Search NASA⌕ Search

SEARCH · Search NASA

Results for “computational cost”

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 325 records · Page 18
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Loop series expansions for tensor networks

Belief propagation (BP) can be a useful tool to approximately contract a tensor network, provided that the contributions from any closed loops in the network are sufficiently weak. In this article, we describe how a loop series expansion can be applied to systematically improve the accuracy of a BP approximation to a tensor network contraction, in principle converging arbitrarily close to the exact result. More generally, our result provides a framework for expanding a tensor network as a sum of component networks in a hierarchy of increasing complexity. We benchmark this proposal for the contraction of infinite projected entangled pair states, either representing the ground state of an Affleck-Kennedy-Lieb-Tasaki (AKLT) model or with randomly defined tensors, where it is shown to improve in accuracy over standard BP by several orders of magnitude while incurring only a minor increase in computational cost. These results indicate that the proposed series expansions could be a useful tool to accurately evaluate tensor networks in cases that otherwise exceed the limits of established contraction routines.

Evenbly, Glen [AWS Center for Quantum Computing, P↗
NASA NTRS2002

Efficient Reformulation of HOTFGM: Heat Conduction with Variable Thermal Conductivity

Functionally graded materials (FGMs) have become one of the major research topics in the mechanics of materials community during the past fifteen years. FGMs are heterogeneous materials, characterized by spatially variable microstructure, and thus spatially variable macroscopic properties, introduced to enhance material or structural performance. The spatially variable material properties make FGMs challenging to analyze. The review of the various techniques employed to analyze the thermodynamical response of FGMs reveals two distinct and fundamentally different computational strategies, called uncoupled macromechanical and coupled micromechanical approaches by some investigators. The uncoupled macromechanical approaches ignore the effect of microstructural gradation by employing specific spatial variations of material properties, which are either assumed or obtained by local homogenization, thereby resulting in erroneous results under certain circumstances. In contrast, the coupled approaches explicitly account for the micro-macrostructural interaction, albeit at a significantly higher computational cost. The higher-order theory for functionally graded materials (HOTFGM) developed by Aboudi et al. is representative of the coupled approach. However, despite its demonstrated utility in applications where micro-macrostructural coupling effects are important, the theory's full potential is yet to be realized because the original formulation of HOTFGM is computationally intensive. This, in turn, limits the size of problems that can be solved due to the large number of equations required to mimic realistic material microstructures. Therefore, a basis for an efficient reformulation of HOTFGM, referred to as user-friendly formulation, is developed herein, and subsequently employed in the construction of the efficient reformulation using the local/global conductivity matrix approach. In order to extend HOTFGM's range of applicability, spatially variable thermal conductivity capability at the local level is incorporated into the efficient reformulation. Analytical solutions to validate both the user-friendly and efficient reformulations am also developed. Volume discretization sensitivity and validation studies, as well as a practical application of the developed efficient reformulation are subsequently carried out. The presented results illustrate the accuracy and implementability of both the user-friendly formulation and the efficient reformulation of HOTFGM.

Zhong, Yi↗
NASA NTRS2026

Parametric-Based Heat Rejection Trade Study for Lunar and Martian Surface Operations

Establishing and maintaining a sustained presence on the lunar and/or Martian surfaces will require a diverse portfolio of surface elements (e.g., habitation, mobility, power generation, etc.). Many of these systems generate excess heat that must be rejected across a wide range of magnitudes, temperatures, and duty cycles and under variable environmental conditions. To identify the most promising heat rejection approaches for this diverse portfolio, a heat rejection trade study was conducted to evaluate the performance of different technology approaches across a spectrum of surface environments and heat-load requirements. The trade study consisted of three stages: (1) development of a parametric-based modeling framework, (2) creation of a database of heat rejection technologies, surface elements, and environmental conditions for the Moon and Mars, and (3) execution of a quantitative analysis of various heat rejection technologies across different operating conditions and surface elements. The modeling framework is developed in Python and Excel to prioritize small model size and hence low computational cost to enable large parametric sweeps while avoiding the reliance on proprietary software. Individual heat rejection processes are represented as simple Excel models, and a centralized Python script interfaces with the models to coordinate the parametric study. These simple sizing models were developed to take heat load requirements and environmental parameters as inputs and compute mass, power, and volume as outputs. Rather than assess each heat rejection technology separately for each surface element, a unified parametric space was developed to evaluate all technologies across all elements. This parametric space includes factors related to heat load (e.g., magnitude or temperature) and environment (e.g., surface temperature, sky temperature, solar flux). This effort generated a database containing information on over 60 heat rejection technologies and 30 surface elements. For each surface element, the expected heat rejection requirements were documented and analyzed to determine the most common needs shared across all elements. Environmental conditions at various lunar and Martian latitudes were also established for worst-case hot and worst-case cold scenarios. High-fidelity heat rejection models are currently under development. Preliminary trades between heat rejection technologies including radiators, venting technologies, convective coolers, and more have been conducted to identify promising options. This presentation will summarize the preliminary trade results and provide an overview and discussion of the expected heat loads and thermal environments for sustained surface operations on the Moon and Mars.

Heat Rejection↗
DOE OSTI2025

High Performance, High Fidelity: A GPU‐Accelerated Doubly‐Periodic Configuration of the Simple Cloud‐Resolving E3SM Atmosphere Model Version 1 (DP‐SCREAMv1)

The development of the Simplified Cloud Resolving Energy Exascale Earth System Atmosphere Model (SCREAMv1) enables global storm-resolving simulations on modern GPU-based supercomputers. However, the high computational cost of SCREAMv1 limits its routine use for process-level studies, creating a need for efficient proxy configurations. This study addresses this gap by introducing DP-SCREAMv1, a doubly periodic cloud-resolving model designed to be fully consistent with SCREAMv1 while enabling high-resolution, long-duration simulations at significantly reduced computational expense by simulating a limited doubly periodic domain rather than the entire globe. Built on a C++/Kokkos architecture, DP-SCREAMv1 achieves exceptional performance scalability on GPU systems and includes a rich library of cases for validation and scientific exploration. In this work, we demonstrate short wall-clock times at SCREAMv1's default resolution and show that DP-SCREAMv1 supports routine execution of large-domain, high-resolution experiments that were previously challenging in practice. Furthermore, we show that DP-SCREAMv1 enables routine execution of “Giga-LES” style simulations and facilitates large-domain, high-resolution simulations that were recently considered burdensome to perform. These results document an efficient, fully consistent process-level configuration for SCREAMv1 (DP-SCREAMv1) and illustrate its use for long-duration and large-domain experiments at cloud-resolving to eddy-permitting resolution.

Environmental sciences↗
NASA NTRSDate not supplied

DeepONet-Assisted Optimization of Surface Topography for Transition Delay in A Mach 4.5 Boundary Layer

We use deep learning, an ensemble variationaltechnique (EnVar), and direct numerical simulations(DNS) to design an optimal topography for a two-dimensional roughness element that delays the on-set of laminar-turbulent transition in a Mach 4.5 flat-plate boundary layer. Deep operator networks (Deep-ONets), which have the known ability to learn com-plex nonlinear operators within dynamical systems,are used for machine learning. For the baseline config-uration of a smooth flat plate, the second-mode wavesat the DNS inflow cause a quick nonlinear breakdownof the high-speed boundary layer within the computa-tional domain. Results reported in the present studyvalidate the ability of DeepONets to model the tran-sition delay via a given topography of the roughnesselement. The computing cost to optimize the rough-ness element for minimal skin-friction drag is substan-tially lowered by the DeepONets-based reduced-ordermodel. In comparison to the baseline method of EnVaroptimization based on DNS alone, the DeepONets-based EnVar optimizer is able to delay transition pastthe outflow boundary of the computational domainwhile utilizing almost 5–6 times fewer DNS.

Machine Learning↗
DOE OSTI2025

DS-TIDE: Harnessing Dynamical Systems for Efficient Time-Independent Differential Equation Solving

Time-Independent Differential Equations (TIDEs) are central to modeling equilibrium behavior across a wide range of scientific and engineering domains, from electrostatics to porous media flow. Conventional numerical solvers offer reliable solutions but incur significant computational costs due to fine-grained discretization and iterative procedures. Machine learning-based approaches address this by replacing iterative solving processes with one-time inference; however, their sophisticated models require extensive training resources that often exceed those of traditional solvers. Consequently, designing a TIDE solver that achieves high accuracy, broad applicability, and exceptional computational efficiency remains a fundamental challenge. In this paper, we propose DS-TIDE, a novel hardware solver that is inspired by, and subsequently leverages, the intrinsic connection between Dynamical Systems (DS) and Differential Equations (DEs) to efficiently and accurately solve TIDEs. DS-TIDE employs a CMOS-compatible DS-based processor, whose physical states evolve under carefully designed DE-driven dynamics and naturally converge to equilibrium -- the solution of the target TIDE -- within ~1µs on a ~1-watt DS-TIDE processor. To enhance expressivity, DS-TIDE incorporates Heterogeneous Dynamics with Temporal Layering (HDTL), which solves TIDEs through a three-stage DS evolution -- conditioning, solving, and decoding -- each governed by specialized dynamics. The entire evolution process is analogous to an infinitely deep neural network temporally unrolled, offering the system the capability of representing complex equations. Furthermore, DS-TIDE is equipped with an on-device DS-DE Auto-Alignment mechanism that dynamically adapts intrinsic hardware dynamics within milliseconds, effectively aligning the system’s dynamics to diverse target DEs. Experimental results across TIDEs from a wide range of scientific and engineering domains demonstrate that DS-TIDE achieves ~10^3× speedup, ~10^5× energy savings, and competitive or superior accuracy compared to state-of-the-art numerical and ML-based solvers.

Liu, Chuan↗
DOE OSTI2025

A dynamic 2D Borehole Thermal Energy Storage (BTES) model for enhanced computational efficiency

Progressing toward a future increasingly reliant on renewable energy sources, the development of effective, durable energy storage solutions becomes essential to balance supply and demand fluctuations. Borehole Thermal Energy Storage (BTES) is a long-duration thermal energy storage technology that captures excess heat generated from renewable energy sources and stores it underground for later use, enabling the efficient utilization of sustainable energy. This approach is particularly valuable in district energy networks when integrated with Ground Source Heat Pumps (GSHP) to provide stable heating and cooling. However, traditional three-dimensional (3D) numerical models of BTES systems demand extensive computational resources, limiting their practicality for real-time and large-scale applications. This study introduces a novel two-dimensional (2D) modeling approach that reduces computational costs while maintaining high accuracy. By employing a radial ring-based discretization method, the model simulates heat injection, retention, and retrieval dynamics over seasonal cycles. A new thermal-mass weighted-average temperature parameter is introduced to evaluate the performance of BTES systems. Model validation against FEFLOW simulations demonstrates a 17-fold improvement in computational speed compared to traditional Computational Fluid Dynamics (CFD) models while achieving a mean absolute percentage error (MAPE) of 2 % during charging and 4 % during discharging. Additionally, a trade-off analysis between computational efficiency and accuracy is conducted, ensuring the model's applicability for real-world scenarios. The findings of this research contribute to the development of computationally efficient BTES models, facilitating better optimization, control, and integration into renewable energy systems. This work provides a foundation for further studies in techno-economic analysis, multi-year performance evaluation, and real-time operational strategies for BTES applications, supporting a more sustainable energy future.

2D modeling↗
DOE OSTI2026

Stochastic GW -GPU: Rapid Quasi-Particle Energies for Molecules beyond 10,000 Atoms

StochasticGW is a code for computing accurate quasi-particle (QP) energies of molecules and material systems in the GW approximation. StochasticGW utilizes the stochastic Resolution of the Identity (sROI) technique to enable a massively parallel implementation with computational costs that scale semilinearly with system size, allowing the method to access systems with tens of thousands of electrons. Here, we introduce a new implementation, StochasticGW-GPU, for which the main bottleneck steps have been ported to GPUs and give substantial performance improvements over previous versions of the code. We showcase the new code by computing band gaps of hydrogenated silicon clusters (Si x H y ) containing up to 10,001 atoms and 35,144 electrons, and we obtain individual QP energies with a statistical precision of better than ±0.03 eV with times-to-solution of less than 1 h.

Thomas, Phillip S. [Lawrence Berkeley National Lab↗
NASA NTRS2018

Multifidelity Uncertainty Quantification of a Commercial Supersonic Transport

The objective of this work was to develop a multifidelity uncertainty quantification approach for efficient analysis of a commercial supersonic transport concept. An approach based on point-collocation, non-intrusive polynomial chaos was formulated in which a low-fidelity model could be corrected using multiple higher-fidelity models. The formulation and methodology also allows for the addition of uncertainty sources not present in the lower fidelity models. To demonstrate the applicability and potential computational savings of the multifidelity polynomial chaos approach, two model problems were explored. The first was a supersonic airfoil with three levels of modeling fidelity, each capturing a gradual increase in modeling of the underlying flow physics. As much as 50% computational cost reduction was observed using the mutlifidelity approach, while predicting nearly the same amount of uncertainty in drag. The second problem was a commercial supersonic transport. This model had three levels of fidelity that included two different modeling approaches and the addition of physics between the fidelity levels. Results of this analysis yielded nearly a 70% computational savings to predict a comparable amount of uncertainty in ground noise. Both problems illustrate the applicability and significant computational savings of the multifidelity method for efficient and accurate uncertainty quantification.

Thomas K. West IV↗
NASA NTRSDate not supplied

DeepONet-Assisted Optimization of Surface Topography for Transition Delay in a Mach 4.5 Boundary Layer

We use deep learning, an ensemble variational technique (EnVar), and direct numerical simulations(DNS) to design an optimal topography for a two-dimensional roughness element that delays the on-set of laminar-turbulent transition in a Mach 4.5 flat-plate boundary layer. Deep operator networks (DeepONets), which have the known ability to learn complex nonlinear operators within dynamical systems, are used for machine learning. For the baseline configuration of a smooth flat plate, the second-mode waves at the DNS inflow cause a quick nonlinear breakdown of the high-speed boundary layer within the computational domain. Results reported in the present study validate the ability of DeepONets to model the transition delay via a given topography of the roughness element. The computing cost to optimize the rough-ness element for minimal skin-friction drag is substantially lowered by the DeepONets-based reduced-order model. In comparison to the baseline method of EnVar optimization based on DNS alone, the DeepONets-based EnVar optimizer is able to delay transition past the outflow boundary of the computational domain while utilizing almost 5–6 times fewer DNS.

Machine Learning↗
DOE OSTI2024

Kilonova light-curve interpolation with neural networks

Kilonovae are the electromagnetic transients created by the radioactive decay of freshly synthesized elements in the environment surrounding a neutron star merger. To study the fundamental physics in these complex environments, kilonova modeling requires, in part, the use of radiative transfer simulations. The microphysics involved in these simulations results in high computational cost, prompting the use of emulators for parameter inference applications. Utilizing a training set of 22 248 high-fidelity simulations (composed of 412 unique ejecta parameter combinations evaluated at 54 viewing angles), we use a neural network to efficiently train on existing radiative transfer simulations and predict light curves for new parameters in a fast and computationally efficient manner. Our neural network can generate millions of new light curves in under a minute. We discuss our emulator's degree of off-sample reliability and parameter inference of the AT2017gfo observational data. Finally, we discuss tension introduced by multiband inference in the parameter inference results, particularly with regard to the neural network's recovery of viewing angle. Published by the American Physical Society 2024

79 ASTRONOMY AND ASTROPHYSICS↗
NASA NTRSDate not supplied

High-precision predictions of properties of chemically disordered crystals

Multiple scattering theory (MST) combined with density functional theory (DFT) allows to predict properties of chemically disordered materials from the first principles. However, such predictions often suffer from the systematic errors, which depend on crystal geometry. Each computed property of a particular crystal structure typically has a relatively small random error and a larger systematic error. Cancellation of systematic errors allows more accurate predictions. We propose a computational methodology based on the subtraction of the systematic errors in MST. To exemplify it, we apply it to the precipitated alloys. Considering precipitation strengthening in Ni superalloys, we compute the relative enthalpies of the competing Ni3(Al,Ti)1 crystal structures with a chemical disorder on the Al+Ti sublattice. Such predicted composition-structure-property dependencies are useful for the guided design of the next-generation alloys with improved strength. Our predictions are validated by comparison with the results of other DFT methods (having a higher computational cost) and with experiment. We acknowledge funding of the guided design of stronger superalloys for airspace by NASA Aeronautics Research Mission Directorate (ARMD) via Transformational Tools and Technologies (TTT) Project.

Computational↗
NASA NTRS1991

An efficient formulation and implementation of the analytic energy gradient method to the single and double excitation coupled-cluster wave function - Application to Cl2O2

The analytic energy gradient for the single and double excitation coupled-cluster (CCSD) wave function has been reformulated and implemented in a new set of programs. The reformulated set of gradient equations have a smaller computational cost than any previously published. The iterative solution of the linear equations and the construction of the effective density matrices are fully vectorized, being based on matrix multiplications. The new method has been used to investigate the Cl2O2 molecule, which has recently been postulated as an important intermediate in the destruction of ozone in the stratosphere. In addition to reporting computational timings, the CCSD equilibrium geometries, harmonic vibrational frequencies, infrared intensities, and relative energetics of three isomers of Cl2O2 are presented.

Rendell, Alistair P.↗
DOE OSTI2026

Stochastic Thermo-Hydro Modeling and Neural Network Surrogate Development for Thermal Resource Assessment of the Galleries-to-Calories Geobattery

The Galleries-to-Calories Geobattery concept explores the use of abandoned coal mine workings for large-scale thermal energy transport and storage. The system involves injecting waste heat from a supercomputing facility into flooded mine galleries, where groundwater flow can store and transport thermal energy for potential recovery in downgradient district heating and cooling applications. To evaluate the feasibility and performance of the Geobattery under geological and operational uncertainty, we developed a suite of stochastic thermo-hydrological (TH) simulations using Monte Carlo sampling of key uncertain parameters (e.g., permeability, porosity, thermal conductivity, specific heat capacity) and operating conditions (e.g., injection rate, injection temperature). Results identified injection rate and temperature as the most influential parameters governing thermal front propagation, while the geometry of the room-and-pillar structure played a critical role in directing the extent and orientation of thermal advancement. Optimal combinations of material properties for maximizing heat recovery were also determined. To address the high computational cost of coupled-process stochastic modeling, we trained a neural network surrogate model on 24,000 physics-based realizations, achieving an R² > 0.99 and MAE < 0.1 for temperature predictions at monitoring locations. This surrogate enabled an additional 100,000 realizations for global sensitivity analysis and probabilistic thermal resource assessment. The integrated stochastic physics–surrogate modeling framework offers a computationally efficient tool for quantifying uncertainty, identifying key drivers, and informing early-stage design decisions for Geobattery systems.

15 - GEOTHERMAL ENERGY↗
DOE OSTI2025

Orthorhombic Structures as Inorganic Halide Perovskite Models for High-Throughput DFT Investigations

Although high-symmetry Pm-3m space group cubic models are computationally efficient for high-throughput density functional theory (DFT) calculations of inorganic ternary (ABX 3 ) halide perovskites (HPs), they frequently predict band gaps (E g ) that disagree with experiment. Conversely, while low-symmetry cubic polymorphous networks (PN) comprised of 160 to 320 atoms incur significantly greater computational cost, they predict E g ’s that are more closely aligned with experiment. Here, in this study, we compare the DFT total energies and E g ’s predicted by four high-symmetry structure models (Pnma orthorhombic, R3m trigonal, P4/mbm tetragonal, and Pm-3m cubic) to cubic PNs for 5 experimentally characterized ternary HPs and find that the orthorhombic model computes Eg’s with the smallest MAD of 0.23 eV relative to the PNs. Pair distribution functions and DFT-computed total energies show that octahedral tilting, which is present in the 20-atom orthorhombic and 160-atom cubic PN models but not in the 5-atom cubic models, stabilizes all 5 compositions in our benchmarking set. We also find that imposing PN constraints when generating and optimizing these orthorhombic structures by fixing the unit cell lattice vectors and displacing the atoms prior to ionic relaxation with DFT, which we call the orthorhombic surrogate model (OSM), lowers the MAD of E g predictions to 0.09 eV. Our OSM predicts the PN band gaps of an additional 95 theoretical inorganic ternary HPs with MAD of 0.08 eV, supporting its usage in high-throughput DFT investigations to closely estimate PN band gaps with much less computational expense.

14 SOLAR ENERGY↗
DOE OSTI2025

Extending SEER for Extreme Heterogeneity

Heterogeneous and multi-device nodes are increasingly common in high-performance computing and data centers, yet existing programming models often lack simple, transparent, and portable support for these diverse architectures. The main contribution of this work is the development of novel SEER capabilities to address this challenge by providing a descriptive programming model that allows applications to seamlessly leverage heterogeneous nodes across various device types. SEER uses efficient memory management and can select the proper device[s] depending on the computational cost of the applications. This is completely transparent to the programmer, thereby providing a highly productive programming environment. Integrating extreme heterogeneity into the SEER library as shown with the use of NVIDIA and AMD GPUs simultaneously allows it to expand and exploit the performance possibilities. Our analysis based on the well-known Conjugate Gradient algorithm reports accelerations above 1.5 × on computationally demanding steps of such an algorithm by using both architectures simultaneously.

Teranishi, Keita [ORNL] (ORCID:0000000166472690)↗
NASA NTRS1998

Modeling Macro- and Micro-Scale Turbulent Mixing and Chemistry in Engine Exhaust Plumes

Simulation of turbulent mixing and chemical processes in the near-field plume and plume-vortex regimes has been successfully carried out recently using a reduced gas phase kinetics mechanism which substantially decreased the computational cost. A detailed mechanism including gas phase HOx, NOx, and SOx chemistry between the aircraft exhaust and the ambient air in near-field aircraft plumes is compiled. A reduced mechanism capturing the major chemical pathways is developed. Predictions by the reduced mechanism are found to be in good agreement with those by the detailed mechanism. With the reduced chemistry, the computer CPU time is saved by a factor of more than 3.5 for the near-field plume modeling. Distributions of major chemical species are obtained and analyzed. The computed sensitivities of major species with respect to reaction step are deduced for identification of the dominant gas phase kinetic reaction pathways in the jet plume. Both the near field plume and the plume-vortex regimes were investigated using advanced mixing models. In the near field, a stand-alone mixing model was used to investigate the impact of turbulent mixing on the micro- and macro-scale mixing processes using a reduced reaction kinetics model. The plume-vortex regime was simulated using a large-eddy simulation model. Vortex plume behind Boeing 737 and 747 aircraft was simulated along with relevant kinetics. Many features of the computed flow field show reasonable agreement with data. The entrainment of the engine plumes into the wing tip vortices and also the partial detrainment of the plume were numerically captured. The impact of fluid mechanics on the chemical processes was also studied. Results show that there are significant differences between spatial and temporal simulations especially in the predicted SO3 concentrations. This has important implications for the prediction of sulfuric acid aerosols in the wake and may partly explain the discrepancy between past numerical studies (that employed parabolic or temporal approximations) and the measured data. Finally to address the major uncertainty in the near-field plume modeling related to the plume processing of sulfur compounds and advanced model was developed to evaluate its impact on the chemical processes in the near wake. A comprehensive aerosol model is developed and it is coupled with chemical kinetics and the axisymmetric turbulent jet flow models. The integrated model is used to simulate microphysical processes in the near-field jet plume, including sulfuric acid and water binary homogeneous nucleation, coagulation, non-equilibrium heteromolecular condensation, and sulfur-induced soot activation. The formation and evolution of aerosols are computed and analyzed. The computed results show that a large number of ultra-fine (0.3--0.6 nm in radius) volatile HSO4 - HO embryos are generated in the near-field plume. These embryos further grow in size by self coagulation and condensation. Soot particles can be activated by both heterogeneous nucleation and scavenging of H2SO4-H2O aerosols. These activated soot particles can serve as water condensation nuclei for contrail formation. Conditions under which ice contrails can form behind aircrafts are studied. The sensitivities of the threshold temperature for contrail formation with respect to aircraft propulsion efficiency, relative humidity, and ambient pressure are evaluated. The computed aerosol properties for different extent of fuel sulfur conversion to S(VI) (SO3 and H2SO4) in engine are examined and the results are found to be sensitive to this conversion fraction.

Menon, Suresh↗
DOE OSTI2025

Comparative Analysis of DNA LLM Classification Techniques Using Intra-Layer Feature Extraction with Autoencoder Stacks [Poster]

This project conducts a comparative analysis of DNA LLM classification techniques using Evo2, Grover, and UTRML, focusing on intra-layer feature extraction in Evo2. By extracting features from multiple layers of Evo2 and integrating them into an autoencoder stack with a binary classification head, we evaluate its effectiveness in classifying genomic sequences compared to smaller DNA language models. My findings demonstrate that Evo2 outperforms Grover and UTRML in classification accuracy on a dataset provided by department 08625, CAO2021, while UTRML offers competitive performance with lower computational costs. This study highlights the potential of advanced embedding techniques in enhancing genomic data analysis and informs future research in bioinformatics.

59 BASIC BIOLOGICAL SCIENCES↗