Model Overview
Astromer-1 is a transformer-based model specifically designed for astronomical time series embeddings. It is pretrained in a self-supervised manner on a large corpus of astronomical time series data (MACHO). It leverages attention mechanisms to capture long-range dependencies in irregularly sampled light curves and handles the unique challenges of astronomical data.
Key Features
- Handles Irregular Sampling: Designed for astronomical light curves with missing observations/seasonal gaps
- Domain Specialization: Designed for light-curve embeddings
