2025 CSDMS meeting-133: Difference between revisions
From CSDMS
Created page with "{{CSDMS meeting personal information template-2025 |CSDMS meeting first name=Donsub |CSDMS meeting last name=Rim |CSDMS meeting institute=Washington University in Saint Louis |CSDMS meeting city=Saint Louis |CSDMS meeting country=United States |CSDMS meeting state=Missouri |CSDMS meeting email address=rim@wustl.edu }} {{CSDMS meeting select clinics1 2025 |CSDMS_meeting_select_clinics1_2025=2) Simulating Glacier Flow with ICEPACK }} {{CSDMS meeting select clinics2 2025 |C..." |
No edit summary |
||
| Line 29: | Line 29: | ||
{{CSDMS meeting abstract template 2025 | {{CSDMS meeting abstract template 2025 | ||
|CSDMS meeting abstract=Neural networks (NNs) enable precise modeling of complicated geophysical phenomena but can be sensitive to small input changes. In this work, we present a new method for analyzing this instability in NNs. We focus our analysis on adversarial examples, test-time inputs with carefully crafted human-imperceptible perturbations that expose the worst-case instability in a model's predictions. Our stability analysis is based on a low-rank expansion of NNs on a fixed input, and we apply our analysis to a NN model for tsunami early warning which takes geodetic measurements as the input and forecasts tsunami waveforms. The result is an improved description of local stability that explains adversarial examples generated by a standard gradient-based algorithm, and allows the generation of other comparable examples. Our analysis can predict whether noise in the geodetic input will produce an unstable output, and identifies a potential approach to filtering the input that enable more robust forecasting. | |CSDMS meeting abstract=Neural networks (NNs) enable precise modeling of complicated geophysical phenomena but can be sensitive to small input changes. In this work, we present a new method for analyzing this instability in NNs. We focus our analysis on adversarial examples, test-time inputs with carefully crafted human-imperceptible perturbations that expose the worst-case instability in a model's predictions. Our stability analysis is based on a low-rank expansion of NNs on a fixed input, and we apply our analysis to a NN model for tsunami early warning which takes geodetic measurements as the input and forecasts tsunami waveforms. The result is an improved description of local stability that explains adversarial examples generated by a standard gradient-based algorithm, and allows the generation of other comparable examples. Our analysis can predict whether noise in the geodetic input will produce an unstable output, and identifies a potential approach to filtering the input that enable more robust forecasting. | ||
|CSDMS meeting posterPDF= Tsunami neuralnetwork stability.pdf | |||
|CSDMS meeting posterPNG= Tsunami neuralnetwork stability.png | |||
}} | }} | ||
{{blank line template}} | {{blank line template}} | ||
Latest revision as of 11:28, 21 May 2025
(if you haven't already)
Log in (or create account for non-CSDMS members)
Forgot username? Search or email:CSDMSweb@colorado.edu
Browse abstracts
A Stability Analysis of Neural Networks and Its Application to Tsunami Early Warning
Donsub Rim,
Washington University in Saint Louis Saint Louis Missouri, United States. rim@wustl.edu
Neural networks (NNs) enable precise modeling of complicated geophysical phenomena but can be sensitive to small input changes. In this work, we present a new method for analyzing this instability in NNs. We focus our analysis on adversarial examples, test-time inputs with carefully crafted human-imperceptible perturbations that expose the worst-case instability in a model's predictions. Our stability analysis is based on a low-rank expansion of NNs on a fixed input, and we apply our analysis to a NN model for tsunami early warning which takes geodetic measurements as the input and forecasts tsunami waveforms. The result is an improved description of local stability that explains adversarial examples generated by a standard gradient-based algorithm, and allows the generation of other comparable examples. Our analysis can predict whether noise in the geodetic input will produce an unstable output, and identifies a potential approach to filtering the input that enable more robust forecasting.
