Search NASA⌕ Search

SEARCH · Search NASA

Results for “evolutionary computation”

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.

At least 433 records · Page 24

Evolutionary Development of the Simulation by Logical Modeling System (SIBYL)

Through the evolutionary development of the Simulation by Logical Modeling System (SIBYL) we have re-engineered the expensive and complex IBM mainframe based Long-term Hardware Projection Model (LHPM) to a robust cost-effective computer based mode that is easy to use. We achieved significant cost reductions and improved productivity in preparing long-term forecasts of Space Shuttle Main Engine (SSME) hardware. The LHPM for the SSME is a stochastic simulation model that projects the hardware requirements over 10 years. SIBYL is now the primary modeling tool for developing SSME logistics proposals and Program Operating Plan (POP) for NASA and divisional marketing studies.

Wu, Helen↗

HARMONY: Large-Scale Architecture Search for Efficient Hybrid Language Models

As large language models scale to trillions of parameters, their computational and memory requirements present critical challenges for efficient training and deployment. While Mixture of Experts (MoE) architectures enable efficient scaling through sparse parameter activation, and state-space models like Mamba offer linear-time complexity, principled methods for combining these paradigms remain undeveloped. We introduce HARMONY (Hybrid Architecture Research for Mamba, Optimized with Neural efficiencY), a multi-objective evolutionary neural architecture search framework for discovering efficient hybrid language models that integrate Transformer attention mechanisms, Mixture-of-Experts routing, and Mamba state-space components. Through large-scale distributed search using 16,384 MI250X GPUs on the Frontier supercomputer, HARMONY explores a comprehensive design space encompassing six attention variants (MHA, MQA, GQA, MLA, SWA, and Mamba-2), variable MoE configurations with both routed and shared experts, and extensive Mamba hyperparameters. Our framework discovers heterogeneous architectures that balance training performance with computational efficiency through multi-objective optimization incorporating latency penalties and fitness-based selection. Analysis of discovered architectures reveals that optimal hybrid designs favor heterogeneous component mixing rather than homogeneous patterns, with Mamba-2 and Multi-Head Latent Attention (MLA) emerging as preferred mechanisms. Discovered architectures demonstrate superior training efficiency: our best configuration achieves a final perplexity of 1.0874 with 2.38B parameters while processing 4,320 tokens/second, outperforming significantly larger manually designed models. Full-scale evaluation shows HARMONY's top architectures achieve better loss trajectories than equivalently-sized models using state-of-the-art configurations including Mixtral, Jamba, and Samba. Additionally, we demonstrate 91% weak scaling efficiency when training discovered 36B-parameter models across 1,024 GPUs. HARMONY is released as an open framework with comprehensive tools for building and training hybrid models using expert-data-pipeline parallelism, democratizing access to automated architecture design for next-generation language models.

Herron, Emily [ORNL] (ORCID:0000000273008172)↗

Microbial Communities in Microgravity: Simulation in Lab and on the Computer

Microorganisms grow differently in spaceflight than they do on Earth. While much remains unexplained about how microgravity affects microbial growth, one dominant hypothesis is that the lack of density-driven convection in the liquid growth environment makes mixing diffusion-limited, and therefore slower. This is supported by evidence that individual microbial strains experience starvation and acid stress in microgravity. However, if it is true, then microgravity would also have measurable effects on microbes in mixed communities, because many interspecies interactions involve the exchange of soluble metabolites through the medium (cross-feeding). Specifically, cooperative cross-feeding communities would grow more slowly in microgravity, and cooperation would be less stable on evolutionary timescales. Here we describe our efforts to test this hypothesis by simulating microgravity in silico and in the lab, using a model system of Eschericia coli and Salmonella enterica that grow only when they can exchange methionine and acetate. We created CAMDLES (CFD-DEM Artificial Microgravity Developments for Living Ecosystem Simulation) as an extension of CFDEM®coupling software, to carry out computational modeling of biological flows, growth, and mass transfer in microgravity and also in laboratory artificial microgravity devices (rotating wall vessels, RWV). Using CAMDLES, we found distinct differences in growth rates between RWV and true microgravity, and we were able to identify several features, such as spatial distribution, biofilm formation, and product yield parameters, that influence the degree to which RWV growth recapitulates microgravity growth. In addition, we report on the development of a laboratory system for monitoring growth rates and species ratios of the community in RWVs, using fluorescent strains. Pairing CAMDLES with the laboratory model system allows us to generate quantitative predictions about the effects of spaceflight on organisms that will be essential to sustaining human space exploration in the long term.

microbiology↗

Lazy Updating of hubs can enable more realistic models by speeding up stochastic simulations

To respect the nature of discrete parts in a system, stochastic simulation algorithms (SSAs) must update for each action (i) all part counts and (ii) each action's probability of occurring next and its timing. This makes it expensive to simulate biological networks with well-connected “hubs” such as ATP that affect many actions. Temperature and volume also affect many actions and may be changed significantly in small steps by the network itself during fever and cell growth, respectively. Such trends matter for evolutionary questions, as cell volume determines doubling times and fever may affect survival, both key traits for biological evolution. Yet simulations often ignore such trends and assume constant environments to avoid many costly probability updates. Such computational convenience precludes analyses of important aspects of evolution. Here we present “Lazy Updating,” an add-on for SSAs designed to reduce the cost of simulating hubs. When a hub changes, Lazy Updating postpones all probability updates for reactions depending on this hub, until a threshold is crossed. Speedup is substantial if most computing time is spent on such updates. We implemented Lazy Updating for the Sorting Direct Method and it is easily integrated into other SSAs such as Gillespie's Direct Method or the Next Reaction Method. Testing on several toy models and a cellular metabolism model showed >10× faster simulations for its use-cases—with a small loss of accuracy. Thus we see Lazy Updating as a valuable tool for some special but important simulation problems that are difficult to address efficiently otherwise.

97 MATHEMATICS AND COMPUTING↗

HYDRODYNAMIC MODELS OF TYPE I X-RAY BURSTS: METALLICITY EFFECTS

Type I X-ray bursts (XRBs) are thermonuclear stellar explosions driven by charged-particle reactions. In the regime for combined H/He-ignition, the main nuclear flow is dominated by the rp-process (rapid proton-captures and {beta}{sup +}-decays), the 3{alpha}-reaction, and the {alpha}p-process (a suite of ({alpha}, p) and (p, {gamma}) reactions). The main flow is expected to proceed away from the valley of stability, eventually reaching the proton drip line beyond A = 38. Detailed analysis of the relevant reactions along the main path has only been scarcely addressed, mainly in the context of parameterized one-zone models. In this paper, we present a detailed study of the nucleosynthesis and nuclear processes powering type I XRBs. The reported 11 bursts have been computed by means of a spherically symmetric (one-dimensional), Lagrangian, hydrodynamic code, linked to a nuclear reaction network that contains 325 isotopes (from {sup 1}H to {sup 107}Te), and 1392 nuclear processes. These evolutionary sequences, followed from the onset of accretion up to the explosion and expansion stages, have been performed for two different metallicities to explore the dependence between the extension of the main nuclear flow and the initial metal content. We carefully analyze the dominant reactions and the products of nucleosynthesis, together with the physical parameters that determine the light curve (including recurrence times, ratios between persistent and burst luminosities, or the extent of the envelope expansion). Results are in qualitative agreement with the observed properties of some well-studied bursting sources. Leakage from the predicted SbSnTe cycle cannot be discarded in some of our models. Production of {sup 12}C (and implications for the mechanism that powers superbursts), light p-nuclei, and the amount of H left over after the bursting episodes will also be discussed.

79 ASTRONOMY AND ASTROPHYSICS↗

Deep Reinforcement Learning Based Control of Wind Turbines for Fast Frequency Response

In order to fulfill vital auxiliary grid services, such as load regulation, spin and non-spin reserve provision, and frequency support during emergencies, there is often a requirement for certain wind farms to operate in de-loaded modes. Leveraging the swift response capabilities of wind farms, this study demonstrates that reserving power in de-loaded modes can significantly enhance power grid stability and reliability during system contingencies. Controlling wind farms optimally for frequency support is intricate due to the nonlinearity of models and controllers and the complexity of wind farm interactions with power systems. Here, to address this challenge, this paper introduces a novel approach that integrates wind turbines into reinforcement learning-based solutions for frequency response. This innovative methodology utilizes the state-of-the-art reinforcement learning algorithm known as the surrogate-gradient-based evolutionary strategy. The proposed learning-based algorithm provides continuous control of wind farm output to rapidly stabilize system frequency and prevent unnecessary trips of under-frequency load shedding relays. To facilitate efficient training, parallel computing techniques are employed. The proposed methodology is evaluated on a modified IEEE-39 bus system, and simulation results reveal its efficacy in reliably supporting power system frequency and preventing the need for unnecessary load shedding.

Gao, Wei [Argonne National Laboratory (ANL), Argon↗

JavaGenes: Evolving Graphs with Crossover

Genetic algorithms usually use string or tree representations. We have developed a novel crossover operator for a directed and undirected graph representation, and used this operator to evolve molecules and circuits. Unlike strings or trees, a single point in the representation cannot divide every possible graph into two parts, because graphs may contain cycles. Thus, the crossover operator is non-trivial. A steady-state, tournament selection genetic algorithm code (JavaGenes) was written to implement and test the graph crossover operator. All runs were executed by cycle-scavagging on networked workstations using the Condor batch processing system. The JavaGenes code has evolved pharmaceutical drug molecules and simple digital circuits. Results to date suggest that JavaGenes can evolve moderate sized drug molecules and very small circuits in reasonable time. The algorithm has greater difficulty with somewhat larger circuits, suggesting that directed graphs (circuits) are more difficult to evolve than undirected graphs (molecules), although necessary differences in the crossover operator may also explain the results. In principle, JavaGenes should be able to evolve other graph-representable systems, such as transportation networks, metabolic pathways, and computer networks. However, large graphs evolve significantly slower than smaller graphs, presumably because the space-of-all-graphs explodes combinatorially with graph size. Since the representation strongly affects genetic algorithm performance, adding graphs to the evolutionary programmer's bag-of-tricks should be beneficial. Also, since graph evolution operates directly on the phenotype, the genotype-phenotype translation step, common in genetic algorithm work, is eliminated.

Globus, Al↗

Inference of Transmission Network Structure from HIV Phylogenetic Trees

Phylogenetic inference is an attractive means to reconstruct transmission histories and epidemics. However, there is not a perfect correspondence between transmission history and virus phylogeny. Both node height and topological differences may occur, depending on the interaction between within-host evolutionary dynamics and between-host transmission patterns. To investigate these interactions, we added a within-host evolutionary model in epidemiological simulations and examined if the resulting phylogeny could recover different types of contact networks. To further improve realism, we also introduced patient-specific differences in infectivity across disease stages, and on the epidemic level we considered incomplete sampling and the age of the epidemic. Second, we implemented an inference method based on approximate Bayesian computation (ABC) to discriminate among three well-studied network models and jointly estimate both network parameters and key epidemiological quantities such as the infection rate. Our ABC framework used both topological and distance-based tree statistics for comparison between simulated and observed trees. Overall, our simulations showed that a virus time-scaled phylogeny (genealogy) may be substantially different from the between-host transmission tree. This has important implications for the interpretation of what a phylogeny reveals about the underlying epidemic contact network. In particular, we found that while the within-host evolutionary process obscures the transmission tree, the diversification process and infectivity dynamics also add discriminatory power to differentiate between different types of contact networks. We also found that the possibility to differentiate contact networks depends on how far an epidemic has progressed, where distance-based tree statistics have more power early in an epidemic. Finally, we applied our ABC inference on two different outbreaks from the Swedish HIV-1 epidemic.

60 APPLIED LIFE SCIENCES↗

A combinatorially complete epistatic fitness landscape in an enzyme active site

Protein engineering often targets binding pockets or active sites which are enriched in epistasis—nonadditive interactions between amino acid substitutions—and where the combined effects of multiple single substitutions are difficult to predict. Few existing sequence-fitness datasets capture epistasis at large scale, especially for enzyme catalysis, limiting the development and assessment of model-guided enzyme engineering approaches. We present here a combinatorially complete, 160,000-variant fitness landscape across four residues in the active site of an enzyme. Assaying the native reaction of a thermostable β-subunit of tryptophan synthase (TrpB) in a nonnative environment yielded a landscape characterized by significant epistasis and many local optima. These effects prevent simulated directed evolution approaches from efficiently reaching the global optimum. There is nonetheless wide variability in the effectiveness of different directed evolution approaches, which together provide experimental benchmarks for computational and machine learning workflows. The most-fit TrpB variants contain a substitution that is nearly absent in natural TrpB sequences—a result that conservation-based predictions would not capture. Thus, although fitness prediction using evolutionary data can enrich in more-active variants, these approaches struggle to identify and differentiate among the most-active variants, even for this near-native function. Overall, this work presents a large-scale testing ground for model-guided enzyme engineering and suggests that efficient navigation of epistatic fitness landscapes can be improved by advances in both machine learning and physical modeling.

biocatalysis↗

Betelgeuse as a Merger of a Massive Star with a Companion

We investigate the merger between a 16M ⊙ star, on its way to becoming a red supergiant (RSG), and a 4M ⊙ main-sequence companion. Our study employs three-dimensional hydrodynamic simulations using the state-of-the-art adaptive mesh refinement code Octo-Tiger. The initially corotating binary undergoes interaction and mass transfer, resulting in the accumulation of mass around the companion and its subsequent loss through the second Lagrangian point (L2). The companion eventually plunges into the envelope of the primary, leading to its spin-up and subsequent merger with the helium core. We examine the internal structural properties of the post-merger star, as well as the merger environment and the outflow driven by the merger. Our findings reveal the ejection of approximately ∼0.6 M ⊙ of material in an asymmetric and somewhat bipolar outflow. We import the post-merger stellar structure into the MESA stellar evolution code to model its long-term nuclear evolution. In certain cases, the post-merger star exhibits persistent rapid equatorial surface rotation as it evolves in the H–R diagram toward the observed location of Betelgeuse. These cases demonstrate surface rotation velocities of a similar magnitude to those observed in Betelgeuse, along with a chemical composition resembling that of Betelgeuse. In other cases, efficient rotationally induced mixing leads to slower surface rotation. This pioneering study aims to model stellar mergers across critical timescales, encompassing dynamical, thermal, and nuclear evolutionary stages.

79 ASTRONOMY AND ASTROPHYSICS↗

An Application Development Platform for Neuromorphic Computing

Dynamic Adaptive Neural Network Arrays (DANNAs) are neuromorphic computing systems developed as a hardware based approach to the implementation of neural networks. They feature highly adaptive and programmable structural elements, which model arti cial neural networks with spiking behavior. We design them to solve problems using evolutionary optimization. In this paper, we highlight the current hardware and software implementations of DANNA, including their features, functionalities and performance. We then describe the development of an Application Development Platform (ADP) to support efficient application implementation and testing of DANNA based solutions. We conclude with future directions.

Dean, Mark↗

How actin crosslinking and bundling proteins cooperate to generate an enhanced cell mechanical response

Actin-crosslinking proteins organize actin filaments into dynamic and complex subcellular scaffolds that orchestrate important mechanical functions, including cell motility and adhesion. Recent mutation studies have shown that individual crosslinking proteins often play seemingly non-essential roles, leading to the hypothesis that they have considerable redundancy in function. We report live-cell, in vitro, and theoretical studies testing the mechanical role of the two ubiquitous actin-crosslinking proteins, alpha-actinin and fascin, which co-localize to stress fibers and the basis of filopodia. Using live-cell particle tracking microrheology, we show that the addition of alpha-actinin and fascin elicits a cell mechanical response that is significantly greater than that originated by alpha-actinin or fascin alone. These live-cell measurements are supported by quantitative rheological measurements with reconstituted actin filament networks containing pure proteins that show that alpha-actinin and fascin can work in concert to generate enhanced cell stiffness. Computational simulations using finite element modeling qualitatively reproduce and explain the functional synergy of alpha-actinin and fascin. These findings highlight the cooperative activity of fascin and alpha-actinin and provide a strong rationale that an evolutionary advantage might be conferred by the cooperative action of multiple actin-crosslinking proteins with overlapping but non-identical biochemical properties. Thus the combination of structural proteins with similar function can provide the cell with unique properties that are required for biologically optimal responses.

Non-NASA Center↗

MODULES FOR EXPERIMENTS IN STELLAR ASTROPHYSICS (MESA)

Stellar physics and evolution calculations enable a broad range of research in astrophysics. Modules for Experiments in Stellar Astrophysics (MESA) is a suite of open source, robust, efficient, thread-safe libraries for a wide range of applications in computational stellar astrophysics. A one-dimensional stellar evolution module, MESAstar, combines many of the numerical and physics modules for simulations of a wide range of stellar evolution scenarios ranging from very low mass to massive stars, including advanced evolutionary phases. MESAstar solves the fully coupled structure and composition equations simultaneously. It uses adaptive mesh refinement and sophisticated timestep controls, and supports shared memory parallelism based on OpenMP. State-of-the-art modules provide equation of state, opacity, nuclear reaction rates, element diffusion data, and atmosphere boundary conditions. Each module is constructed as a separate Fortran 95 library with its own explicitly defined public interface to facilitate independent development. Several detailed examples indicate the extensive verification and testing that is continuously performed and demonstrate the wide range of capabilities that MESA possesses. These examples include evolutionary tracks of very low mass stars, brown dwarfs, and gas giant planets to very old ages; the complete evolutionary track of a 1 M {sub sun} star from the pre-main sequence (PMS) to a cooling white dwarf; the solar sound speed profile; the evolution of intermediate-mass stars through the He-core burning phase and thermal pulses on the He-shell burning asymptotic giant branch phase; the interior structure of slowly pulsating B Stars and Beta Cepheids; the complete evolutionary tracks of massive stars from the PMS to the onset of core collapse; mass transfer from stars undergoing Roche lobe overflow; and the evolution of helium accretion onto a neutron star. MESA can be downloaded from the project Web site (http://mesa.sourceforge.net/).

79 ASTRONOMY AND ASTROPHYSICS↗

Technology test results from an intelligent, free-flying robot for crew and equipment retrieval in space

The ground-based demonstrations of Extra Vehicular Activity (EVA) Retriever, a voice-supervised, intelligent, free-flying robot, are designed to evaluate the capability to retrieve objects (astronauts, equipment, and tools) which have accidentally separated from the Space Station. The EVA Retriever software is required to autonomously plan and execute a target rendezvous, grapple, and return to base while avoiding stationary and moving obstacles with subsequent object handover. The software architecture incorporates a heirarchical decomposition of the control system that is horizontally partitioned into five major functional subsystems: sensing, perception, world model, reasoning, and acting. The design provides for supervised autonomy as the primary mode of operation. It is intended to be an evolutionary system improving in capability over time and as it earns crew trust through reliable and safe operation. This paper gives an overview of the hardware, a focus on software, and a summary of results achieved recently from both computer simulations and air bearing floor demonstrations. Limitations of the technology used are evaluated. Plans for the next phase, during which moving targets and obstacles drive realtime behavior requirements, are discussed.

Erickson, J.↗

Functional phylogenomics analysis of bacteria and archaea using consistent genome annotation with UniFam

To supply some background, phylogenetic studies have provided detailed knowledge on the evolutionary mechanisms of genes and species in Bacteria and Archaea. However, the evolution of cellular functions, represented by metabolic pathways and biological processes, has not been systematically characterized. Many clades in the prokaryotic tree of life have now been covered by sequenced genomes in GenBank. This enables a large-scale functional phylogenomics study of many computationally inferred cellular functions across all sequenced prokaryotes. Our results show a total of 14,727 GenBank prokaryotic genomes were re-annotated using a new protein family database, UniFam, to obtain consistent functional annotations for accurate comparison. The functional profile of a genome was represented by the biological process Gene Ontology (GO) terms in its annotation. The GO term enrichment analysis differentiated the functional profiles between selected archaeal taxa. 706 prokaryotic metabolic pathways were inferred from these genomes using Pathway Tools and MetaCyc. The consistency between the distribution of metabolic pathways in the genomes and the phylogenetic tree of the genomes was measured using parsimony scores and retention indices. The ancestral functional profiles at the internal nodes of the phylogenetic tree were reconstructed to track the gains and losses of metabolic pathways in evolutionary history. In conclusion, our functional phylogenomics analysis shows divergent functional profiles of taxa and clades. Such function-phylogeny correlation stems from a set of clade-specific cellular functions with low parsimony scores. On the other hand, many cellular functions are sparsely dispersed across many clades with high parsimony scores. These different types of cellular functions have distinct evolutionary patterns reconstructed from the prokaryotic tree.

59 BASIC BIOLOGICAL SCIENCES↗

NGC 4656UV: A UV-SELECTED TIDAL DWARF GALAXY CANDIDATE

We report the discovery of a UV-bright tidal dwarf galaxy (TDG) candidate in the NGC 4631/4656 galaxy group, which we designate NGC 4656UV. Using survey and archival data spanning from 1.4 GHz to the ultraviolet, we investigate the gas kinematics and stellar properties of this system. The H I morphologies of NGC 4656UV and its parent galaxy NGC 4656 are extremely disturbed, with significant amounts of counterrotating and extraplanar gas. From UV-FIR photometry, computed using a new method to correct for surface gradients on faint objects, we find that NGC 4656UV has no significant dust opacity and a blue spectral energy distribution. We compute a star formation rate of 0.027 M{sub Sun} yr{sup -1} from the far-ultraviolet flux and measure a total H I mass of 3.8 Multiplication-Sign 10{sup 8} M{sub Sun} for the object. Evolutionary synthesis modeling indicates that NGC 4656UV is a low-metallicity system whose only major burst of star formation occurred within the last {approx}260-290 Myr. The age of the stellar population is consistent with a rough timescale for a recent tidal interaction between NGC 4656 and NGC 4631, although we discuss the true nature of the object-whether it is tidal or pre-existing in origin-in the context of its metallicity being a factor of 10 lower than its parent galaxy. We estimate that NGC 4656UV is either marginally bound or unbound. If bound, it contains relatively low amounts of dark matter. The abundance of archival data allows for a deeper investigation into this dynamic system than is currently possible for most TDG candidates.

79 ASTRONOMY AND ASTROPHYSICS↗

Classifier-Guided Sampling for Complex Energy System Optimization

This report documents the results of a Laboratory Directed Research and Development (LDRD) effort entitled “Classifier-Guided Sampling for Complex Energy System Optimization” that was conducted during FY 2014 and FY 2015. The goal of this project was to develop, implement, and test major improvements to the classifier-guided sampling (CGS) algorithm. CGS is type of evolutionary algorithm for performing search and optimization over a set of discrete design variables in the face of one or more objective functions. Existing evolutionary algorithms, such as genetic algorithms, may require a large number of objective function evaluations to identify optimal or near-optimal solutions. Reducing the number of evaluations can result in significant time savings, especially if the objective function is computationally expensive. CGS reduces the evaluation count by using a Bayesian network classifier to filter out non-promising candidate designs, prior to evaluation, based on their posterior probabilities. In this project, both the single-objective and multi-objective versions of the CGS are developed and tested on a set of benchmark problems. As a domain-specific case study, CGS is used to design a microgrid for use in islanded mode during an extended bulk power grid outage.

24 POWER TRANSMISSION AND DISTRIBUTION↗