Active Learning for Computationally Efficient Distribution of Binary Evolution Simulations
arXiv:2203.16683 · doi:10.3847/1538-4357/ac8b05
Abstract
Binary stars undergo a variety of interactions and evolutionary phases, critical for predicting and explaining observed properties. Binary population synthesis with full stellar-structure and evolution simulations are computationally expensive requiring a large number of mass-transfer sequences. The recently developed binary population synthesis code POSYDON incorporates grids of MESA binary star simulations which are then interpolated to model large-scale populations of massive binaries. The traditional method of computing a high-density rectilinear grid of simulations is not scalable for higher-dimension grids, accounting for a range of metallicities, rotation, and eccentricity. We present a new active learning algorithm, psy-cris, which uses machine learning in the data-gathering process to adaptively and iteratively select targeted simulations to run, resulting in a custom, high-performance training set. We test psy-cris on a toy problem and find the resulting training sets require fewer simulations for accurate classification and regression than either regular or randomly sampled grids. We further apply psy-cris to the target problem of building a dynamic grid of MESA simulations, and we demonstrate that, even without fine tuning, a simulation set of only the size of a rectilinear grid is sufficient to achieve the same classification accuracy. We anticipate further gains when algorithmic parameters are optimized for the targeted application. We find that optimizing for classification only may lead to performance losses in regression, and vice versa. Lowering the computational cost of producing grids will enable future versions of POSYDON to cover more input parameters while preserving interpolation accuracies.
arXiv comment21 pages, 10 figures, ApJ in press
References in corpus (50)
- SciPy 1.0--Fundamental Algorithms for Scientific Computing in Python
- The NumPy array: a structure for efficient numerical computation
- Modules for Experiments in Stellar Astrophysics (MESA)
- Modules for Experiments in Stellar Astrophysics (MESA): Giant Planets, Oscillations, Rotation, and Massive Stars
- Modules for Experiments in Stellar Astrophysics (MESA): Binaries, Pulsations, and Explosions
- MESA Isochrones and Stellar Tracks (MIST). I: Solar-Scaled Models
- E pur si muove: Galiliean-invariant cosmological hydrodynamical simulations on a moving mesh
- Modules for Experiments in Stellar Astrophysics (MESA): Convective Boundaries, Element Diffusion, and Massive Star Explosions
- Evolution of binary stars and the effect of tides on binary populations
- Modules for Experiments in Stellar Astrophysics (MESA): Pulsating Variable Stars, Rotation, Convective Boundaries, and Energy Conservation
- Comprehensive analytic formulae for stellar evolution as a function of mass and metallicity
- First results from the IllustrisTNG simulations: the galaxy color bimodality
- Parallel Tempering: Theory, Applications, and New Perspectives
- A Comprehensive Study of Binary Compact Objects as Gravitational Wave Sources: Evolutionary Channels, Rates, and Physical Properties
- Binary Population and Spectral Synthesis Version 2.1: construction, observational verification and new results
- Compact Object Modeling with the StarTrack Population Synthesis Code
- Surrogate models for precessing binary black hole simulations with unequal masses
- Progenitors of gravitational wave mergers: Binary evolution with the stellar grid-based code ComBinE
- Formation of the first three gravitational-wave observations through isolated binary evolution
- Supernova Type Ia progenitors from merging double white dwarfs: Using a new population synthesis model
- Merging black hole binaries: the effects of progenitor's metallicity, mass-loss rate and Eddington factor
- COSMIC Variance in Binary Population Synthesis
- Merging black hole binaries with the SEVN code
- Supernova Simulations from a 3D Progenitor Model -- Impact of Perturbations and Evolution of Explosion Properties
- A Complete Survey of Case A Binary Evolution with Comparison to Observed Algol-Type Systems
- Self-consistent 3D Supernova Models From -7 Minutes to +7 Seconds: a 1-bethe Explosion of a ~19 Solar-mass Progenitor
- A Successful 3D Core-Collapse Supernova Explosion Model
- Efficiency of mass transfer in massive close binaries, Tests from double-lined eclipsing binaries in the SMC
- The role of mass transfer and common envelope evolution in the formation of merging binary black holes
- Population Synthesis of Binary Carbon-enhanced Metal-poor Stars
- A Benchmark and Comparison of Active Learning for Logistic Regression
- Accuracy of inference on the physics of binary evolution from gravitational-wave observations
- Binary Population Synthesis
- Binary Black Hole Formation with Detailed Modeling: Stable Mass Transfer Leads to Lower Merger Rates
- The Ecological Impact of High-performance Computing in Astrophysics
- Optimizing spectroscopic follow-up strategies for supernova photometric classification with active learning
- On Carbon Burning in Super Asymptotic Giant Branch Stars
- Active Learning to Overcome Sample Selection Bias: Application to Photometric Variable Star Classification
- Statistical Gravitational Waveform Models: What to Simulate Next?
- The origin of pulsating ultra-luminous X-ray sources: Low- and intermediate-mass X-ray binaries containing neutron star accretors
- Description of the Scenario Machine
- On the Formation of Ultraluminous X-ray Sources with Neutron Star Accretors: the Case of M82 X-2
- STROOPWAFEL: Simulating rare outcomes from astrophysical populations, with application to gravitational-wave sources
- The role of core-collapse physics in the observability of black-hole neutron-star mergers as multi-messenger sources
- Constraining the Parameters of High-Dimensional Models with Active Learning
- dart_board: Binary Population Synthesis with Markov Chain Monte Carlo
- Progenitors of low-mass binary black-hole mergers in the isolated binary evolution scenario
- Interpolating Detailed Simulations of Kilonovae: Adaptive Learning and Parameter Inference Applications
- An Active Instance-based Machine Learning method for Stellar Population Studies
- Estimation of the Galaxy Quenching Rate in the Illustris Simulation