The Machine Learning and Instrument Autonomy strives to research, develop, and infuse data driven solutions machine learning and other data science supporting technologies to advance robotic exploration and science.
Recent News
-
▸ December 3rd, 2024 by Ryan McGranaghan
Ryan McGranaghan co-authored a paper entitled INDUS: Effective and Efficient Language Models for Scientific Applications (Bhattacharjee, et al.) presented at The 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP 2024). This paper presents INDUS, a comprehensive suite of domain-specific large language models for Earth science, biology, physics, heliophysics, planetary sciences, and astrophysics, which includes an encoder model, a contrastive text embedding model, and knowledge-distilled variants, demonstrating superior performance on specialized scientific tasks and benchmarks.
-
▸ November 21st, 2024 by Jake Lee
The MLIA group will have a large presence at the American Geophysical Union Annual Meeting 2024:
ID Time & Loc Title Authors B11J-1454 Mon 9th
08:30-12:20 (Poster)Global Estimates of Mineral Mass Fractional Cover from the Earth Surface Mineral Dust Source Investigation Philip Brodrick, Roger Nelson Clark, Carlos Pérez García-Pando, Maria Gonçalves, Gregory S Okin, Ron L Miller, David R Thompson, Robert O Green, Abigail Keebler and Bethany L Ehlmann A11K-1773 Mon 9th
08:30-12:20 (Poster)Towards practical Cloud Tomography: A Deep Learning Framework for Retrieval and Uncertainty Quantification of Cloud Optical Thickness from Multi-Angle Imaging Nicholas LaHaye, Linda Forster, Steffen Mauceri, Marcin Kurowski and Anthony B Davis IN13A-2142A Mon 9th
13:40-17:30 (Poster)A foundation for understanding scientific knowledge commons: Examining the ‘society of science’ Ryan Michael McGranaghan P13D-3094 Mon 9th
13:40-17:30 (Poster)Characterizing the Lifecycle of Small Martian Dust Storms for Science and Mission Planning Mark Wronkiewicz, Marek Slipski, Joseph Michael Battalio, Huiqun Wang and David M Kass G14B-09 Mon 9th
16:24-16:27 (eLighting 5)Slow Slip Event Detection in the Cascadia Subduction Zone using Machine Learning and GNSS Time Series Umaa Rebbapragada, Zhen Liu, Yehuda Bock, Shibani Likhite, Angelyn W Moore, Steven Lu and Graciela Bachu B23B-1557 Tues 10th
13:40-17:30 (Poster)Increasing data accessibility in multi-sensor airborne campaigns: Lessons from BioSCape Philip Brodrick, Erin Lee Hestir, Drew Meyers, Adam Wilson, Jasper Slingsby and Anabelle Cardoso NG24A-07 Tues 10th
17:00-17:10 (Marquis 12-13)Prediction of High-Latitude Ionospheric Scintillation using Deep Learning Models Umaa Rebbapragada, Steven Lu, Ryan M McGranaghan, Samuel Berndt and Hannah Rae Kerner B31H-1382A Wed 11th
08:30-12:20 (Poster)Comparison of Model Complexity, Representative Capabilities, and Performance for Self-Supervised Multi-Sensor Wildfire and Smoke Segmentation and Tracking Nicholas LaHaye, Hugo Lee, Anastasija Easley, Michael Garay, Kyongsik Yun, Alex Goodman and Olga V. Kalashnikova P34A-02 Wed 11th
16:16-16:26 (Liberty M)Image Content and Similarity Based Search for the Planetary Data System Cartography and Imaging Sciences Node Steven Lu, Sara Bond, Steven Liu, Tariq K Soliman and Robert G Deen P41E-2937 Thurs 12th
08:30-12:20 (Poster)Creating a Novel and Comprehensive Map of Martian Frost Based on Visible, Thermal, and Spectral Observations Mark Wronkiewicz, Serina Diniega, Gary B Doran, Steven Lu, Umaa Rebbapragada and Jacob Widmer GP43B-3573 Thurs 12th
13:40-17:30 (Poster)The science of space weather risk: Demonstrating frontier discovery of the geospace-power grid system with complexity science Ryan Michael McGranaghan Lauren Orr, Sandra C. Chapman and Jui-We Chang IN51B-04 Fri 13th
09:00-09:10 (Marquis 3-4)Investigating the Benefits of a Large Mars Model Umaa Rebbapragada, Mirali Purohit, Hannah Rae Kerner, Steven Lu and Serina Diniega A52C-07 Fri 13th
11:25-11:35 (154 A-B)Challenges and Opportunities in Operational Automated Methane Plume Detection Brian Bue, Jake Lee, Andrew K Thorpe, Philip Brodrick, Daniel Cusworth, Robert O Green and Riley Duren A53I-2209 Fri 13th
13:40-17:30 (Poster)Spectral Transformers for EMIT Methane Retrieval Jake Lee, Brian Bue, Philip G. Brodrick, Andrew K. Thorpe, William Keely, Michael Kiper and Jay Fahlen A05-06 iPoster Online DisCO: Diffusion for conditional retrieval of atmospheric CO2 from satellite observations William Keely, Steffen Mauceri, Otto Lamminpää, Gregory McGarragh, Christopher O’Dell and Sean Crowell A05-07 iPoster Online Methane emission quantification using diffusion applied to imaging spectroscopy time series for EMIT William Keely, Philip G. Brodrick, Andrew K. Thorpe, Jake Lee and Jay Fahlen -
▸ June 12th, 2024 by Mario Damiano
Mario Damiano, Aaron Bello-Arufe, Jeehyun Yang, and Renyu Hu published the paper entitled LHS 1140 b is a potentially habitable water world to the Astrophysical Journal Letters. The paper describe the data analysis of JWST observations, scientific interpretation, and scientific impact of the results.
-
▸ June 4th, 2024 by Philip Brodrick
Phil Brodrick was the JPL lead on the publication of a paper led by Longlei Li (Cornell University) entitled “Improved constraints on hematite refractive index for estimating climatic effects of dust aerosols”, in Nature Communications Earth and Environment. The work shows how new lab measurements of dust composition help constrain global radiative forcing, and demonstrates the need for improved remote measurements of surface minerals.