# ExoMiner: A Highly Accurate and Explainable Deep Learning Classifier that Validates 301 New Exoplanets

**Authors:** Hamed Valizadegan, Miguel Martinho, Laurent S. Wilkens, Jon M. Jenkins, Jeffrey Smith, Douglas A. Caldwell, Joseph D. Twicken, Pedro C. Gerum, Nikash Walia, Kaylie Hausknecht, Noa Y. Lubin, Stephen T. Bryson, Nikunj C. Oza  
**arXiv:** [2111.10009](https://arxiv.org/abs/2111.10009)  
**Date:** 2021-11-19  
**Summary page:** https://www.summarizepaper.com/en/paper/2111.10009v2/

## Key points

- Machine learning is being used to analyze data from Kepler and TESS missions to search for new exoplanets.
- Processing over 100,000 potential transit signals to create a catalog of planet candidates is challenging and requires accurate classifiers.
- ExoMiner is a deep learning classifier proposed by researchers that mimics how domain experts examine diagnostic tests to vet a transit signal.
- ExoMiner is highly accurate, explainable, and robust due to its modular design.
- The proposed classifier allows for the validation of 301 new exoplanets from the MAST Kepler Archive and can be applied across missions such as TESS.
- An extensive experimental study was conducted to verify that ExoMiner is more reliable and accurate than existing transit signal classifiers in terms of different classification and ranking metrics.
- The researchers plan to expand their study by utilizing more advanced explainability techniques, analyzing label noise and underrepresented scenarios, and publishing results on utilizing an adapted version of ExoMiner for TESS transit signals classification.
- Overall, ExoMiner represents a significant advancement in the field of exoplanet discovery with great potential to improve our understanding of these distant worlds.

## Summary

In recent years, there has been a growing interest in using machine learning to analyze the vast amount of data generated by the Kepler and TESS missions in search of new exoplanets. To process over 100,000 potential transit signals and create a catalog of planet candidates is a challenging task that requires accurate and reliable classifiers. In this study, the researchers propose ExoMiner - a deep learning classifier that mimics how domain experts examine diagnostic tests to vet a transit signal. ExoMiner is highly accurate, explainable and robust due to its modular design. The proposed classifier allows for the validation of 301 new exoplanets from the MAST Kepler Archive and can be applied across missions such as TESS. An extensive experimental study was conducted to verify that ExoMiner is more reliable and accurate than existing transit signal classifiers in terms of different classification and ranking metrics. For example, for a fixed precision value of 99%, ExoMiner retrieves 93.6% of all exoplanets in the test set (i.e., recall=0.936), while this rate is only 76.3% for the best existing classifier. The researchers plan to expand their study by utilizing more advanced explainability techniques and designing a framework that can provide experts with insight about ExoMiner's labels and scores; they also aim to conduct an analysis on label noise and underrepresented scenarios as well as publish results on utilizing an adapted version of ExoMiner for TESS transit signals classification by extending its features set similarly to Kepler data processing. Overall, ExoMiner represents a significant advancement in the field of exoplanet discovery with great potential to improve our understanding of these distant worlds.