SEARCH · Search NASA
Results for “neural generative models”
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Integrating deep neural networks with full-waveform inversion: Reparameterization, regularization, and uncertainty quantification
Full-waveform inversion (FWI) is an accurate imaging approach for modeling the velocity structure by minimizing the misfit between recorded and predicted seismic waveforms. However, the strong nonlinearity of FWI resulting from fitting oscillatory waveforms can trap the optimization in local minima. We have adopted a neural-network-based full-waveform inversion (NNFWI) method that integrates deep neural networks with FWI by representing the velocity model with a generative neural network. Neural networks can naturally introduce spatial correlations as regularization to the generated velocity model, which suppresses noise in the gradients and mitigates local minima. Furthermore, the velocity model generated by neural networks is input to the same partial differential equation (PDE) solvers used in conventional FWI. The gradients of the neural networks and PDEs are calculated using automatic differentiation, which back propagates gradients through the acoustic PDEs and neural network layers to update the weights of the generative neural network. Experiments on 1D velocity models, the Marmousi model, and the 2004 BP model determine that NNFWI can mitigate local minima, especially for imaging high-contrast features such as salt bodies, and it significantly improves the inversion in the presence of noise. Adding dropout layers to the neural network model also allows analyzing the uncertainty of the inversion results through Monte Carlo dropout. NNFWI opens a new pathway to combine deep learning and FWI for exploiting the characteristics of deep neural networks and the high accuracy of PDE solvers. Because NNFWI does not require extra training data and optimization loops, it provides an attractive and straightforward alternative to conventional FWI.
Structured Neural Network Modeling for Developing Digital Twins Models of Hydropower Generation Units
Dynamic modeling is a key part in the development of digital twin (DT) for dynamic systems. This is true for hydropower systems, where whole system modeling including penstock, turbine and generators, etc is important in realizing actuate modeling for the real systems. On the other hand, in response to the large variations of the power demand due to increased penetration of renewables such as wind and solar, hydropower systems are now required to operate in a large power generation range. This situation triggers the nonlinear characteristics of the generation unit with respect to its models. As such, it is imperative to use data driven modeling such as neural networks to learn the nonlinear dynamics of the hydropower generation unit. To achieve this objective, this study constructs a modeling and learning algorithm integrated with multiple structured neural network models for the modeling of turbine shaft speed, penstock pressure, and generator power output based on the generator power control setpoint, field current, and field voltage. In addition, the study uses the hydropower data from Tacoma Public Utilities to train and validate the proposed neural network algorithm. The results have shown that this structured neural network modeling approach can learn the system dynamics effectively by using the real-time data collected from the hydropower system with the desired modeling results.
Near-Infrared Spectroscopy can Predict Anatomical Abundance in Corn Stover
Feedstock heterogeneity is a key challenge impacting the deconstruction and conversion of herbaceous lignocellulosic biomass to biobased fuels, chemicals, and materials. Upstream processing to homogenize biomass feedstock streams into their anatomical components via air classification allows for a more tailored approach to subsequent mechanical and chemical processing. Here, we show that differing corn stover anatomical tissues respond differently to pretreatment and enzymatic hydrolysis and therefore, a one-size-fits-all approach to chemical processing biomass is inappropriate. To inform on-line downstream processing, a robust and high-throughput analytical technique is needed to quantitatively characterize the separated biomass. Predictive correlation of near-infrared spectra to biomass chemical composition is such a technique. Here, we demonstrate the capability of models developed using an “off-the-shelf,” industrially relevant spectrometer with limited spectral range to make strong predictions of both cell wall chemical composition and the relative abundance of anatomical components of the corn stover, the latter for the first time ever. Gaussian process regression (GPR) yields stronger correlations (average R 2 v = 88% for chemical composition and 95% for anatomical relative abundance) than the more commonly used partial least squares (PLS) regression (average R 2 v = 84% for chemical composition and 92% for anatomical relative abundance). In nearly all cases, both GPR and PLS outperform models generated using neural networks. These results highlight the potential for coupling NIRS with predictive models based on GPR due to the potential to yield more robust correlations.
Attention-based quantum tomography
Abstract With rapid progress across platforms for quantum systems, the problem of many-body quantum state reconstruction for noisy quantum states becomes an important challenge. There has been a growing interest in approaching the problem of quantum state reconstruction using generative neural network models. Here we propose the ‘attention-based quantum tomography’ (AQT), a quantum state reconstruction using an attention mechanism-based generative network that learns the mixed state density matrix of a noisy quantum state. AQT is based on the model proposed in ‘Attention is all you need’ by Vaswani et al (2017 NIPS ) that is designed to learn long-range correlations in natural language sentences and thereby outperform previous natural language processing (NLP) models. We demonstrate not only that AQT outperforms earlier neural-network-based quantum state reconstruction on identical tasks but that AQT can accurately reconstruct the density matrix associated with a noisy quantum state experimentally realized in an IBMQ quantum computer. We speculate the success of the AQT stems from its ability to model quantum entanglement across the entire quantum system much as the attention model for NLP captures the correlations among words in a sentence.
Non-Blind Deblurring for Fluorescence: A Deformable Latent Space Approach with Kernel Parameterization
We report N\non-blind deblurring (NBD) is a modeling method of the image deblurring problem in computer vision, where the blurring kernel is known or can be externally estimated. In this paper, we attempt to solve a parametric NBD problem, inspired by the simultaneous acquisition of ptychography and fluorescent imaging (FI). Ptychography is an imaging method that favors larger probes, i.e. convolutional kernels, while FI relies on a small probe for high resolution. Also, the kernel can be solved during ptychographic reconstruction. With Ptycho-FI using the same larger kernel, we can perform NBD on the blurred fluorescent images to achieve high-resolution FI, and thus speed up the experiments. To this end, we design a deep latent space deformation network that is directly parameterized by the kernel. The network consists of three components: encoder, deformer, and decoder, where the deformer is specifically meant to rectify the latent space representations of blurred images to a standard latent space, regardless of the kernel. The deformation network is trained with a two-stage training scheme. We conduct extensive experiments to confirm that our parametric model can adapt to drastically different blurring kernels and perform robust deblurring.
From clutter to clarity: Emergent neural operators via questionnaire metrics
Real-world datasets in chemical engineering and bioengineering processes—such as those from catalytic reactors, multiphase flows, polymerization reactors, bioreactors, and clinical trials—can often be unlabeled or disorganized, rendering the training of existing supervised learning models ineffective at learning the underlying dynamics. To salvage these datasets for decision-making, we first seek to obtain clarity from the cluttered data. Here, we present a framework for developing “structural” generative models, discovering emergent equations, and constructing efficient emulators from scrambled datasets by integrating unsupervised organizational learning techniques (Questionnaires) with advanced deep learning architectures (Deep Hidden Physics Models and Deep Operator Networks). Our approach is demonstrated on two illustrative model systems: (a) a 1D advection–diffusion partial differential equation representing a winding underground pipe and (b) an ensemble of Stuart–Landau oscillators, an agent-based system of coupled ordinary differential equations. In both cases, we successfully reconstruct meaningful spatial, temporal, and parameter embeddings from scrambled data, enabling good predictions of system dynamics. As a result, we highlight the framework’s potential for broader applications, enabling data-driven system identification in fields with inherently disorganized or hidden parameter spaces.
Evolutionary Architecture Search for Generative Adversarial Networks Based on Weight Sharing
Generative adversarial networks (GANs) are a powerful generative technique but frequently face challenges with training stability. Network architecture plays a significant role in determining the final output of GANs, but designing a fine architecture demands extensive domain expertise. This article aims to address this issue by searching for high-performance generator’s architectures through neural architecture search (NAS). The proposed approach, called evolutionary weight sharing GANs (EWSGAN), is based on weight sharing and comprises two steps. First, a supernet of the generator is trained using weight sharing. Second, a multiobjective evolutionary algorithm (MOEA) is employed to identify optimal subnets from the supernet. These subnets inherit weights directly from the supernet for fitness assessment. Two strategies are used to stabilize the training of the generator supernet: 1) a fair single-path sampling strategy and 2) a discarding strategy. Experimental results indicate that the architecture searched by our method achieved a new state-of-the-art among NAS–GAN methods with a Fréchet inception distance (FID) of 9.09 and an inception score (IS) of 8.99 on the CIFAR-10 dataset. Finally, it also demonstrates competitive performance on the STL-10 dataset, achieving FID of 21.89 and IS of 10.51.
Generative Vulnerability Assessment for Cyber-Physical Systems
Cyber-physical systems (CPS) are highly susceptible to malicious attacks due to their complex dynamics and interconnectivity. A comprehensive understanding of their vulnerabilities is essential for designing effective resilience measures. This paper presents a data-driven attack generative system for evaluating the vulnerability of CPS. The proposed approach formulates the vulnerability assessment problem as determining the feasibility of a specific attack set based on two boundary functions that represent the effectiveness and stealthiness of attacks. The attack generative model is trained using a custom loss function, with two universal approximators designed to learn the effectiveness and stealthiness functions simultaneously. Theoretical results for successful generation and asymptotic convergence of the resulting training algorithm are given. As a result, the proposed approach is evaluated via numerical simulation of an IEEE 14-bus system and gas pipeline systems, demonstrating its viability in learning how to attack nonlinear CPS and identify potential vulnerabilities.
Neural Network Control of a Magnetically Suspended Rotor System
Magnetic bearings offer significant advantages because they do not come into contact with other parts during operation, which can reduce maintenance. Higher speeds, no friction, no lubrication, weight reduction, precise position control, and active damping make them far superior to conventional contact bearings. However, there are technical barriers that limit the application of this technology in industry. One of them is the need for a nonlinear controller that can overcome the system nonlinearity and uncertainty inherent in magnetic bearings. At the NASA Lewis Research Center, a neural network was selected as a nonlinear controller because it generates a neural model without any detailed information regarding the internal working of the magnetic bearing system. It can be used even for systems that are too complex for an accurate system model to be derived. A feed-forward architecture with a back-propagation learning algorithm was selected because of its proven performance, accuracy, and relatively easy implementation.
Learning generative neural networks with physics knowledge
Deep generative neural networks have enabled modeling complex distributions, but incorporating physics knowledge into the neural networks is still challenging and is at the core of current physics-based machine learning research. To this end, we propose a physics generative neural network (PhysGNN), a new class of generative neural networks for learning unknown distributions in a physical system described by partial differential equations (PDE). PhysGNN couples PDE systems with generative neural networks. It is a fully differentiable model that allows back-propagation of gradients through both numerical PDE solvers and generative neural networks, and is trained by minimizing the discrete Wasserstein distance between generated and observed probability distributions of the PDE outputs using the stochastic gradient descent method. Moreover, PhysGNN does not require adversarial training like standard generative neural networks, which offers better stability than adversarial training. We show that PhysGNN can learn complex distributions in stochastic inverse problems, where conventional methods such as maximum likelihood estimation and momentum matching methods may be inapplicable when little knowledge is known about the form of unknown distributions or the physical model is too complex. Furthermore, our method allows physics-based generative neural network training for learning complex distributions in the context of differential equations.
Deep Image Prior Enabled Full Waveform Inversion (Final Technical Report)
MS Student Naveen Gupta worked on the problem of full waveform inversion (FWI) using neural networks as shown in Figure 1. Our goal was to learn a neural network to represent the subsurface velocity model, which when fed into the FWI module (implemented using a numerical forward model of wave equations) produces amplitude estimates that match with ground-truth observations of amplitude. We used neural networks to solve the inverse problem of estimating velocity distributions for a given seismic amplitude data such that, once trained, our neural network model can generate a distribution of velocity profiles for different random vectors fed as inputs to the neural network model.
Predicting critical heat flux with uncertainty quantification and domain generalization using conditional variational autoencoders and deep neural networks
Deep generative models (DGMs) can generate synthetic data samples that closely resemble the original dataset, addressing data scarcity. In this work, we developed a conditional variational autoencoder (CVAE) to augment critical heat flux (CHF) data used for the 2006 Groeneveld lookup table. To compare with traditional methods, a fine-tuned deep neural network (DNN) regression model was evaluated on the same dataset. Both models achieved small mean absolute relative errors, with the CVAE showing more favorable results. Uncertainty quantification (UQ) was performed using repeated CVAE sampling and DNN ensembling. The DNN ensemble improved performance over the baseline, while the CVAE maintained consistent results with less variability and higher confidence. Both models achieved small errors inside and outside the training domain, with slightly larger errors outside. Altogether, the CVAE performed better than the DNN in predicting CHF and exhibited better uncertainty behavior.
Diffusion-Model-Assisted Supervised Learning of Generative Models for Density Estimation
Here, we present a supervised learning framework of training generative models for density estimation. Generative models, including generative adversarial networks (GANs), normalizing flows, and variational auto-encoders (VAEs), are usually considered as unsupervised learning models, because labeled data are usually unavailable for training. Despite the success of the generative models, there are several issues with the unsupervised training, e.g., requirement of reversible architectures, vanishing gradients, and training instability. To enable supervised learning in generative models, we utilize the score-based diffusion model to generate labeled data. Unlike existing diffusion models that train neural networks to learn the score function, we develop a training-free score estimation method. This approach uses mini-batch-based Monte Carlo estimators to directly approximate the score function at any spatial-temporal location in solving an ordinary differential equation (ODE), corresponding to the reverse-time stochastic differential equation (SDE). This approach can offer both high accuracy and substantial time savings in neural network training. Once the labeled data are generated, we can train a simple, fully connected neural network to learn the generative model in the supervised manner. Compared with existing normalizing flow models, our method does not require the use of reversible neural networks and avoids the computation of the Jacobian matrix. Compared with existing diffusion models, our method does not need to solve the reverse-time SDE to generate new samples. As a result, the sampling efficiency is significantly improved. We demonstrate the performance of our method by applying it to a set of 2D datasets as well as real data from the University of California Irvine (UCI) repository.
System and Method for Modeling the Flow Performance Features of an Object
The method and apparatus includes a neural network for generating a model of an object in a wind tunnel from performance data on the object. The network is trained from test input signals (e.g., leading edge flap position, trailing edge flap position, angle of attack, and other geometric configurations, and power settings) and test output signals (e.g., lift, drag, pitching moment, or other performance features). In one embodiment, the neural network training method employs a modified Levenberg-Marquardt optimization technique. The model can be generated 'real time' as wind tunnel testing proceeds. Once trained, the model is used to estimate performance features associated with the aircraft given geometric configuration and/or power setting input. The invention can also be applied in other similar static flow modeling applications in aerodynamics, hydrodynamics, fluid dynamics, and other such disciplines. For example, the static testing of cars, sails, and foils, propellers, keels, rudders, turbines, fins, and the like, in a wind tunnel, water trough, or other flowing medium.
Neural Network and Regression Methods Demonstrated in the Design Optimization of a Subsonic Aircraft
The neural network and regression methods of NASA Glenn Research Center s COMETBOARDS design optimization testbed were used to generate approximate analysis and design models for a subsonic aircraft operating at Mach 0.85 cruise speed. The analytical model is defined by nine design variables: wing aspect ratio, engine thrust, wing area, sweep angle, chord-thickness ratio, turbine temperature, pressure ratio, bypass ratio, fan pressure; and eight response parameters: weight, landing velocity, takeoff and landing field lengths, approach thrust, overall efficiency, and compressor pressure and temperature. The variables were adjusted to optimally balance the engines to the airframe. The solution strategy included a sensitivity model and the soft analysis model. Researchers generated the sensitivity model by training the approximators to predict an optimum design. The trained neural network predicted all response variables, within 5-percent error. This was reduced to 1 percent by the regression method. The soft analysis model was developed to replace aircraft analysis as the reanalyzer in design optimization. Soft models have been generated for a neural network method, a regression method, and a hybrid method obtained by combining the approximators. The performance of the models is graphed for aircraft weight versus thrust as well as for wing area and turbine temperature. The regression method followed the analytical solution with little error. The neural network exhibited 5-percent maximum error over all parameters. Performance of the hybrid method was intermediate in comparison to the individual approximators. Error in the response variable is smaller than that shown in the figure because of a distortion scale factor. The overall performance of the approximators was considered to be satisfactory because aircraft analysis with NASA Langley Research Center s FLOPS (Flight Optimization System) code is a synthesis of diverse disciplines: weight estimation, aerodynamic analysis, engine cycle analysis, propulsion data interpolation, mission performance, airfield length for landing and takeoff, noise footprint, and others.
LENS: Learning Enabled Network Synthesis
RTRC and UMD have developed novel machine learning based methods under the ARPA-E DIFFERENTIATE program for rapid acceleration of hypothesis generation in complex architecture design spaces involving both discrete choices of component inclusion and interconnection and continuous parametric decisions. The project named Learning Enabled Network Synthesis (LENS) further demonstrated the developed methods on challenging electrical power converter design problems by identifying the most suitable circuit topologies and simultaneously selecting the most appropriate components to achieve optimized design of power converter with improved performances. We demonstrated that LENS could enable exploration of very large design space of circuit topologies and components by addressing the limitations of conventional design process in non-linear, high switching speed, multi-dimensional power converter design and optimization. The key innovation developed in LENS is the seamless integration of statistical learning and logical reasoning techniques and building on the individual strengths of these techniques for rapid hypothesis discovery. The main component of LENS comprises of: 1) Graph Reasoning Engine (GRE) to enforce composition rules that rapidly reject all discrete architectures that are composed incorrectly and generates an adaptive database of feasible designs which can be used by ML modules, 2) Graph Generative Learning module which is a deep neural network based generative model for graph architectures which can enable design space exploration beyond the dataset generated by the GRE, 3) Graph Reduced Order Model (ROM) for graph domains for accelerating computation of output metrics, and 4) Active learning and Rule Discovery module for sample efficient learning and extracting logical rules from the learned ML models which will be integrated in the GRE to enhance the filtering effectiveness. LENS approach can be applied to any design domains where designs can be represented as multi-attribute graphs. The LENS team integrated the various technical innovations listed above into an optimization pipeline and exercised the optimization pipeline on the converter design problem. The LENS project demonstrated that the developed AI/ML technologies can be used to generate novel converter circuits >45x faster than experts on chosen use-cases. This can enable faster design space exploration and identification of new designs which are not considered by experts due to the increasing design space complexity. This has significant potential impact on the public and energy needs of the country. It is currently estimated that 30% of all electrical powers generated passes through power converters. The future estimate is that 80% of all power generated would be passing through converters. LENS fills a critical gap in this space since by accelerating the design process the designers would be able to generate more efficient converters which can lead to significant energy savings for the country.