| Title: |
Leveraging Open Science Machine Learning Challenges for Data Constrained Planetary Mission Instruments |
| Authors: |
da Poian, Victoria; Lyness, Eric, I; Qi, Jay, Y; Shah, Isha; Lipstein, Greg; Archer, P, Doug; Chou, Luoth; Freissinet, Caroline; Malespin, Charles, A; Mcadam, Amy, C; Knudson, Christine, A; Theiling, Bethany, P; Hörst, Sarah, M |
| Contributors: |
NASA Goddard Space Flight Center (GSFC); Microtel LLC; Morton K. Blaustein Department of Earth and Planetary Sciences Baltimore; Johns Hopkins University Baltimore (JHU); DrivenData Inc.; Jacobs Technology ESCG; NASA Johnson Space Center (JSC); NASA; University of Maryland Baltimore County (UMBC); University System of Maryland; Center for Research and Exploration in Space Science and Technology Baltimore (CRESST); University System of Maryland-University System of Maryland; PLANETO - LATMOS; Laboratoire Atmosphères, Milieux, Observations Spatiales (LATMOS); Université de Versailles Saint-Quentin-en-Yvelines (UVSQ)-Institut national des sciences de l'Univers (INSU - CNRS)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Université de Versailles Saint-Quentin-en-Yvelines (UVSQ)-Institut national des sciences de l'Univers (INSU - CNRS)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS); University of Maryland College Park (UMD); Center for Research and Exploration in Space Science and Technology College Park (CRESST) |
| Source: |
EISSN: 2752-8200 ; RAS Techniques and Instruments ; https://insu.hal.science/insu-04515256 ; RAS Techniques and Instruments, 2024, 3 (1), pp.156-165. ⟨10.1093/rasti/rzae009⟩ |
| Publisher Information: |
CCSD; Oxford Academic |
| Publication Year: |
2024 |
| Collection: |
Université de Versailles Saint-Quentin-en-Yvelines: HAL-UVSQ |
| Subject Terms: |
[SDU]Sciences of the Universe [physics] |
| Description: |
International audience ; We set up two open-science machine learning (ML) challenges focusing on building models to automatically analyze massspectrometry (MS) data for Mars exploration. ML challenges provide an excellent way to engage a diverse set of experts withbenchmark training data, explore a wide range of ML and data science approaches, and identify promising models based onempirical results, as well as to get independent external analyses to compare to those of the internal team. These two challengeswere proof-of-concept projects to analyze the feasibility of combining data collected from different instruments in a singleML application. We selected mass spectrometry data from 1) commercial instruments and 2) the Sample Analysis at Mars(SAM, an instrument suite that includes a mass spectrometer subsystem onboard the Curiosity rover) testbed. These challenges,organized with DrivenData, gathered more than 1,150 unique participants from all over the world, and obtained more than 600solutions contributing powerful models to the analysis of rock and soil samples relevant to planetary science using various massspectrometry datasets. These two challenges demonstrated the suitability and value of multiple ML approaches to classifyingplanetary analog datasets from both commercial and flight-like instruments.We present the processes from the problem identification, challenge setups, and challenge results that gathered creative anddiverse solutions from worldwide participants, in some cases with no backgrounds in mass spectrometry. We also present thepotential and limitations of these solutions for ML application in future planetary missions. Our longer-term goal is to deploythese powerful methods onboard the spacecraft to autonomously guide space operations and reduce ground-in-the-loop reliance. |
| Document Type: |
article in journal/newspaper |
| Language: |
English |
| DOI: |
10.1093/rasti/rzae009 |
| Availability: |
https://insu.hal.science/insu-04515256; https://insu.hal.science/insu-04515256v2/document; https://insu.hal.science/insu-04515256v2/file/rzae009%20%281%29.pdf; https://doi.org/10.1093/rasti/rzae009 |
| Rights: |
http://creativecommons.org/licenses/by/ ; info:eu-repo/semantics/OpenAccess |
| Accession Number: |
edsbas.8907663D |
| Database: |
BASE |