Transformer Embeddings for Fast Microlensing Inference

TL;DR

This paper introduces a Transformer-based pipeline for rapid and accurate inference of free-floating planets from noisy microlensing data, significantly outperforming traditional methods in speed.

Contribution

The novel use of Transformer encoders to learn compressed summaries of microlensing time-series data for fast Bayesian inference is presented.

Findings

Accurate posterior estimation over three orders of magnitude faster.

Effective on real microlensing event data.

Well-calibrated probabilistic outputs.

Abstract

The search for free-floating planets (FFPs) is a key science driver for upcoming microlensing surveys like the Nancy Grace Roman Galactic Exoplanet Survey. These rogue worlds are typically detected via short-duration microlensing events, the characterization of which often requires analyzing noisy, irregularly-sampled observations. We present a pipeline for this task using simulation-based inference. We use a Transformer encoder to learn a compressed summary representation of the raw time-series data, which in turn conditions a neural posterior estimator. We demonstrate that our method produces accurate and well-calibrated posteriors over three orders of magnitude faster than traditional methods. We also demonstrate its performance on KMT-BLG-2019-2073, a short-duration FFP candidate event.