Search NASA⌕ Search

SEARCH · Search NASA

Results for “overparameterization”

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.

15 records
All sourcesNASA NTRSDOE OSTI
NASA NTRSDate not supplied

A Look at the Truths and Misconceptions of the Variational Quantum Eigensolver and the Implications of Overparameterization

In this work, we investigate loss landscapes of the variational quantum eigensolver (VQE) by quantifying the number of local minima through empirical analyses. We focus on minimal models in chemistry and physics so that we can do a complete analysis using more computationally expensive tools. We employ Hessian eigenvalue calculations and the nudged elastic band algorithm to characterize these landscapes. Our results expand upon the existing literature by highlighting the optimization challenges faced by VQE. We find that, as the number of parameters in our ansatz increases, the number of basins increases while the corresponding loss function values converge toward the global minimum value. This observation implies that overparameterization may lead to an ``effective convexity'' in VQE loss landscapes, a phenomenon supported by theoretical and numerical work in classical machine learning.

quantum computing↗
DOE OSTI2020

Trainability of ReLU networks and Data-dependent Initialization

Here, in this paper, we study the trainability of rectified linear unit (ReLU) networks at initialization. A ReLU neuron is said to be dead if it only outputs a constant for any input. Two death states of neurons are introduced–tentative and permanent death. A network is then said to be trainable if the number of permanently dead neurons is sufficiently small for a learning task. We refer to the probability of a randomly initialized network being trainable as trainability. We show that a network being trainable is a necessary condition for successful training, and the trainability serves as an upper bound of training success rates. In order to quantify the trainability, we study the probability distribution of the number of active neurons at initialization. In many applications, overspecified or overparameterized neural networks are successfully employed and shown to be trained effectively. With the notion of trainability, we show that overparameterization is both a necessary and a sufficient condition for achieving a zero training loss. Furthermore, we propose a data-dependent initialization method in an overparameterized setting. Numerical examples are provided to demonstrate the effectiveness of the method and our theoretical findings.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Physics-informed latent neural operator for real-time predictions of time-dependent parametric PDEs

Deep operator network (DeepONet) has shown significant promise as surrogate models for systems governed by partial differential equations (PDEs), enabling accurate mappings between infinite-dimensional function spaces. However, when applied to systems with high-dimensional input-output mappings arising from large numbers of spatial and temporal collocation points, these models often require heavily overparameterized networks, leading to long training times. Latent DeepONet addresses some of these challenges by introducing a two-step approach: first learning a reduced latent space using a separate model, followed by operator learning within this latent space. While efficient, this method is inherently data-driven and lacks mechanisms for incorporating physical laws, limiting its robustness and generalizability in data-scarce settings. Here, in this work, we propose PI-Latent-NO, a physics-informed latent neural operator framework that integrates governing physics directly into the learning process. Our architecture features two coupled DeepONets trained end-to-end: a Latent-DeepONet that learns a low-dimensional representation of the solution, and a Reconstruction-DeepONet that maps this latent representation back to the physical space. By embedding PDE constraints into the training via automatic differentiation, our method eliminates the need for labeled training data and ensures physics-consistent predictions. The proposed framework is both memory and compute-efficient, exhibiting near-constant scaling with problem size and demonstrating significant speedups over traditional physics-informed operator models. We validate our approach on a range of parametric PDEs, showcasing its accuracy, scalability, and suitability for real-time prediction in complex physical systems.

Latent representations↗
DOE OSTI2022

Logistic Regression in Clinical Studies

• A logistic regression model is used when the outcome of interest is binary. The term “logistic” refers to the underlying “logit” (log odds) function that is used to model the binary outcome. • Odds ratios are produced from a logistic regression model and have a useful interpretation. • Tips, tricks and concepts used to fit logistic regression models are similar to those used in linear regression models. • Modeling building that is knowledge-based rather than automatic is preferred in most applications of logistic regression. • A logistic regression model that is overparameterized (ie, too many variables for too few events) can result in odds ratios that are implausibly large and confidence intervals that are wide and uninterpretable. These types of “overfitted” models should be avoided. • Logistic regression models can be fit using most standard statistical software.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2023

Encoding nonlinear and unsteady aerodynamics of limit cycle oscillations using nonlinear sparse Bayesian learning

This article investigates the applicability of a recently proposed, nonlinear sparse Bayesian learning (NSBL) algorithm to identify and estimate the complex aerodynamics of limit cycle oscillations. NSBL provides a semi-analytical framework for determining the data-optimal sparse model nested within a (potentially) over-parameterized model. This is particularly relevant to nonlinear dynamical systems where modelling approaches involve the use of physics-based and data-driven components. In such cases, the data-driven components, where analytical descriptions of the physical processes are not readily available, are often prone to overfitting, meaning that the empirical aspects of these models will often involve the calibration of an unnecessarily large number of parameters. While an overparameterized model may fit the observed data well, such models may be inadequate for making predictions in regimes that are different from those wherein the data were recorded. In view of this, it is desirable to not only calibrate the model parameters, but also identify the optimal compromise between data fit and model complexity. In this article, we exhibit the optimal model discovery for an aeroelastic system wherein the structural dynamics are well-known and described by a differential equation model, coupled with a semi-empirical aerodynamic model for laminar separation flutter, resulting in low-amplitude limit cycle oscillations (LCO). To illustrate the performance of the algorithm, in this article, we use synthetic data and demonstrate the ability of the algorithm to correctly rediscover the optimal model and model parameters, given a known data-generating model. The synthetic data are generated from a forward simulation of a known differential equation model with parameters selected so as to mimic the dynamics observed in wind-tunnel experiments. Subsequently, we demonstrate the performance of the algorithm for model selection using noisy LCO data from wind tunnel experiments. As there is no ground truth available for the experimental data case, we provide a comparison between NSBL and Bayesian model selection to validate the results, and demonstrate the use of NSBL as an efficient alternative to traditional methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

End block dynamics in unentangled polymers by dielectric spectroscopy

Dielectric spectroscopy measures the dynamics of polymer melts over a broad frequency range. Developing a theory for the spectral shape can extend the analysis of dielectric spectra beyond determining relaxation times from the peak maxima and adds physical meaning to shape parameters determined with empirical fit functions. Toward this goal, we use the experimental results on unentangled poly(isoprene), and unentangled poly(butylene oxide), polymer melts, to test whether the concept of end blocks could be one reason for the Rouse model deviating from experimental data. These end blocks have been suggested by simulations and neutron spin echo spectroscopy and are a consequence of the monomeric friction coefficient depending on the position of the bead in the chain. The concept of an end block is an approximation which partitions the chain in a middle and two end blocks to avoid overparameterization by a continuous position dependent change of the friction parameter. Analysis of dielectric spectra shows that the deviations of the calculated from the experimental normal mode cannot be related to the end block relaxation. However, the results do not contradict an end block hiding below the segmental relaxation peak. Finally, it seems that the results are compatible with an end block being the specific part of the sub-Rouse chain interpretation close to the chain ends.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Progress toward favorable landscapes in quantum combinatorial optimization

The performance of variational quantum algorithms relies on the success of using quantum and classical computing resources in tandem. Here, we study how these quantum and classical components interrelate. In particular, we focus on algorithms for solving the combinatorial optimization problem MaxCut, and study how the structure of the classical optimization landscape relates to the quantum circuit used to evaluate the MaxCut objective function. In order to analytically characterize the impact of quantum features on the critical points of the landscape, we consider a family of quantum circuit ansätze composed of mutually commuting elements. We identify multiqubit operations as a key resource and show that overparameterization allows for obtaining favorable landscapes. Namely, we prove that an ansatz from this family containing exponentially many variational parameters yields a landscape free of local optima for generic graphs. However, we further prove that these ansätze do not offer superpolynomial advantages over purely classical MaxCut algorithms. Here, we then present a series of numerical experiments illustrating that noncommutativity and entanglement are important features for improving algorithm performance.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2020

Reinforcement Learning via Gaussian Processes with Neural Network Dual Kernels

While deep neural networks (DNNs) and Gaussian Processes (GPs) are both popularly utilized to solve problems in reinforcement learning, both approaches feature undesirable drawbacks for challenging problems. DNNs learn complex non-linear embeddings, but do not naturally quantify uncertainty and are often data-inefficient to train. GPs infer posterior distributions over functions, but popular kernels exhibit limited expressivity on complex and high-dimensional data. Fortunately, recently discovered conjugate and neural tangent kernel functions encode the behavior of overparameterized neural networks in the kernel domain. We demonstrate that these kernels can be efficiently applied to regression and reinforcement learning problems by analyzing a baseline case study.We apply GPs with neural network dual kernels to solve reinforcement learning tasks for the first time. We demonstrate, using the well understood mountain-car problem, that GPs empowered with dual kernels perform at least as well as those using the conventional radial basis function kernel. Finally, we conjecture that by inheriting the probabilistic rigor of GPs and the powerful embedding properties of DNNs, GPs using NN dual kernels will empower future reinforcement learning models on difficult domains.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Spike-and-Slab Shrinkage Priors for Structurally Sparse Bayesian Neural Networks

Network complexity and computational efficiency have become increasingly significant aspects of deep learning. Sparse deep learning addresses these challenges by recovering a sparse representation of the underlying target function by reducing heavily overparameterized deep neural networks. Specifically, deep neural architectures compressed via structured sparsity (e.g., node sparsity) provide low-latency inference, higher data throughput, and reduced energy consumption. In this article, we explore two well-established shrinkage techniques, Lasso and Horseshoe, for model compression in Bayesian neural networks (BNNs). To this end, we propose structurally sparse BNNs, which systematically prune excessive nodes with the following: 1) spike-and-slab group Lasso (SS-GL) and 2) SS group Horseshoe (SS-GHS) priors, and develop computationally tractable variational inference, including continuous relaxation of Bernoulli variables. We establish the contraction rates of the variational posterior of our proposed models as a function of the network topology, layerwise node cardinalities, and bounds on the network weights. Furthermore, we empirically demonstrate the competitive performance of our models compared with the baseline models in prediction accuracy, model compression, and inference latency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

“Understanding Robustness Lottery”: A Geometric Visual Comparative Analysis of Neural Network Pruning Approaches

Deep learning approaches have provided state-of-the-art performance in many applications by relying on large and overparameterized neural networks. However, such networks are very brittle and are difficult to deploy on resource-limited platforms. Model pruning, i.e., reducing the size of the network, is a widely adopted strategy that can lead to a more robust and compact model. Many heuristics exist for model pruning, but our understanding of the pruning process remains limited due to the black-box nature of a neural network model. Empirical studies show that some heuristics improve performance whereas others can make models more brittle. Here, this work aims to shed light on how different pruning methods alter the network’s internal feature representation and the corresponding impact on model performance. To facilitate a comprehensive comparison and characterization of the high-dimensional model feature space, we introduce a visual geometric analysis of feature representations. We evaluated a set of critical geometric concepts decomposed from the commonly adopted classification loss and used them to design a visualization system to compare and highlight the impact of pruning on model performance and feature representation. The proposed tool provides an environment for an in-depth comparison of pruning methods and a comprehensive understanding of how the model responds to common data corruption. By leveraging the proposed visualization, machine learning researchers can reveal the similarities between pruning methods and redundancy in robustness evaluation benchmarks, obtain geometric insights about the differences between pruned models that achieve superior robustness performance, and identify samples that are robust or fragile to model pruning and common data corruption.

Li, Zhimin [Univ. of Utah, Salt Lake City, UT (Uni↗
DOE OSTI2021

Correspondence of NNGP Kernel and the Matérn Kernel

Kernels representing limiting cases of neural network architectures have recently gained popularity. However, the application and performance of these new kernels compared to existing options, such as the Matérn kernel, is not well studied. We take a practical approach to explore the neural network Gaussian process (NNGP) kernel and its application to data in Gaussian process regression. We first demonstrate the necessity of normalization to produce valid NNGP kernels and explore related numerical challenges. We further demonstrate that the predictions from this model are quite inflexible, and therefore do not vary much over the valid hyperparameter sets. We then demonstrate a surprising result that the predictions given from the NNGP kernel correspond closely to those given by the Matérn kernel under specific circumstances, which suggests a deep similarity between overparameterized deep neural networks and the Matérn kernel. Finally, we demonstrate the performance of the NNGP kernel as compared to the Matérn kernel on three benchmark data cases, and we conclude that for its flexibility and practical performance, the Matérn kernel is preferred to the novel NNGP in practical applications.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1977

Theoretic aspects of the identification of the parameters in the optimal control model

The identification of the parameters of the optimal control model from input-output data of the human operator is considered. Accepting the basic structure of the model as a cascade of a full-order observer and a feedback law, and suppressing the inherent optimality of the human controller, the parameters to be identified are the feedback matrix, the observer gain matrix, and the intensity matrices of the observation noise and the motor noise. The identification of the parameters is a statistical problem, because the system and output are corrupted by noise, and therefore the solution must be based on the statistics (probability density function) of the input and output data of the human operator. However, based on the statistics of the input-output data of the human operator, no distinction can be made between the observation and the motor noise, which shows that the model suffers from overparameterization.

Vanwijk, R. A.↗
NASA NTRS1983

Algorithms for real-time flutter identification

The work reported here addresses the important algorithm issues necessary to achieve a real-time flutter monitoring system; namely, the guidelines for choosing appropriate model forms, reduction of the parameter convergence transient, handling multiple modes, the effect of overparameterization, and estimate accuracy predictions, both online and for experiment design. An approach for efficiently computing continuous-time flutter parameter Cramer-Rao estimate error bounds has been developed. This enables a convincing comparison of theoretical and simulation results, as well as off-line studies in preparation for a flight test. Theoretical predictions, simulation and flight test results from the NASA/Dryden Drones for Aerodynamic and Structural Test (DAST) Program are compared.

Walker, R. A.↗
NASA NTRS1991

Indirect techniques for adaptive input-output linearization of non-linear systems

A technique of indirect adaptive control based on certainty equivalence for input output linearization of nonlinear systems is proven convergent. It does not suffer from the overparameterization drawbacks of the direct adaptive control techniques on the same plant. This paper also contains a semiindirect adaptive controller which has several attractive features of both the direct and indirect schemes.

Teel, Andrew↗
NASA NTRS2009

Quantifying Parameter Sensitivity, Interaction and Transferability in Hydrologically Enhanced Versions of Noah-LSM over Transition Zones

We use sensitivity analysis to identify the parameters that are most responsible for shaping land surface model (LSM) simulations and to understand the complex interactions in three versions of the Noah LSM: the standard version (STD), a version enhanced with a simple groundwater module (GW), and version augmented by a dynamic phenology module (DV). We use warm season, high-frequency, near-surface states and turbulent fluxes collected over nine sites in the US Southern Great Plains. We quantify changes in the pattern of sensitive parameters, the amount and nature of the interaction between parameters, and the covariance structure of the distribution of behavioral parameter sets. Using Sobol s total and first-order sensitivity indexes, we show that very few parameters directly control the variance of the model output. Significant parameter interaction occurs so that not only the optimal parameter values differ between models, but the relationships between parameters change. GW decreases parameter interaction and appears to improve model realism, especially at wetter sites. DV increases parameter interaction and decreases identifiability, implying it is overparameterized and/or underconstrained. A case study at a wet site shows GW has two functional modes: one that mimics STD and a second in which GW improves model function by decoupling direct evaporation and baseflow. Unsupervised classification of the posterior distributions of behavioral parameter sets cannot group similar sites based solely on soil or vegetation type, helping to explain why transferability between sites and models is not straightforward. This evidence suggests a priori assignment of parameters should also consider climatic differences.

Rosero, Enrique↗