KilonovaSCORER

A Simulation-Based Scoring Framework for Early Identification of Kilonova Candidates

[] [] []

KilonovaScorer is an open-source statistical framework designed to score and rank transient candidates according to their consistency with physically motivated kilonova light-curve simulations. The package is optimized for the early-time, sparse-data regime encountered in multimessenger astronomy follow-up campaigns.

The framework enables rapid candidate vetting using simulation-based statistical diagnostics, making it suitable for alert-driven environments such as LSST brokers and gravitational-wave follow-up infrastructures.

---

Scientific Motivation

Kilonovae evolve rapidly, often reaching peak brightness within a few days after merger. Under realistic survey cadences (e.g., LSST), only a small number of photometric observations are typically available before the transient fades.

In this low-data regime:

• Bayesian parameter inference becomes weakly constrained,
• model comparison is dominated by prior assumptions,
• rapid follow-up prioritization becomes challenging.

KilonovaScorer addresses this problem by comparing observations directly against ensembles of simulated light curves using prior predictive checks, producing quantitative compatibility scores suitable for real-time decision making.

Installation

Install directly with pip:

pip install git+https://github.com/phelipedarc/KilonovaSCORER.git

or from source:

git clone https://github.com/phelipedarc/KilonovaSCORER
cd KilonovaSCORER
pip install -e .

Creating your set of simulations:

data_sim_all = ks.simulation.simulate_kilonova(N_SIM=1000, MODEL_NAME='two_component_kilonova_model', SAVE_CSV=False)
data_sim_all['filter_mapped'] = data_sim_all['band'].str.lower().map(FILTER_LOOKUP)

or running the script

simulating_kne_pop.py --nsim 100000 

Core Functionality

The current beta release provides the following high-level interface:

Function	Description
kilonovascorer_v1	Main scoring engine. Compares observational data against simulation grids and returns scoring metrics and diagnostics.
load_observations	Data ingestion utility. Cleans and formats CSV/JSON light-curve data for compatibility.
preprocess_lsst_like	Cadence preprocessing. Converts arbitrary light curves into LSST-like follow-up cadence.
plot_final_all_metrics	Visualization module. Produces multi-panel diagnostic plots of scoring results.
binned_stats_cumulative_ptail	Statistical diagnostics. Computes cumulative scores and ABC-style consistency metrics.
plotting.plot_survivor_param_kde_grid	Visualization diagnostics. Produces a multi-panel for each kilonova parameter constraint over time from the ABC-diagnostic.

---

Project Development

This project was developed at Northwestern University (December 2025 – March 2026) in collaboration with LAB-IA — Laboratório de Inteligência Artificial (CBPF).

• Northwestern University — https://ciera.northwestern.edu/
• LAB-IA — https://labia.cbpf.br/

Corresponding email: phelipedarc@gmail.com

About me: https://phelipedarc.github.io/