Explorable INR: An Implicit Neural Representation for Ensemble Simulation Enabling Efficient Spatial and Parameter Exploration

2025-04-01

AI Summary

High-resolution ensemble simulations (e.g., in cosmology and oceanography) incur prohibitive storage and computational costs, while existing surrogate models lack support for flexible point/region queries and efficient parameter-space exploration. Method: We propose an interactive implicit neural representation (INR) framework that innovatively integrates probabilistic affine forms (PAFs) to enable uncertainty propagation and statistical summarization; it formulates parameter optimization as a gradient-driven KL-divergence minimization problem, enabling scalable joint exploration of spatial and parameter spaces. Contribution/Results: Our approach achieves millisecond-level on-demand predictions at arbitrary points or regions—without compromising accuracy—while drastically reducing memory and computational overhead. It enables real-time ensemble analysis and visualization, delivering the first INR solution for large-scale scientific simulations that simultaneously ensures interpretability, explorability, and efficiency.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Reasoning under Uncertainty: Stochastic Optimization
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

With the growing computational power available for high-resolution ensemble simulations in scientific fields such as cosmology and oceanology, storage and computational demands present significant challenges. Current surrogate models fall short in the flexibility of point- or region-based predictions as the entire field reconstruction is required for each parameter setting, hence hindering the efficiency of parameter space exploration. Limitations exist in capturing physical attribute distributions and pinpointing optimal parameter configurations. In this work, we propose Explorable INR, a novel implicit neural representation-based surrogate model, designed to facilitate exploration and allow point-based spatial queries without computing full-scale field data. In addition, to further address computational bottlenecks of spatial exploration, we utilize probabilistic affine forms (PAFs) for uncertainty propagation through Explorable INR to obtain statistical summaries, facilitating various ensemble analysis and visualization tasks that are expensive with existing models. Furthermore, we reformulate the parameter exploration problem as optimization tasks using gradient descent and KL divergence minimization that ensures scalability. We demonstrate that the Explorable INR with the proposed approach for spatial and parameter exploration can significantly reduce computation and memory costs while providing effective ensemble analysis.

Problem

• Efficient spatial and parameter exploration in ensemble simulations
• Flexible point-based predictions without full field reconstruction
• Scalable optimization for parameter space analysis

Innovation

• Implicit neural representation for flexible spatial queries
• Probabilistic affine forms for uncertainty propagation
• Gradient descent and KL divergence for parameter optimization