (L2C) tasks. Tackle advanced constrained parametric optimization, model fluid dynamics using physics-informed neural networks, or learn how to control indoor air temperature in buildings to maximize building efficiency.\n* **Symbolic programming** interface makes it very easy to define embed prior knowledge of physics, domain knowledge, and constraints into those learning paradigms. \n* **Comprehensive Learning Tools**: Access a wide array of tutorials and example applications\u2014from basic system identification to advanced predictive control\u2014making it easy for users to learn and apply NeuroMANCER to real-world problems.\n* **State-of-the-art methods**: NeuroMANCER is up-to-date with SOTA methods such as Kolgomorov-Arnold Networks (KANs) for function approximation, neural ordinary differential equations (NODEs) and sparse identification of non-linear dynamics (SINDy) for learning to model dynamical systems, and differentiable convex optimization layers for safety constraints in learning to optimize and learning to control.\n\n\n## What's New in v1.5.2\n### Load Forecasting Capabilities and Transformers\nWe expand our energy systems domain examples with load forecasting for buildings. \nWe showcase the use of time-series modelling and forecasting using the Short-Term Electricity Load Forecasting (Panama case study) dataset. We demonstrate forecasting capaibilities using a Transformer model, a new block added to our (neural) blocks.py, as well as other standard blocks. We also utilize historical weather data to assist in energy forecasting. \n\n\n### Finite-Basis Kolgomorov Arnold Networks\n\nKolmogorov-Arnold networks (KANs) have attracted attention recently as an alternative to multilayer\nperceptrons (MLPs) for scientific machine learning. However, KANs can be expensive to train, even\nfor relatively small networks. We have implemented an FBKAN block (FBPINNs),\nfor domain decomposition method for KANs that allows for several small\nKANs to be trained in parallel to give accurate solutions for multiscale problems.\n\n\n\n**New Colab Examples:**\n> \u2b50 [Load Forecasting](#energy-systems)\n\n> \u2b50 [Function Approximation with Kolgomorov-Arnold Networks ](#function-approximation)\n\n## Getting Started\n\n```\npip install neuromancer\n```\n\nExtensive set of tutorials can be found in the \n[examples](https://github.com/pnnl/neuromancer/tree/master/examples) folder and the [Tutorials](#domain-examples) below.\nInteractive notebook versions of examples are available on Google Colab!\nTest out NeuroMANCER functionality before cloning the repository and setting up an\nenvironment.\n\nThe notebooks below introduce the core abstractions of the NeuroMANCER library, in particular our symbolic programming interface and Node classes. \n\n### Symbolic Variables, Nodes, Constraints, Objectives, and Systems Classes\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/tutorials/part_1_linear_regression.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\nPart 1: Linear regression in PyTorch vs NeuroMANCER. \n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/tutorials/part_2_variable.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\nPart 2: NeuroMANCER syntax tutorial: variables, constraints, and objectives. \n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/tutorials/part_3_node.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\nPart 3: NeuroMANCER syntax tutorial: modules, Node, and System class.\n\n\n\n### PyTorch Lightning Integration\n\nWe have integrated PyTorch Lightning to streamline code, enable custom training logic, support GPU and multi-GPU setups, and handle large-scale, memory-intensive learning tasks.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/lightning_integration_examples/Part_1_lightning_basics_tutorial.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 1: Lightning Integration Basics.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/lightning_integration_examples/Part_2_lightning_advanced_and_gpu_tutorial.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 2: Lightning Advanced Features and Automatic GPU Support.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/lightning_integration_examples/other_examples/lightning_custom_training_example.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 4: Defining Custom Training Logic via Lightning Modularized Code.\n\n### Example\nQuick example for how to solve parametric constrained optimization problem using NeuroMANCER, leveraging our symbolic programming interface, Node and Variable, Blocks, SLiM library, and PenaltyLoss classes. \n\n```python \n# Neuromancer syntax example for constrained optimization\nimport neuromancer as nm\nimport torch \n\n# define neural architecture \nfunc = nm.modules.blocks.MLP(insize=1, outsize=2, \n linear_map=nm.slim.maps['linear'], \n nonlin=torch.nn.ReLU, hsizes=[80] * 4)\n# wrap neural net into symbolic representation via the Node class: map(p) -> x\nmap = nm.system.Node(func, ['p'], ['x'], name='map')\n \n# define decision variables\nx = nm.constraint.variable(\"x\")[:, [0]]\ny = nm.constraint.variable(\"x\")[:, [1]]\n# problem parameters sampled in the dataset\np = nm.constraint.variable('p')\n\n# define objective function\nf = (1-x)**2 + (y-x**2)**2\nobj = f.minimize(weight=1.0)\n\n# define constraints\ncon_1 = 100.*(x >= y)\ncon_2 = 100.*(x**2+y**2 <= p**2)\n\n# create penalty method-based loss function\nloss = nm.loss.PenaltyLoss(objectives=[obj], constraints=[con_1, con_2])\n# construct differentiable constrained optimization problem\nproblem = nm.problem.Problem(nodes=[map], loss=loss)\n```\n\n\n## Domain Examples\n\nNeuroMANCER is built to tackle a variety of domain-specific modeling and control problems using its array of methods. Here we show how to model and control building energy systems, as well as apply load forecasting techniques. \n\nFor more in-depth coverage of our methods, please see our general [Tutorials](#tutorials-on-methods-for-modeling-optimization-and-control) section below. \n\n### Energy Systems\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/NODE_building_dynamics.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Learning Building Thermal Dynamics using Neural ODEs \n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/NODE_RC_networks.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Multi-zone Building Thermal Dynamics Resistance-Capacitance network with Neural ODEs\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/NODE_swing_equation.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Learning Swing Equation Dynamics using Neural ODEs\n\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/DPC_building_control.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Learning to Control Indoor Air Temperature in Buildings\n\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/HVAC_load_forecasting.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Energy Load Forecasting for the Air Handling System of an Office Building with MLP and CNN models\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/building_load_forecasting_Transformers.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Energy Load Forecasting for Building with Transformers Model\n\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/domain_examples/DPC_PSH.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Learning to Control Pumped-storage Hyrdoelectricity System\n\n\n## Tutorials on Methods for Modeling, Optimization, and Control\n### Learning to Optimize (L2O) Parametric Programming\n\nNeuromancer allows you to formulate and solve a broad class of parametric optimization problems leveraging machine learning to learn the solutions to such problems. [More information on Parametric programming](https://github.com/pnnl/neuromancer/tree/develop/examples/parametric_programming)\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_1_basics.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\nPart 1: Learning to solve a constrained optimization problem.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_2_pQP.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\nPart 2: Learning to solve a quadratically-constrained optimization problem.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_3_pNLP.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>\nPart 3: Learning to solve a set of 2D constrained optimization problems.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_4_projectedGradient.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> \nPart 4: Learning to solve a constrained optimization problem with the projected gradient.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_5_cvxpy_layers.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> \nPart 5: Using Cvxpylayers for differentiable projection onto the polytopic feasible set. \n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/parametric_programming/Part_6_pQp_lopoCorrection.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> \nPart 6: Learning to optimize with metric learning for Operator Splitting layers. \n\n### Learning to Control (L2C)\nNeuromancer allows you to learn control policies for full spectrum of white/grey/black-box dynamical systems, subject to choice constraints and objective functions. \n[More information on Differential Predictive Control](https://github.com/pnnl/neuromancer/tree/develop/examples/control)\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_1_stabilize_linear_system.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 1: Learning to stabilize a linear dynamical system.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_2_stabilize_ODE.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 2: Learning to stabilize a nonlinear differential equation.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_3_ref_tracking_ODE.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 3: Learning to control a nonlinear differential equation.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_4_NODE_control.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 4: Learning neural ODE model and control policy for an unknown dynamical system.\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/control/Part_5_neural_Lyapunov.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 5: Learning neural Lyapunov function for a nonlinear dynamical system.\n\n### Function Approximation\nNeuromancer is up-to-date with state-of-the-art methods. Here we showcase the powerful Kolgomorov-Arnold networks [More information on Kolgomorov-Arnold Networks](https://github.com/pnnl/neuromancer/tree/develop/examples/KANs)\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/feature/fbkans/examples/KANs/p1_fbkan_vs_kan_noise_data_1d.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 1: A comparison of KANs and FBKANs in learning a 1D multiscale function with noise\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/feature/fbkans/examples/KANs/p2_fbkan_vs_kan_noise_data_2d.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 2: A comparison of KANs and FBKANs in learning a 2D multiscale function with noise\n\n\n### Neural Operators\nNeuromancer allows one to use machine learning, prior physics and domain knowledge, to construct mathematical and differentiabl models of dynamical systems given the measured observations of the system behavior.\n[More information on System ID via Neural State Space Models and ODEs](https://github.com/pnnl/neuromancer/tree/develop/examples/ODEs)\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_1_NODE.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 1: Neural Ordinary Differential Equations (NODEs)\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_2_param_estim_ODE.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 2: Parameter estimation of ODE system\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_3_UDE.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 3: Universal Differential Equations (UDEs)\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_4_nonauto_NODE.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 4: NODEs with exogenous inputs\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_5_nonauto_NSSM.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 5: Neural State Space Models (NSSMs) with exogenous inputs\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_6_NetworkODE.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 6: Data-driven modeling of resistance-capacitance (RC) network ODEs\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_7_DeepKoopman.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 7: Deep Koopman operator\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_8_nonauto_DeepKoopman.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 8: control-oriented Deep Koopman operator\n\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/ODEs/Part_9_SINDy.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 9: Sparse Identification of Nonlinear Dynamics (SINDy)\n\n\n### Physics-Informed Neural Networks (PINNs)\nNeuromancer's symbolic programming design is perfectly suited for solving PINNs. [More information on PINNs](https://github.com/pnnl/neuromancer/tree/develop/examples/PDEs)\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_1_PINN_DiffusionEquation.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 1: Diffusion Equation\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_2_PINN_BurgersEquation.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 2: Burgers' Equation\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_3_PINN_BurgersEquation_inverse.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 3: Burgers' Equation w/ Parameter Estimation (Inverse Problem)\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_4_PINN_LaplaceEquationSteadyState.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 4: Laplace's Equation (steady-state)\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_5_Pendulum_Stacked.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 5: Damped Pendulum (stacked PINN)\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/PDEs/Part_6_PINN_NavierStokesCavitySteady_KAN.ipynb\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> Part 6: Navier-Stokes equation (lid-driven cavity flow, steady-state, KAN)\n\n### Stochastic Differential Equations (SDEs) \nNeuromancer has been integrated with TorchSDE to handle stochastic dynamical systems. [More information on SDEs](https://github.com/pnnl/neuromancer/tree/develop/examples/SDEs)\n+ <a target=\"_blank\" href=\"https://colab.research.google.com/github/pnnl/neuromancer/blob/master/examples/SDEs/sde_walkthrough.ipynb\">\n <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a> LatentSDEs: \"System Identification\" of Stochastic Processes using Neuromancer x TorchSDE\n\n\n\n\n## Documentation and User Guides\nThe documentation for the library can be found [online](https://pnnl.github.io/neuromancer/). \nThere is also an [introduction video](https://www.youtube.com/watch?v=YkFKz-DgC98) covering \ncore features of the library. \n\nFor more information, including those for developers, please go to our [Developer and User Guide](USER_GUIDE.md)\n\n\n## Installation\nSimply run \n```\npip install neuromancer\n```\nFor manual installation please refer to [Installation Instructions](INSTALLATION.md)\n\n## Community Information\nWe welcome contributions and feedback from the open-source community! \n\n### Contributions, Discussions, and Issues\nPlease read the [Community Development Guidelines](https://github.com/pnnl/neuromancer/blob/master/CONTRIBUTING.md) \nfor further information on contributions, [discussions](https://github.com/pnnl/neuromancer/discussions), and [Issues](https://github.com/pnnl/neuromancer/issues).\n\n### Release notes\nSee the [Release notes](https://github.com/pnnl/neuromancer/blob/master/RELEASE_NOTES.md) documenting new features.\n\n### License\nNeuroMANCER comes with [BSD license](https://en.wikipedia.org/wiki/BSD_licenses).\nSee the [license](https://github.com/pnnl/neuromancer/blob/master/LICENSE.md) for further details. \n\n\n## Publications \n+ [Jan Drgona, Aaron Tuor, Draguna Vrabie, Learning Constrained Parametric Differentiable Predictive Control Policies With Guarantees, IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2024](https://ieeexplore.ieee.org/abstract/document/10479163)\n+ [Renukanandan Tumu, Wenceslao Shaw Cortez, J\u00e1n Drgo\u0148a, Draguna L. Vrabie, Sonja Glavaski, Differentiable Predictive Control for Large-Scale Urban Road Networks, \tarXiv:2406.10433, 2024](https://arxiv.org/abs/2406.10433)\n+ [Ethan King, James Kotary, Ferdinando Fioretto, Jan Drgona, Metric Learning to Accelerate Convergence of Operator Splitting Methods for Differentiable Parametric Programming, arXiv:2404.00882, 2024](https://arxiv.org/abs/2404.00882)\n+ [James Koch, Madelyn Shapiro, Himanshu Sharma, Draguna Vrabie, Jan Drgona, Neural Differential Algebraic Equations, arXiv:2403.12938, 2024](https://arxiv.org/abs/2403.12938)\n+ [Wenceslao Shaw Cortez, Jan Drgona, Draguna Vrabie, Mahantesh Halappanavar, A Robust, Efficient Predictive Safety Filter, arXiv:2311.08496, 2024](https://arxiv.org/abs/2311.08496)\n+ [Shrirang Abhyankar, Jan Drgona, Andrew August, Elliott Skomski, Aaron Tuor, Neuro-physical dynamic load modeling using differentiable parametric optimization, 2023 IEEE Power & Energy Society General Meeting (PESGM), 2023](https://ieeexplore.ieee.org/abstract/document/10253098)\n+ [James Koch, Zhao Chen, Aaron Tuor, Jan Drgona, Draguna Vrabie, Structural Inference of Networked Dynamical Systems with Universal Differential Equations, arXiv:2207.04962, (2022)](https://aps.arxiv.org/abs/2207.04962)\n+ [J\u00e1n Drgo\u0148a, Sayak Mukherjee, Aaron Tuor, Mahantesh Halappanavar, Draguna Vrabie, Learning Stochastic Parametric Differentiable Predictive Control Policies, IFAC ROCOND conference (2022)](https://www.sciencedirect.com/science/article/pii/S2405896322015877)\n+ [Sayak Mukherjee, J\u00e1n Drgo\u0148a, Aaron Tuor, Mahantesh Halappanavar, Draguna Vrabie, Neural Lyapunov Differentiable Predictive Control, IEEE Conference on Decision and Control Conference 2022](https://arxiv.org/abs/2205.10728)\n+ [Wenceslao Shaw Cortez, Jan Drgona, Aaron Tuor, Mahantesh Halappanavar, Draguna Vrabie, Differentiable Predictive Control with Safety Guarantees: A Control Barrier Function Approach, IEEE Conference on Decision and Control Conference 2022](https://arxiv.org/abs/2208.02319)\n+ [Ethan King, Jan Drgona, Aaron Tuor, Shrirang Abhyankar, Craig Bakker, Arnab Bhattacharya, Draguna Vrabie, Koopman-based Differentiable Predictive Control for the Dynamics-Aware Economic Dispatch Problem, 2022 American Control Conference (ACC)](https://ieeexplore.ieee.org/document/9867379)\n+ [Drgo\u0148a, J., Tuor, A. R., Chandan, V., & Vrabie, D. L., Physics-constrained deep learning of multi-zone building thermal dynamics. Energy and Buildings, 243, 110992, (2021)](https://www.sciencedirect.com/science/article/pii/S0378778821002760)\n+ [E. Skomski, S. Vasisht, C. Wight, A. Tuor, J. Drgo\u0148a and D. Vrabie, \"Constrained Block Nonlinear Neural Dynamical Models,\" 2021 American Control Conference (ACC), 2021, pp. 3993-4000, doi: 10.23919/ACC50511.2021.9482930.](https://ieeexplore.ieee.org/document/9482930)\n+ [Skomski, E., Drgo\u0148a, J., & Tuor, A. (2021, May). Automating Discovery of Physics-Informed Neural State Space Models via Learning and Evolution. In Learning for Dynamics and Control (pp. 980-991). PMLR.](https://proceedings.mlr.press/v144/skomski21a.html)\n+ [Drgo\u0148a, J., Tuor, A., Skomski, E., Vasisht, S., & Vrabie, D. (2021). Deep Learning Explicit Differentiable Predictive Control Laws for Buildings. IFAC-PapersOnLine, 54(6), 14-19.](https://www.sciencedirect.com/science/article/pii/S2405896321012933)\n+ [Tuor, A., Drgona, J., & Vrabie, D. (2020). Constrained neural ordinary differential equations with stability guarantees. arXiv preprint arXiv:2004.10883.](https://arxiv.org/abs/2004.10883)\n+ [Drgona, Jan, et al. \"Differentiable Predictive Control: An MPC Alternative for Unknown Nonlinear Systems using Constrained Deep Learning.\" Journal of Process Control Volume 116, August 2022, Pages 80-92](https://www.sciencedirect.com/science/article/pii/S0959152422000981)\n+ [Drgona, J., Skomski, E., Vasisht, S., Tuor, A., & Vrabie, D. (2020). Dissipative Deep Neural Dynamical Systems, in IEEE Open Journal of Control Systems, vol. 1, pp. 100-112, 2022](https://ieeexplore.ieee.org/document/9809789)\n+ [Drgona, J., Tuor, A., & Vrabie, D., Learning Constrained Adaptive Differentiable Predictive Control Policies With Guarantees, arXiv preprint arXiv:2004.11184, (2020)](https://arxiv.org/abs/2004.11184)\n\n\n## Cite as\n```yaml\n@article{Neuromancer2023,\n title={{NeuroMANCER: Neural Modules with Adaptive Nonlinear Constraints and Efficient Regularizations}},\n author={Drgona, Jan and Tuor, Aaron and Koch, James and Shapiro, Madelyn and Jacob, Bruno and Vrabie, Draguna},\n Url= {https://github.com/pnnl/neuromancer}, \n year={2023}\n}\n```\n\n## Development team\n\n**Active core developers**: [Jan Drgona](https://drgona.github.io/), [Rahul Birmiwal](https://www.linkedin.com/in/rahul-birmiwal009/), [Bruno Jacob](https://brunopjacob.github.io/) \n**Notable contributors**: [Aaron Tuor](https://sw.cs.wwu.edu/~tuora/aarontuor/), Madelyn Shapiro, James Koch, Seth Briney, Bo Tang, Ethan King, Elliot Skomski, Zhao Chen, Christian M\u00f8ldrup Legaard \n**Scientific advisors**: Draguna Vrabie, Panos Stinis \n\nOpen-source contributions made by: \n<a href=\"https://github.com/pnnl/neuromancer/graphs/contributors\">\n <img src=\"https://contrib.rocks/image?repo=pnnl/neuromancer\" />\n</a>\n\nMade with [contrib.rocks](https://contrib.rocks).\n\n## Acknowledgments\nThis research was partially supported by the Mathematics for Artificial Reasoning in Science (MARS) and Data Model Convergence (DMC) initiatives via the Laboratory Directed Research and Development (LDRD) investments at Pacific Northwest National Laboratory (PNNL), by the U.S. Department of Energy, through the Office of Advanced Scientific Computing Research's \u201cData-Driven Decision Control for Complex Systems (DnC2S)\u201d project, and through the Energy Efficiency and Renewable Energy, Building Technologies Office under the \u201cDynamic decarbonization through autonomous physics-centric deep learning and optimization of building operations\u201d and the \u201cAdvancing Market-Ready Building Energy Management by Cost-Effective Differentiable Predictive Control\u201d projects. This project was also supported from the U.S. Department of Energy, Advanced Scientific Computing Research program, under the Uncertainty Quantification for Multifidelity Operator Learning (MOLUcQ) project (Project No. 81739). PNNL is a multi-program national laboratory operated for the U.S. Department of Energy (DOE) by Battelle Memorial Institute under Contract No. DE-AC05-76RL0-1830.\n\n",
"bugtrack_url": null,
"license": "BSD-2-Clause",
"summary": "Neural Modules with Adaptive Nonlinear Constraints and Efficient Regularization",
"version": "1.5.2",
"project_urls": {
"documentation": "https://github.com/pnnl/neuromancer/",
"homepage": "https://github.com/pnnl/neuromancer/",
"repository": "https://github.com/pnnl/neuromancer/"
},
"split_keywords": [
"deep learning",
" pytorch",
" linear models",
" dynamical systems",
" data-driven control"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "d75298dfa2b9a197a2de146823ec5d993dab72bea0bfe6b44a4e608034a88385",
"md5": "7e0e20e92fc2927dbc20997b2e76509d",
"sha256": "a6efdc0b921dce6fbc15d1095a62244186288e6384dfba15bb027cfae7d946d2"
},
"downloads": -1,
"filename": "neuromancer-1.5.2-py3-none-any.whl",
"has_sig": false,
"md5_digest": "7e0e20e92fc2927dbc20997b2e76509d",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": "<3.11,>=3.9",
"size": 177049,
"upload_time": "2024-11-08T20:41:40",
"upload_time_iso_8601": "2024-11-08T20:41:40.292155Z",
"url": "https://files.pythonhosted.org/packages/d7/52/98dfa2b9a197a2de146823ec5d993dab72bea0bfe6b44a4e608034a88385/neuromancer-1.5.2-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "c95d5eb20f67f8527ae34fbf14eb6840bafc58a8597371a2b6da78260c7d9f6a",
"md5": "8941e96b9507c697b945298e0aab119a",
"sha256": "a8163ad92aa09ca2dc5e07baa4c1a957ac1acac1306c2d9db3d6c811b9ef99c5"
},
"downloads": -1,
"filename": "neuromancer-1.5.2.tar.gz",
"has_sig": false,
"md5_digest": "8941e96b9507c697b945298e0aab119a",
"packagetype": "sdist",
"python_version": "source",
"requires_python": "<3.11,>=3.9",
"size": 112067327,
"upload_time": "2024-11-08T20:42:23",
"upload_time_iso_8601": "2024-11-08T20:42:23.867976Z",
"url": "https://files.pythonhosted.org/packages/c9/5d/5eb20f67f8527ae34fbf14eb6840bafc58a8597371a2b6da78260c7d9f6a/neuromancer-1.5.2.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-11-08 20:42:23",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "pnnl",
"github_project": "neuromancer",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"lcname": "neuromancer"
}
JSON
