SEARCH · Search NASA
Results for “Disentangling”
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.
INSURE: An Information Theory iNspired diSentanglement and pURification modEl for Domain Generalization
Domain Generalization (DG) aims to learn a generalizable model on the unseen target domain by only training on the multiple observed source domains. Although a variety of DG methods have focused on extracting domain-invariant features, the domain-specific class-relevant features have attracted attention and been argued to benefit generalization to the unseen target domain. To take into account the class-relevant domain-specific information, in this paper we propose an Information theory iNspired diSentanglement and pURification modEl (INSURE) to explicitly disentangle the latent features to obtain sufficient and compact (necessary) class-relevant feature for generalization to the unseen domain. Specifically, we first propose an information theory inspired loss function to ensure the disentangled class-relevant features contain sufficient class label information and the other disentangled auxiliary feature has sufficient domain information. Additionally, we further propose a paired purification loss function to let the auxiliary feature discard all the class-relevant information and thus the class-relevant feature will contain sufficient and compact (necessary) class-relevant information. Moreover, instead of using multiple encoders, we propose to use a learnable binary mask as our disentangler to make the disentanglement more efficient and make the disentangled features complementary to each other. We conduct extensive experiments on five widely used DG benchmark datasets including PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet. The proposed INSURE achieves state-of-the-art performance. We also empirically show that domain-specific class-relevant features are beneficial for domain generalization. The code is available at https://github.com/yuxi120407/INSURE .
Disentangling Alzheimer’s disease neurodegeneration from typical brain ageing using machine learning
Abstract Neuroimaging biomarkers that distinguish between changes due to typical brain ageing and Alzheimer’s disease are valuable for determining how much each contributes to cognitive decline. Supervised machine learning models can derive multivariate patterns of brain change related to the two processes, including the Spatial Patterns of Atrophy for Recognition of Alzheimer’s Disease (SPARE-AD) and of Brain Aging (SPARE-BA) scores investigated herein. However, the substantial overlap between brain regions affected in the two processes confounds measuring them independently. We present a methodology, and associated results, towards disentangling the two. T1-weighted MRI scans of 4054 participants (48–95 years) with Alzheimer’s disease, mild cognitive impairment (MCI), or cognitively normal (CN) diagnoses from the Imaging-based coordinate SysTem for AGIng and NeurodeGenerative diseases (iSTAGING) consortium were analysed. Multiple sets of SPARE scores were investigated, in order to probe imaging signatures of certain clinically or molecularly defined sub-cohorts. First, a subset of clinical Alzheimer’s disease patients (n = 718) and age- and sex-matched CN adults (n = 718) were selected based purely on clinical diagnoses to train SPARE-BA1 (regression of age using CN individuals) and SPARE-AD1 (classification of CN versus Alzheimer’s disease) models. Second, analogous groups were selected based on clinical and molecular markers to train SPARE-BA2 and SPARE-AD2 models: amyloid-positive Alzheimer’s disease continuum group (n = 718; consisting of amyloid-positive Alzheimer’s disease, amyloid-positive MCI, amyloid- and tau-positive CN individuals) and amyloid-negative CN group (n = 718). Finally, the combined group of the Alzheimer’s disease continuum and amyloid-negative CN individuals was used to train SPARE-BA3 model, with the intention to estimate brain age regardless of Alzheimer’s disease-related brain changes. The disentangled SPARE models, SPARE-AD2 and SPARE-BA3, derived brain patterns that were more specific to the two types of brain changes. The correlation between the SPARE-BA Gap (SPARE-BA minus chronological age) and SPARE-AD was significantly reduced after the decoupling (r = 0.56–0.06). The correlation of disentangled SPARE-AD was non-inferior to amyloid- and tau-related measurements and to the number of APOE ε4 alleles but was lower to Alzheimer’s disease-related psychometric test scores, suggesting the contribution of advanced brain ageing to the latter. The disentangled SPARE-BA was consistently less correlated with Alzheimer’s disease-related clinical, molecular and genetic variables. By employing conservative molecular diagnoses and introducing Alzheimer’s disease continuum cases to the SPARE-BA model training, we achieved more dissociable neuroanatomical biomarkers of typical brain ageing and Alzheimer’s disease.
Interactively Assessing Disentanglement in GANs
Abstract Generative adversarial networks (GAN) have witnessed tremendous growth in recent years, demonstrating wide applicability in many domains. However, GANs remain notoriously difficult for people to interpret, particularly for modern GANs capable of generating photo‐realistic imagery. In this work we contribute a visual analytics approach for GAN interpretability, where we focus on the analysis and visualization of GAN disentanglement. Disentanglement is concerned with the ability to control content produced by a GAN along a small number of distinct, yet semantic, factors of variation. The goal of our approach is to shed insight on GAN disentanglement, above and beyond coarse summaries, instead permitting a deeper analysis of the data distribution modeled by a GAN. Our visualization allows one to assess a single factor of variation in terms of groupings and trends in the data distribution, where our analysis seeks to relate the learned representation space of GANs with attribute‐based semantic scoring of images produced by GANs. Through use‐cases, we show that our visualization is effective in assessing disentanglement, allowing one to quickly recognize a factor of variation and its overall quality. In addition, we show how our approach can highlight potential dataset biases learned by GANs.
Disentangling ( 2 + 1 ) D topological states of matter with entanglement negativity
We use the entanglement negativity, a bipartite measure of entanglement in mixed quantum states, to study how multipartite entanglement constrains the real-space structure of the ground state wavefunctions of (2 + 1)-dimensional topological phases. We focus on the (Abelian) Laughlin and (non-Abelian) Moore-Read states at filling fraction ν = 1/m. We show that a combination of entanglement negativities, calculated with respect to specific cylinder and torus geometries, determines a necessary condition for when a topological state can be disentangled, i.e., factorized into a tensor product of states defined on cylinder subregions. This condition, which requires the ground state to lie in a definite topological sector, is sufficient for the Laughlin state. On the other hand, we find that a general Moore-Read ground state cannot be disentangled even when the disentangling condition holds.
Fast tensor disentangling algorithm
Many recent tensor network algorithms apply unitary operators to parts of a tensor network in order to reduce entanglement. However, many of the previously used iterative algorithms to minimize entanglement can be slow. We introduce an approximate, fast, and simple algorithm to optimize disentangling unitary tensors. Our algorithm is asymptotically faster than previous iterative algorithms and often results in a residual entanglement entropy that is within 10 to 40% of the minimum. For certain input tensors, our algorithm returns an optimal solution. When disentangling order-4 tensors with equal bond dimensions, our algorithm achieves an entanglement spectrum where nearly half of the singular values are zero. We further validate our algorithm by showing that it can efficiently disentangle random 1D states of qubits.
Enhancing Interpretability in Generative Modeling: Statistically Disentangled Latent Spaces Guided by Generative Factors in Scientific Datasets
This study addresses the challenge of statistically extracting generative factors from complex, high-dimensional datasets in unsupervised or semi-supervised settings. We investigate encoder-decoder-based generative models for nonlinear dimensionality reduction, focusing on disentangling low-dimensional latent variables corresponding to independent physical factors. Introducing Aux-VAE, a novel architecture within the classical Variational Autoencoder framework, we achieve disentanglement with minimal modifications to the standard VAE loss function by leveraging prior statistical knowledge through auxiliary variables. These variables guide the shaping of the latent space by aligning latent factors with learned auxiliary variables. We validate the efficacy of Aux-VAE through comparative assessments on multiple datasets, including astronomical simulations.
Symmetric-asymmetric collision comparison: Disentangling nuclear structure and subnucleonic structure effects for small system flow
Previous flow measurements in small collision systems were mostly based on highly asymmetric collisions (p+Pb, p+Au, d+Au, 3 He+Au), where both nuclear structure and subnucleonic fluctuations are important. Comparing these asymmetric systems with the newly available symmetric 16 O+ 16 O collisions at RHIC and LHC provides a unique opportunity to disentangle these two contributions. Using Glauber models incorporating both nucleon and quark-level substructure, we analyze multiplicity distributions and initial-state estimators: eccentricities ε n for anisotropic flow ν n and inverse transverse size d ⟂ for radial flow. We find that subnucleonic fluctuations impact O+O collisions differently from asymmetric systems, creating specific patterns in flow observables that enable disentangling the competing contributions. Such experimental comparisons will reduce uncertainties in the initial conditions and improve our understanding of the properties of the QGP-like medium produced in small systems.
Disentangling Rotational Dynamics and Ordering Transitions in a System of Self-Organizing Protein Nanorods via Rotationally Invariant Latent Representations
The dynamics of complex ordering systems with active rotational degrees of freedom exemplified by protein self-assembly is explored using a machine learning workflow that combines deep learning-based semantic segmentation and rotationally invariant variational autoencoder-based analysis of orientation and shape evolution. The latter allows for disentanglement of the particle orientation from other degrees of freedom and compensates for lateral shifts. The disentangled representations in the latent space encode the rich spectrum of local transitions that can now be visualized and explored via continuous variables. The time dependence of ensemble averages allows insight into the time dynamics of the system and, in particular, illustrates the presence of the potential ordering transition. Finally, analysis of the latent variables along the single-particle trajectory allows tracing these parameters on a single-particle level. The proposed approach is expected to be universally applicable for the description of the imaging data in optical, scanning probe, and electron microscopy seeking to understand the dynamics of complex systems where rotations are a significant part of the process.
Disentangling quantum matter with measurements
Measurements destroy entanglement. Building on ideas used to study ‘quantum disentangled liquids,’ we explore the use of this effect to characterize states of matter. In this work, we focus on systems with multiple components, such as charge and spin in a Hubbard model or local moments and conduction electrons in a Kondo lattice model. In such systems, measurements of (a subset of) one of the components can leave behind a quantum state of the other that is easy to understand, for example in terms of scaling of entanglement entropy of subregions. We bound the outcome of this protocol, for any choice of measurement, in terms of more standard information-theoretic quantities. We apply this quantum disentangling protocol to several problems of physical interest, including gapless topological phases, heavy fermions, and scar states in the Hubbard model.
Disentangling contributions to small-system collectivity via scans of light nucleus-nucleus collisions
The observation of multiparticle azimuthal correlations in high-energy small-system collisions has led to intense debate on its origin and the possible coexistence from two competing theoretical scenarios: one based on initial-state intrinsic momentum anisotropy, and the other based on final-state interaction model (FSM) collective response to the collision geometry. To complement the previous scan of asymmetric collision systems ( p +Au, d +Au, and He+Au), we propose a scan of small symmetric collision systems at the Brookhaven National Laboratory Relativistic Heavy Ion Collider, such as C+C, O+O, Al+Al, and Ar+Ar$\sqrt{s_\text{NN}}$ = 0.2 TeV , to provide further insights in disentangling contributions from these two scenarios. These symmetric small systems have the advantage of providing a better controlled initial geometry dominated by the average shape of the overlap region as opposed to fluctuation-driven geometries in asymmetric systems. In this work, a transport model is employed to investigate the expected geometry response in the FSM scenario. Different trends of elliptic flow with increasing charge particle multiplicity are observed between symmetric and asymmetric systems, whereas triangular flow appears to show a similar behavior. Furthermore, a comparison of O+O collisions at $\sqrt{s_\text{NN}}$ = 0.2 TeV and at $\sqrt{s_\text{NN}}$ = 2.76 ₋ 7 TeV as proposed at the CERN Large Hadron Collider, provides a unique opportunity to disentangle the collision geometry effects at the nucleon level from those arising from subnucleon fluctuations.
CAD-VAE: Leveraging Correlation-Aware Latents for Comprehensive Fair Disentanglement
While deep generative models have significantly advanced representation learning, they may inherit or amplify biases and fairness issues by encoding sensitive attributes alongside predictive features. Enforcing strict independence in disentanglement is often unrealistic when target and sensitive factors are naturally correlated. To address this challenge, we propose CAD-VAE(Correlation-Aware Disentangled VAE), which introduces a correlated latent code to capture the information shared between the target and sensitive attributes. Given this correlated latent, our method effectively separates over-lapping factors without extra domain knowledge by directly minimizing the conditional mutual information between target and sensitive codes. A relevance-driven optimization strategy refines the correlated code by efficiently capturing essential correlated features and eliminating redundancy. Extensive experiments on benchmark datasets demonstrate that CAD-VAE produces fairer representations, realistic counterfactuals, and improved fairness-aware image editing.
Towards interpretable Cryo-EM: disentangling latent spaces of molecular conformations
Molecules are essential building blocks of life and their different conformations (i.e., shapes) crucially determine the functional role that they play in living organisms. Cryogenic Electron Microscopy (cryo-EM) allows for acquisition of large image datasets of individual molecules. Recent advances in computational cryo-EM have made it possible to learn latent variable models of conformation landscapes. However, interpreting these latent spaces remains a challenge as their individual dimensions are often arbitrary. The key message of our work is that this interpretation challenge can be viewed as an Independent Component Analysis (ICA) problem where we seek models that have the property of identifiability. That means, they have an essentially unique solution, representing a conformational latent space that separates the different degrees of freedom a molecule is equipped with in nature. Thus, we aim to advance the computational field of cryo-EM beyond visualizations as we connect it with the theoretical framework of (nonlinear) ICA and discuss the need for identifiable models, improved metrics, and benchmarks. Moving forward, we propose future directions for enhancing the disentanglement of latent spaces in cryo-EM, refining evaluation metrics and exploring techniques that leverage physics-based decoders of biomolecular systems. Moreover, we discuss how future technological developments in time-resolved single particle imaging may enable the application of nonlinear ICA models that can discover the true conformation changes of molecules in nature. The pursuit of interpretable conformational latent spaces will empower researchers to unravel complex biological processes and facilitate targeted interventions. This has significant implications for drug discovery and structural biology more broadly. More generally, latent variable models are deployed widely across many scientific disciplines. Thus, the argument we present in this work has much broader applications in AI for science if we want to move from impressive nonlinear neural network models to mathematically grounded methods that can help us learn something new about nature.
Disentangling AGN and Star Formation in Soft X-Rays
We have explored the interplay of star formation and active galactic nucleus (AGN) activity in soft X-rays (0.5-2 keV) in two samples of Seyfert 2 galaxies (Sy2s). Using a combination of low-resolution CCD spectra from Chandra and XMM-Newton, we modeled the soft emission of 34 Sy2s using power-law and thermal models. For the 11 sources with high signal-to-noise Chandra imaging of the diffuse host galaxy emission, we estimate the luminosity due to star formation by removing the AGN, fitting the residual emission. The AGN and star formation contributions to the soft X-ray luminosity (i.e., L(sub x,AGN) and L(sub x,SF)) for the remaining 24 Sy2s were estimated from the power-law and thermal luminosities derived from spectral fitting. These luminosities were scaled based on a template derived from XSINGS analysis of normal star-forming galaxies. To account for errors in the luminosities derived from spectral fitting and the spread in the scaling factor, we estimated L(sub x,AGN) and L(sub x,SF))from Monte Carlo simulations. These simulated luminosities agree with L(sub x,AGN) and L(sub x,SF) derived from Chandra imaging analysis within a 3sigma confidence level. Using the infrared [Ne ii]12.8 micron and [O iv]26 micron lines as a proxy of star formation and AGN activity, respectively, we independently disentangle the contributions of these two processes to the total soft X-ray emission. This decomposition generally agrees with L(sub x,SF) and L(sub x,AGN) at the 3 sigma level. In the absence of resolvable nuclear emission, our decomposition method provides a reasonable estimate of emission due to star formation in galaxies hosting type 2 AGNs.
Disentangling Ferroelectric Wall Dynamics and Identification of Pinning Mechanisms via Deep Learning
Field-induced domain-wall dynamics in ferroelectric materials underpins multiple applications ranging from actuators to information technology devices and necessitates a quantitative description of the associated mechanisms including giant electromechanical couplings, controlled nonlinearities, or low coercive voltages. While the advances in dynamic piezoresponse force microscopy measurements over the last two decades have rendered visualization of polarization dynamics relatively straightforward, the associated insights into the local mechanisms have been elusive. This work explores the domain dynamics in model polycrystalline materials using a workflow combining deep-learning-based segmentation of the domain structures with nonlinear dimensionality reduction using multilayer rotationally invariant autoencoders (rVAE). The former allows unambiguous identification and classification of the ferroelectric and ferroelastic domain walls. Here, the rVAE discovers the latent representations of the domain wall geometries and their dynamics, consequently providing insight into the intrinsic mechanisms of polarization switching, that can further be compared to simple physical models. The rVAE disentangles the factors affecting the pinning efficiency of ferroelectric walls, offering insights into the correlation of ferroelastic wall distribution and ferroelectric wall pinning.
Disentangling observable dependence in SCET I and SCET II anomalous dimensions: angularities at two loops
The resummation of radiative corrections to collider jet observables using soft collinear effective theory is encoded in differential renormalization group equations (RGEs), with anomalous dimensions depending on the observable under consideration. This observable dependence arises from the ultraviolet (UV) singular structure of real phase space integrals in the effective field theory. We show that the observable dependence of anomalous dimensions in SCET I problems can be disentangled by introducing a suitable UV regulator in real radiation integrals. Resummation in the presence of the new regulator can be performed by solving a two-dimensional system of RGEs in the collinear and soft sectors, and resembles many features of resummation in SCET II theories by means of the rapidity renormalization group. We study the properties of SCET I with the additional regulator and explore the connection with the system of RGEs in SCET II theories, highlighting some universal patterns that can be exploited in perturbative calculations. As an application, we compute the two-loop soft and jet anomalous dimensions for a family of recoil-free angularities and give new analytic results. This allows us to study the relations between the SCET I and SCET II limits for these observables. We also discuss how the extra UV regulator can be exploited to calculate anomalous dimensions numerically, and the prospects for numerical resummation.
Disentangling error structures of precipitation datasets using decision trees
Characterizing error structures in precipitation products not only facilitates their proper applications for scientific and practical purposes but also helps improve their retrieval algorithms and processing methods. Despite the fact that multiple precipitation products have been assessed in the literature, factors that affect their error structures remain inadequately addressed. By interpreting 60 binary decision trees, this study disentangles the error characteristics of precipitation products in terms of their spatiotemporal patterns and geographical factors. Three independent precipitation products - two satellite-based and one reanalysis datasets: the Integrated Multi-satellitE Retrievals for GPM (Global Precipitation Measurement) late run (IMERG-L), Soil Moisture to Rain-Advanced SCATterometer (SM2RAIN-ASCAT), and the Modern-Era Retrospective analysis for Research and Applications, Version 2 uncorrected precipitation output (MERRA2-UC), are evaluated across the contiguous United States from 2010 to 2019. Here, the ground-based Stage IV precipitation dataset is used as the ground truth. Results indicate that the MERRA2-UC outperforms the IMERG-L and SM2RAIN-ASCAT with higher accuracy and more stable interannual patterns for the analysis period. Decision trees cross-assess three spatiotemporal factors and find that the underestimation of MERRA2-UC occurs in the east of the Rocky Mountains, and SM2RAIN-ASCAT underestimates precipitation over high latitudes, especially in winter. Additionally, the decision tree method ascribes system errors to nine different geographical characteristics, of which the distance to the coast, soil type, and DEM are the three dominant features. On the other hand, the land cover type, topography position index, and aspect are three relatively weak factors.
Rapid Crystallization of Zeolites with Controllable Defects: Disentangling Fluoride Concentration and pH Using NH 4 F
Zeolite synthesis is typically conducted either under basic conditions or in neutral fluoride media using hydrofluoric acid (HF). While basic (OH − ) conditions generally result in faster zeolite crystallization, they can also increase the likelihood of framework defects and crystal intergrowths. In contrast, synthesis in neutral fluoride media tends to produce fewer defects because fluoride balances positive charges from structure-directing agents. However, this method often requires significantly longer crystallization times and involves the handling of dangerous HF. In the present study, we pursue the best of both synthesis conditions, rapid syntheses with controllable defect concentrations, by disentangling of mineralizing agent and chargebalancing agent using ammonium fluoride (NH 4 F) as an alternative to HF. We have investigated the use of NH 4 F in the syntheses of siliceous and aluminum-containing zeolite A (LTA, small pore), ZSM-5 (MFI, medium pore), and siliceous Beta(*BEA, large pore). The crystallization times of all four zeolites decreased substantially with an increasing NH 4 F concentration. Crystallization times were reduced from 24 to 4 h (Si-LTA), 96 to 36 h (Al-LTA), 240 to 6 h (ZSM-5), and 24 to 3 h (Si-*BEA). Additionally, increasing the NH 4 F concentration in the synthesis mixtures decreases the defect densities of siliceous zeolites. Raman spectroscopy, along with 29 Si MAS NMR, 19 F MAS NMR, 13 C MAS NMR, and fluorine elemental analysis of Si-LTA samples confirms that the reduction in charged defects (Si−O − ) is due to the higher incorporation of F − within the double four-membered ring (D4R) present in the LTA samples. We show that the accelerated crystallization is due to the role of F − in enhancing the silica mineralization rate (formation of silicon hexafluoride species) and stabilizing D4Rs under basic conditions. As a result, combining basic and fluoride-mediated synthesis could therefore be advantageous for faster zeolite production and improved control over structural properties for a wide variety of zeolite structures.