Dalek -- a deep-learning emulator for TARDIS

Abstract

Supernova spectral time series contain a wealth of information about the progenitor and explosion process of these energetic events. The modeling of these data requires the exploration of very high dimensional posterior probabilities with expensive radiative transfer codes. Even modest parametrizations of supernovae contain more than ten parameters and a detailed exploration demands at least several million function evaluations. Physically realistic models require at least tens of CPU minutes per evaluation putting a detailed reconstruction of the explosion out of reach of traditional methodology. The advent of widely available libraries for the training of neural networks combined with their ability to approximate almost arbitrary functions with high precision allows for a new approach to this problem. Instead of evaluating the radiative transfer model itself, one can build a neural network proxy trained on the simulations but evaluating orders of magnitude faster. Such a framework is called an emulator or surrogate model. In this work, we present an emulator for the TARDIS supernova radiative transfer code applied to Type Ia supernova spectra. We show that we can train an emulator for this problem given a modest training set of a hundred thousand spectra (easily calculable on modern supercomputers). The results show an accuracy on the percent level (that are dominated by the Monte Carlo nature of TARDIS and not the emulator) with a speedup of several orders of magnitude. This method has a much broader set of applications and is not limited to the presented problem.

6 pages;5 figures submitted to AAS Journals. Constructive Criticism invited

References in corpus (32)

• Scikit-learn: Machine Learning in Python
• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Improving neural networks by preventing co-adaptation of feature detectors
• ADADELTA: An Adaptive Learning Rate Method
• Popular Ensemble Methods: An Empirical Study
• Three-dimensional delayed-detonation models with nucleosynthesis for Type Ia supernovae
• Normal Type Ia supernovae from violent mergers of white dwarf binaries
• Double-detonation sub-Chandrasekhar supernovae: can minimum helium shell masses detonate the core?
• Sub-luminous type Ia supernovae from the mergers of equal-mass white dwarfs with M~0.9 M_sun
• Detonations in Sub-Chandrasekhar Mass C+O White Dwarfs
• Three-dimensional pure deflagration models with nucleosynthesis and synthetic observables for Type Ia supernovae
• Double-detonation sub-Chandrasekhar supernovae: synthetic observables for minimum helium shell mass models
• A statistical test for Nested Sampling algorithms
• Constraining Type Ia supernova models: SN 2011fe as a test case
• 3D deflagration simulations leaving bound remnants: a model for 2002cx-like Type Ia supernovae
• Abundance stratification in Type Ia supernovae - I. The case of SN 2002bo
• A spectral synthesis code for rapid modelling of supernovae
• 2D simulations of the double-detonation model for thermonuclear transients from low-mass carbon-oxygen white dwarfs
• Constructing A Flexible Likelihood Function For Spectroscopic Inference
• Deflagrations in hybrid CONe white dwarfs: a route to explain the faint Type Iax supernova 2008ha
• Synthetic light curves and spectra for three-dimensional delayed-detonation models of Type Ia supernovae
• Early light curves for Type Ia supernova explosion models
• The peculiar Type Ia supernova iPTF14atg: Chandrasekhar-mass explosion or violent merger?
• Type Ia supernovae from exploding oxygen-neon white dwarfs
• SN 2010lp---a Type Ia supernova from a violent merger of two carbon--oxygen White Dwarfs
• Three-dimensional simulations of gravitationally confined detonations compared to observations of SN 1991T
• Spectral modeling of type II supernovae. I. Dilution factors
• The white dwarf's carbon fraction as a secondary parameter of Type Ia supernovae
• Spectral modeling of type II supernovae II. A machine learning approach to quantitative spectroscopic analysis
• Gamma-ray diagnostics of Type Ia supernovae: Predictions of observables from three-dimensional modeling
• Thermonuclear explosions of rapidly differentially rotating white dwarfs: Candidates for superluminous Type Ia supernovae?
• Type Ia supernovae with and without blueshifted narrow Na I D lines - how different is their structure?

Cited by in corpus (9)

• ParSNIP: Generative Models of Transient Light Curves with Physics-Enabled Deep Learning
• Spectroscopic Studies of Type Ia Supernovae Using LSTM Neural Networks
• A New Emulated Monte Carlo Radiative Transfer Disk-Wind Model: X-Ray Accretion Disk-wind Emulator -- XRADE
• Probabilistic Reconstruction of Type Ia Supernova SN 2002bo
• Scalable stellar evolution forecasting: Deep learning emulation vs. hierarchical nearest neighbor interpolation
• DeepGlow: an efficient neural-network emulator of physical afterglow models for gamma-ray bursts and gravitational-wave events
• Exploring the origins of high-velocity features in SNe Ia with the spectral synthesis code TARDIS
• Ultraviolet Spectroscopy and TARDIS Models of the Broad-lined Type-Ic Supernova 2014ad
• New mass estimates for massive binary systems: a probabilistic approach using polarimetric radiative transfer