Search NASA⌕ Search

SEARCH · Search NASA

Results for “combinatorial reasoning”

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 19 records
All sourcesNASA NTRSDOE OSTI
NASA NTRSDate not supplied

Combinatorial Reasoning: Selecting Reasons in Generative AI Pipelines via Combinatorial Optimization

Recent Large Language Models (LLMs) have demonstrated impressive capabilities at tasks that require human intelligence and are a significant step towards human-like artificial intelligence (AI). Yet the performance of LLMs at reasoning tasks have been subpar and the reasoning capability of LLMs is a matter of significant debate. While it has been shown that the choice of the prompting technique to the LLM can alter its performance on a multitude of tasks, including reasoning, the best performing techniques require human-made prompts with the knowledge of the tasks at hand. We introduce a framework for what we call Combinatorial Reasoning (CR), a fully-automated prompting method, where reasons are sampled from an LLM pipeline and mapped into a Quadratic Unconstrained Binary Optimization (QUBO) problem. The framework investigates whether QUBO solutions can be profitably used to select a useful subset of the reasons to construct a Chain-of-Thought style prompt. We explore the acceleration of CR with specialized solvers. We also investigate the performance of simpler zero-shot strategies such as linear majority rule or random selection of reasons. Our preliminary study indicates that coupling a combinatorial solver to generative AI pipelines is an interesting avenue for AI reasoning and elucidates design principles for future CR methods.

combinatorial reasoning↗
NASA NTRS1992

Uncertainty management by relaxation of conflicting constraints in production process scheduling

Mathematical-analytical methods as used in Operations Research approaches are often insufficient for scheduling problems. This is due to three reasons: the combinatorial complexity of the search space, conflicting objectives for production optimization, and the uncertainty in the production process. Knowledge-based techniques, especially approximate reasoning and constraint relaxation, are promising ways to overcome these problems. A case study from an industrial CIM environment, namely high-grade steel production, is presented to demonstrate how knowledge-based scheduling with the desired capabilities could work. By using fuzzy set theory, the applied knowledge representation technique covers the uncertainty inherent in the problem domain. Based on this knowledge representation, a classification of jobs according to their importance is defined which is then used for the straightforward generation of a schedule. A control strategy which comprises organizational, spatial, temporal, and chemical constraints is introduced. The strategy supports the dynamic relaxation of conflicting constraints in order to improve tentative schedules.

Dorn, Juergen↗
DOE OSTI2023

Monte Carlo Thought Search: Large Language Model Querying for Complex Scientific Reasoning in Catalyst Design

Discovering novel catalysts requires complex reasoning involving multiple chemical properties and resultant trade-offs, leading to a combinatorial growth in the search space. While large language models (LLM) have demonstrated novel capabilities for chemistry through complex instruction following capabilities and high quality reasoning, a goal-driven combinatorial search using LLMs has not been explored in detail. In this work, we present a Monte Carlo Tree Search-based approach that improves beyond state-of-the-art chain-of-thought prompting variants to augment scientific reasoning. We introduce two new reasoning datasets: 1) a curation of computational chemistry simulations, and 2) diverse questions written by catalysis researchers for reasoning about novel chemical conversion processes. We improve over the best baseline by 25.8\% and find that our approach can augment scientist's reasoning and discovery process with novel insights.\footnote{All resources will be publicly available upon publication.

Sprueill, Henry W.↗
DOE OSTI2022

A differentiable approach to the maximum independent set problem using dataless neural networks

The success of machine learning solutions for reasoning about discrete structures has brought attention to its adoption within combinatorial optimization algorithms. Such approaches generally rely on supervised learning by leveraging datasets of the combinatorial structures of interest drawn from some distribution of problem instances. Reinforcement learning has also been employed to find such structures. Here, in this paper, we propose a different approach in that no data is required for training the neural networks that produce the solution. In this sense, what we present is not a machine learning solution, but rather one that is dependent on neural networks and where backpropagation is applied to a loss function defined by the structure of the neural network architecture as opposed to a training dataset. In particular, we reduce the popular combinatorial optimization problem of finding a maximum independent set to a neural network and employ a dataless training scheme to refine the parameters of the network such that those parameters yield the structure of interest. Additionally, we propose a universal graph reduction procedure to handle large-scale graphs. The reduction exploits community detection for graph partitioning and is applicable to any graph type and/or density. Experimental results on both real and synthetic graphs demonstrate that our proposed method performs on par or outperforms state-of-the-art learning-based methods in terms of the size of the found set without requiring any training data.

97 MATHEMATICS AND COMPUTING↗
NASA NTRSDate not supplied

Runway Configuration Management with Offline Reinforcement Learning

Runway configuration management (RCM) is a challenging task, and it affects the efficiency of the National Airspace System (NAS) and airport surface operations significantly. Each airport, depending on the geometry, capacity, local climate patterns, etc. has multiple configurations for the runway usage for arriving and departing flights. Many factors such as the incoming/outgoing traffic load, wind direction and speed, convective weather, cloud ceiling and other environmental factors might affect the choice of a runway configuration at any point in time. However, other factors such as safety measures and regulations, noise abatement, capacity of each configuration, and preference of the air traffic controllers (ATCs) can also play a significant role in selecting the configuration. A sub-optimal selection of the runway configuration, or delay in making configuration changes might result in significant increase in taxi times for aircraft on the surface of the airport, fuel and energy use of the aircraft, and maintenance costs. It can also lead to safety concerns, such as an aircraft performing one or more go-arounds before being able to land. All these factors make RCM an extremely important and challenging decision-making process for the ATCs. The current state of practice sets the runway configuration by the ATCs based on relevant information available at the time including weather, traffic, noise abatement, safety bounds, etc. This makes the decision-making process subjective based on the accuracy of the available information and the bias in human decision making. Unfortunately, this approach yields poor results (e.g., significant delays) if the predicted outcomes are uncertain and their relative impact is not well understood. This is especially evident when the uncertainty increases the size of possible predicted outcomes (combinatorial explosion in possible scenarios) that cannot be handled by human reasoning. On the other hand, an automated approach based on machine intelligence can make use of historical data and search through all (or significant amount of) possible scenarios under uncertainty and make well-informed decisions.

Milad Memarzadeh↗
DOE OSTI2024

A Case Study of Multimodal, Multi-institutional Data Management for the Combinatorial Materials Science Community

Although the convergence of high-performance computing, automation, and machine learning has significantly altered the materials design timeline, transformative advances in functional materials and acceleration of their design will require addressing the deficiencies that currently exist in materials informatics, particularly a lack of standardized experimental data management. The challenges associated with experimental data management are especially true for combinatorial materials science, where advancements in automation of experimental workflows have produced datasets that are often too large and too complex for human reasoning. The data management challenge is further compounded by the multimodal and multi-institutional nature of these datasets, as they tend to be distributed across multiple institutions and can vary substantially in format, size, and content. Furthermore, modern materials engineering requires the tuning of not only composition but also of phase and microstructure to elucidate processing–structure–property–performance relationships. To adequately map a materials design space from such datasets, an ideal materials data infrastructure would contain data and metadata describing (i) synthesis and processing conditions, (ii) characterization results, and (iii) property and performance measurements. In this work, we present a case study for the low-barrier development of such a dashboard that enables standardized organization, analysis, and visualization of a large data lake consisting of combinatorial datasets of synthesis and processing conditions, X-ray diffraction patterns, and materials property measurements generated at several different institutions. While this dashboard was developed specifically for data-driven thermoelectric materials discovery, we envision the adaptation of this prototype to other materials applications, and, more ambitiously, future integration into an all-encompassing materials data management infrastructure.

36 MATERIALS SCIENCE↗
NASA NTRS2020

Scheduling the Mapping of a planet under geometrical constraints

Scheduling the coverage of a planet by scientific instruments on board spacecrafts under observational constraints is central in a significant number of current and future missions for the exploration of the solar system. In this paper, we describe the components and algorithms of a software used to study early-phase mission design or to schedule daily operations of currently in-flight spacecraft. The scheduling problem at hand is usually a large combinatorial problem. A discretization process is described and several ordering algorithms are designed and compared. Experiments show that high-quality schedules can be produced by approaches combining reasoning about rolling, coverage and priorities.

Wells, Christopher↗
DOE OSTI2021

Transformative Materials for High-Efficiency Thermochemical Production of Solar Fuels (Final Report)

Metal oxide-based two-step solar thermochemical (STC) H 2 O and CO 2 -splitting cycles are a promising route to convert solar thermal energy into fuels. The metal oxide materials are reduced at high temperatures (Step 1), and then at low (but still elevated) temperatures, the reduced oxide is used to split H 2 O or CO 2 (Step 2). However, current applications of these cycles are limited by the efficiency of the metal oxide materials. A lower temperature for reduction is desirable, but that brings a concomitant reduction in the driving force for gas splitting. So, designing novel, high-efficiency materials is challenging. Here, we devised and performed a joint computational-experimental project, combined with materials design strategies and high-throughput approaches with the goal to quickly discover and demonstrate novel thermochemical materials with superior properties. Our approach involves an active feedback loop between experimental and computational research. We have use our previously developed materials design map, to efficiently screen results from high-throughput first- principles computation to predict new compositions and subsequently to experimentally study the properties of novel, predicted materials. Experimental measurements of the thermodynamic properties of selected materials serve as a critical benchmark, not only in evaluating the STCH properties of synthesized compounds, but also serving as a validation dataset for computational approaches. We experimentally explored a set of predicted ABO3 perovskites and focus on obtaining high quality thermodynamic properties for validation of computational prediction of enthalpy and entropy of reduction. These thermodynamic quantities play a major role in designing materials with reduced temperatures of reduction but sufficient gas-splitting rates. Finding good validation between experimental and computational data for enthalpies of reduction, we turn to promising ABO3 materials modified by A and B site substitutions, opening an enormous combinatorial space of materials. We use our high-throughput approach to “tune in” the desired solar thermochemical (STC) properties for STC. This vast composition space can only be reasonably explored using the high-throughput approaches, both computational and experimental, of this proposal. Several promising novel oxide materials have been predicted, synthesized, and verified using our materials design approach.

08 HYDROGEN↗
NASA NTRS2002

Overall Traveling-Wave-Tube Efficiency Improved By Optimized Multistage Depressed Collector Design

Depressed Collector Design The microwave traveling wave tube (TWT) is used widely for space communications and high-power airborne transmitting sources. One of the most important features in designing a TWT is overall efficiency. Yet, overall TWT efficiency is strongly dependent on the efficiency of the electron beam collector, particularly for high values of collector efficiency. For these reasons, the NASA Glenn Research Center developed an optimization algorithm based on simulated annealing to quickly design highly efficient multistage depressed collectors (MDC's). Simulated annealing is a strategy for solving highly nonlinear combinatorial optimization problems. Its major advantage over other methods is its ability to avoid becoming trapped in local minima. Simulated annealing is based on an analogy to statistical thermodynamics, specifically the physical process of annealing: heating a material to a temperature that permits many atomic rearrangements and then cooling it carefully and slowly, until it freezes into a strong, minimum-energy crystalline structure. This minimum energy crystal corresponds to the optimal solution of a mathematical optimization problem. The TWT used as a baseline for optimization was the 32-GHz, 10-W, helical TWT developed for the Cassini mission to Saturn. The method of collector analysis and design used was a 2-1/2-dimensional computational procedure that employs two types of codes, a large signal analysis code and an electron trajectory code. The large signal analysis code produces the spatial, energetic, and temporal distributions of the spent beam entering the MDC. An electron trajectory code uses the resultant data to perform the actual collector analysis. The MDC was optimized for maximum MDC efficiency and minimum final kinetic energy of all collected electrons (to reduce heat transfer). The preceding figure shows the geometric and electrical configuration of an optimized collector with an efficiency of 93.8 percent. The results show the improvement in collector efficiency from 89.7 to 93.8 percent, resulting in an increase of three overall efficiency points. In addition, the time to design a highly efficient MDC was reduced from a month to a few days. All work was done in-house at Glenn for the High Rate Data Delivery Program. Future plans include optimizing the MDC and TWT interaction circuit in tandem to further improve overall TWT efficiency.

Vaden, Karl R.↗
DOE OSTI2026

CDRL: Certification-Driven Reinforcement Learning for Neutrino Flavor Model Discovery

Many scientific discovery problems require searching combinatorial hypothesis spaces under complex domain constraints. Reinforcement learning (RL) offers a promising approach, but existing methods rely on scalar rewards that provide limited information about why candidate solutions fail, leading agents to repeatedly explore invalid regions. We introduce Certification-Driven Reinforcement Learning (CDRL), a framework that leverages structured feedback from symbolic reasoning tools. When a candidate violates domain constraints, these tools produce certificates identifying the actions responsible for failure. CDRL converts these certificates into reusable constraints that eliminate classes of invalid solutions and guide exploration toward valid regions. We evaluate CDRL on neutrino flavor model discovery in theoretical particle physics, where the hypothesis space exceeds $10^{26}$ possible models, and compare it with the state-of-the-art RL approach previously used for this task. Across three theory spaces, CDRL achieves up to 1.95$\times$ higher valid model rates and up to 6.33$\times$ higher neutrino model rates while evaluating up to 4$\times$ fewer candidates. We further extract 40 interpretable rules from search trajectories using a post-hoc decision-tree framework and show that reusing them as soft constraints yields gains of up to 2$\times$ in valid model rates and 3$\times$ in neutrino model discovery across all three theory spaces. These results suggest that CDRL uncovers reusable structure in combinatorial search spaces and provides a general framework for scientific model discovery.

Jha, Piyush [Georgia Tech., Atlanta; Georgia Tech]↗
NASA NTRS2011

Optimizing Muscle Parameters in Musculoskeletal Modeling Using Monte Carlo Simulations

Astronauts assigned to long-duration missions experience bone and muscle atrophy in the lower limbs. The use of musculoskeletal simulation software has become a useful tool for modeling joint and muscle forces during human activity in reduced gravity as access to direct experimentation is limited. Knowledge of muscle and joint loads can better inform the design of exercise protocols and exercise countermeasure equipment. In this study, the LifeModeler(TM) (San Clemente, CA) biomechanics simulation software was used to model a squat exercise. The initial model using default parameters yielded physiologically reasonable hip-joint forces. However, no activation was predicted in some large muscles such as rectus femoris, which have been shown to be active in 1-g performance of the activity. Parametric testing was conducted using Monte Carlo methods and combinatorial reduction to find a muscle parameter set that more closely matched physiologically observed activation patterns during the squat exercise. Peak hip joint force using the default parameters was 2.96 times body weight (BW) and increased to 3.21 BW in an optimized, feature-selected test case. The rectus femoris was predicted to peak at 60.1% activation following muscle recruitment optimization, compared to 19.2% activation with default parameters. These results indicate the critical role that muscle parameters play in joint force estimation and the need for exploration of the solution space to achieve physiologically realistic muscle activation.

Hanson, Andrea↗
DOE OSTI2023

Computational Investigation of MAX as Intercalation Host for Rechargeable Aluminum‐Ion Battery

Abstract Layered carbides and their analogs with MAX phase (general formula AM n+1 X n ) have emerged as promising candidates for energy storage and conversion applications. One frontier for energy storage is using MAX as an Al‐ion intercalation electrode. Given that many MAXs have Al as the A sites, the structure can potentially serve as a stable host for Al intercalation. Here in this work, 425 ternary MAX Al‐ion battery electrodes are computationally enumerated. Specifically, first principal phase diagram calculations are performed on the combinatorial space of 17 types of typical transition metals, five types of anions (C, N, B, Si, and P), three types of stoichiometries (n = 1, 2, and 3) and two types of layered stackings (α and β). Among all the ternary MAX materials, 44 candidates show reasonable synthetic accessibility, and six with extraordinary performance are predicted to be promising Al‐ion battery electrodes. With the phase stability, and electrochemical performance (average voltage, theoretical capacity, energy density, and Al diffusion barrier), the work provides a comprehensive computational assessment of the great opportunities behind MAX‐based Al‐ion batteries.

Chemistry↗
DOE OSTI2026

Cross-Domain Reasoning for Neuromorphic Model Design

Designing performant neuromorphic models requires reasoning across neuroscience, neuromorphic computing, and machine learning, making it a natural target for cross-domain hypothesis generation. Our primary contribution is a multi-corpus knowledge graph spanning all three domains, which we show substantially increases cross-domain retrieval novelty over single-corpus baselines. We additionally introduce NeuKReAct, an agentic reasoning framework that iteratively retrieves from this graph and synthesizes design hypotheses via a step-by-step blackboard architecture, enabling structured compartmentalization of design decisions. Lastly, we introduce an execution head that translates hypotheses into structured design documents and runnable code. We evaluate novelty using a combinatorial creativity metric that measures cross-domain retrieval distance across the citation graph. Our results confirm that corpus breadth is the dominant driver of novelty. Moreover, we highlight a concrete instance of the novelty-utility tradeoff within NeuKReAct, underscoring a need for joint creativity evaluation, balancing both novelty and utility.

Ramavarapu, Vikram [ORNL] (ORCID:0009000188757213)↗
DOE OSTI2025

Comparing three generations of D-Wave quantum annealers for minor embedded combinatorial optimization problems

Abstract Quantum annealing (QA) is a novel type of analog computation that aims to use quantum mechanical fluctuations to search for optimal solutions of Ising problems. QA in the transverse Ising model, implemented on D-Wave quantum processing units, are available as cloud computing resources. In this study we report concise benchmarks across three generations of D-Wave quantum annealers, consisting of four different devices, for the NP-hard discrete combinatorial optimization problems unweighted maximum clique and unweighted maximum cut on random graphs. The Ising, or equivalently quadratic unconstrained binary optimization, formulation of these problems do not require auxiliary variables for order reduction, and their overall structure and weights are not highly variable, which makes these problems simple test cases to understand the sampling capability of current D-Wave quantum annealers. All-to-all minor embeddings of size 52, with relatively uniform chain lengths, are used for a direct comparison across the Chimera, Pegasus, and Zephyr device topologies. A grid-search over annealing times and the minor embedding chain strengths is performed in order to determine the level of reasonable performance for each device and problem type. Experiment metrics that are reported are approximation ratios for non-broken chain samples, chain break proportions, and time-to-solution for the maximum clique problem instances. How fairly the quantum annealers sample optimal maximum cliques, for instances which contain multiple maximum cliques, is quantified using entropy of the measured ground state distributions. The newest generation of quantum annealing hardware, which has a Zephyr hardware connectivity, performed the best overall with respect to approximation ratios and chain break frequencies.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS2000

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↗
NASA NTRS2017

A Dose of Reality: Radiation Analysis for Realistic Human Spacecraft

INTRODUCTION As with most computational analyses, a tradeoff exists between problem complexity, resource availability and response accuracy when modeling radiation transport from the source to a detector. The largest amount of analyst time for setting up an analysis is often spent ensuring that any simplifications made have minimal impact on the results. The vehicle shield geometry of interest is typically simplified from the original CAD design in order to reduce computation time, but this simplification requires the analyst to "re-draw" the geometry with a limited set of volumes in order to accommodate a specific radiation transport software package. The resulting low-fidelity geometry model cannot be shared with or compared to other radiation transport software packages, and the process can be error prone with increased model complexity. The work presented here demonstrates the use of the DAGMC (Direct Accelerated Geometry for Monte Carlo) Toolkit from the University of Wisconsin, to model the impacts of several space radiation sources on a CAD drawing of the US Lab module. METHODS The DAGMC toolkit workflow begins with the export of an existing CAD geometry from the native CAD to the ACIS format. The ACIS format file is then cleaned using SpaceClaim to remove small holes and component overlaps. Metadata is then assigned to the cleaned geometry file using CUBIT/Trelis from csimsoft (Registered Trademark). The DAGMC plugin script removes duplicate shared surfaces, facets the geometry to a specified tolerance, and ensures that the faceted geometry is water tight. This step also writes the material and scoring information to a standard input file format that the analyst can alter as desired prior to running the radiation transport program. The scoring results can be transformed, via python script, into a 3D format that is viewable in a standard graphics program. RESULTS The CAD model of the US Lab module of the International Space Station, inclusive of all the racks and components, was simplified to remove holes and volume overlaps. Problematic features within the drawing were also removed or repaired to prevent runtime issues. The cleaned drawing was then run through the DAGMC workflow to prepare for analysis. Pilot tests modeling transport of 1GeV proton and 800MeV/A oxygen sources show that reasonable results are converged upon in an acceptable amount of overall computation time from drawing preparation to data analysis. The FLUKA radiation transport code will next be used to model both a GCR and a trapped radiation source. These results will then be compared with measurements that have been made by the radiation instrumentation deployed inside the US Lab module. DISCUSSION Early analyses have indicated that the DAGMC workflow is a promising toolkit for running vehicle geometries of interest to NASA through multiple radiation transport codes. In addition, recent work has shown that a realistic human phantom, provided via a subcontract with the University of Florida, can be placed inside any vehicle geometry for a combinatorial analysis. This added functionality gives the user the ability to score various parameters at the organ level, and the results can then be used as input for cancer risk models.

Barzilla, J. E.↗
NASA NTRS2000

Biological Moleculars: Have Most of Our Problems Already Been Solved?

Evolution has resulted in biological machinery that engineers have great reason to envy and at present can only poorly mimic. This is not just a curiosity as biological systems perform many functions that are desired industrial processes. Examples include photosynthesis, chemosynthesis, energy storage, low temperature chemical conversion, reproducible manufacture of chemical compounds, etc. The bases of biological machinery are the proteins and nucleic acids that comprise living organisms. Each molecule functions as a part of a biological machine. In many cases the molecule can be properly regarded as a stand alone machine of its own. Concepts and methods for harnessing the power of biological molecules exist but are often overlooked in the industrial world. Some are old and appear crude but are quite effective, e.g. the fermentation of grains and fruits. Currently, there is a revolution in progress regarding the harnessing biological processes. These include techniques such as genetic manipulation via polymerase chain reaction, forced evolution also known as evolution in a test tube, determination of molecular structure, and combinatorial chemistry. The following is a brief discussion on how these processes are performed and how they may relate to industrial and aerospace processes.

Downey, James P.↗
DOE OSTI2022

PWR loading pattern optimization with reinforcement learning

The core loading pattern optimization problem belongs to the class of combinatorial optimization problem and has been studied since the dawn of commercial nuclear energy industry. It is characterized by multiple objectives and constraints, with a very high number of candidate patterns, which makes it impossible to solve explicitly. Stochastic optimization methodologies including Genetic Algorithms and Simulated Annealing are used by different nuclear utilities and vendors to perform fuel cycle reload design. Nevertheless, hand-designed solutions continue to be the prevalent method in the industry. To improve the state-of-the-art core reload patterns, we aim to create a method as scalable as possible, that agrees with the designer's goal of performance and safety. To help in this task Deep Reinforcement Learning (DRL), in particular Proximal Policy Optimization is leveraged. DRL has recently experienced a strong impetus from its successes applied to games, sometimes even reaching 'super-human' performances. This paper lays out the foundation of this method and proposes to study the behavior of several hyper-parameters that influence the DRL algorithm. The algorithm is highly dependent on multiple factors such as an exploration/exploitation trade-off that manifests through different parameters such as the number of loading patterns seen and the number of samples collected before a policy update, but also the shape of the objective function derived for the core design. Experimental results also demonstrate the effectiveness of the method in finding high-quality solutions from scratch within a reasonable amount of time. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗