Microlensify 1.0.2

Microlensing Single Lens Classification – Atousa Kalantari & Somayeh Khakpash

Microlensify

Deep-learning microlensing classifier

Microlensify is a deep learning model that detects single lens microlensing events in light curves. It works with light curves from any telescope, either via URLs (e.g., MAST FITS files) or your own data files.

Microlensify splits your light curve into chunks and downsamples them to a fixed length for prediction. Predictions are made for each chunk individually and also for the whole light curve.

---

Installation

pip install Microlensify

Usage

Microlensify <input_file> <compute_stats> <n_cores>

Arguments

<input_file>

• URL input: a file containing URLs of light curves and their flux & time column names.
• Local files: a file containing paths to your files and their flux & time column names.

<compute_stats>

• yes — compute flux statistics (min, max, median, std, std/(max-min)) from raw flux.
• no — flux is already normalized (like TESS QLP pipeline flux), so the model uses fixed values from the training set for these statistics.

<n_cores>

• Number of CPU cores to use for parallel processing.

Input File Format

• Tab-separated file.

• Column 1: time

• Column 2: flux

• URL input: list the URLs and the flux/time column names.

• Local files: list file paths and the flux/time column names.

• Example input files are provided in the input_examples folder.

Output

All results are saved to prediction_results.csv with the following columns:

Column	Description
Source	URL or file path of the light curve
Class	1 if probability > 0.99, else 0
Probability	Model probability of microlensing
Real_4FWHM_days	Predicted 4FWHM duration (trustworthy if light curve has 940 points over 27.4 days)
Latent_Space	20 latent space values from the model (can be used for reconstruction)
Points	Number of points in the chunk
Chunk_Description	Description of the chunk used

---

Citation

If you use Microlensify in your research, please cite it appropriately.

---

Contact

For questions or support, please open an issue or contact [e-mail].