An official website of the United States government
Accurate prediction of dynamical systems in unstructured meshes has recently shown successes in scientific simulations. Many dynamical systems have a nonnegligible level of stochasticity introduced by various factors (e.g. chaoticity), so there is a need for a unified framework that captures both deterministic and stochastic components in the rollouts of these systems. Inspired by regeneration learning, we propose a new model that combines generative and sequential networks to model dynamical systems. Specifically, we use an autoencoder to learn compact representations of full-space physical variables in a low-dimensional space. We then integrate a transformer with a conditional normalizing flow model to model the temporal sequence of latent representations. We evaluate the new model in both deterministic and stochastic systems. The model outperforms several competitive baseline models and makes more accurate predictions of deterministic systems. Its own prediction error is also reflected in its uncertainty estimations. When predicting stochastic systems, the proposed model generates high-quality rollout samples. The mean and variance of these samples well match the statistics of samples computed from expensive numerical simulations.
more »
« less
Competing Models
Abstract Different agents need to make a prediction. They observe identical data, but have different models: they predict using different explanatory variables. We study which agent believes they have the best predictive ability—as measured by the smallest subjective posterior mean squared prediction error—and show how it depends on the sample size. With small samples, we present results suggesting it is an agent using a low-dimensional model. With large samples, it is generally an agent with a high-dimensional model, possibly including irrelevant variables, but never excluding relevant ones. We apply our results to characterize the winning model in an auction of productive assets, to argue that entrepreneurs and investors with simple models will be overrepresented in new sectors, and to understand the proliferation of “factors” that explain the cross-sectional variation of expected stock returns in the asset-pricing literature.
more »
« less
- Award ID(s):
- 1763349
- PAR ID:
- 10472528
- Publisher / Repository:
- Oxford Academic Press
- Date Published:
- Journal Name:
- The Quarterly Journal of Economics
- Volume:
- 137
- Issue:
- 4
- ISSN:
- 0033-5533
- Page Range / eLocation ID:
- 2419 to 2457
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Many recent studies have explored remote sensing approaches to facilitate non‐destructive sampling of aboveground biomass (AGB). Lidar platforms (e.g., iPhone and iPad PRO models) have recently made remote sensing technologies widely available and present an alternative to traditional approaches for estimating AGB. Lidar approaches can be completed within a fraction of the time required by many analog methods. However, it is unknown if handheld sensors are capable of accurately predicting AGB or how different modeling techniques affect prediction accuracy. Here, we collected AGB from 0.25‐m2plots (N = 45) from three sites along an elevational gradient within rangelands surrounding Flagstaff, Arizona, USA. Each plot was scanned with a mobile laser scanner (MLS) and iPad before plants were clipped, dried, and weighed. We compared the capability of iPad and MLS sensors to estimate AGB via minimization of model normalized root mean square error (NRMSE). This process was performed on predictor subsets describing structural, spectral, and field‐based characteristics across a suite of modeling approaches including simple linear, stepwise, lasso, and random forest regression. We found that models developed from MLS and iPad data were equally capable of predicting AGB (NRMSE 26.6% and 29.3%, respectively) regardless of the variable subsets considered. We also found that stepwise regression regularly resulted in the lowest NRMSE. Structural variables were consistently selected during each modeling approach, while spectral variables were rarely included. Field‐based variables were important in linear regression models but were not included after variable selection within random forest models. These findings support the notion that remote sensing techniques offer a valid alternative to analog field‐based data collection methods. Together, our results demonstrate that data collected using a more widely available platform will perform similarly to a more costly option and outline a workflow for modeling AGB using remote sensing systems alone.more » « less
-
Abstract Large datasets make it possible to build predictive models that can capture heterogenous relationships between the response variable and features. The mixture of high-dimensional linear experts model posits that observations come from a mixture of high-dimensional linear regression models, where the mixture weights are themselves feature-dependent. In this article, we show how to construct valid prediction sets for an ℓ1-penalized mixture of experts model in the high-dimensional setting. We make use of a debiasing procedure to account for the bias induced by the penalization and propose a novel strategy for combining intervals to form a prediction set with coverage guarantees in the mixture setting. Synthetic examples and an application to the prediction of critical temperatures of superconducting materials show our method to have reliable practical performance.more » « less
-
Abstract Complex machine learning architectures and high-dimensional gridded input data are increasingly used to develop high-performance geoscience models, but model complexity obfuscates their decision-making strategies. Understanding the learned patterns is useful for model improvement or scientific investigation, motivating research in eXplainable artificial intelligence (XAI) methods. XAI methods often struggle to produce meaningful explanations of correlated features. Gridded geospatial data tends to have extensive autocorrelation so it is difficult to obtain meaningful explanations of geoscience models. A recommendation is to group correlated features and explain those groups. This is becoming common when using XAI to explain tabular data. Here, we demonstrate that XAI algorithms are highly sensitive to the choice of how we group raster elements. We demonstrate that reliance on a single partition scheme yields misleading explanations. We propose comparing explanations from multiple grouping schemes to extract more accurate insights from XAI. We argue that each grouping scheme probes the model in a different way so that each asks a different question of the model. By analyzing where the explanations agree and disagree, we can learn information about the scale of the learned features. FogNet, a complex three-dimensional convolutional neural network for coastal fog prediction, is used as a case study for investigating the influence of feature grouping schemes on XAI. Our results demonstrate that careful consideration of how each grouping scheme probes the model is key to extracting insights and avoiding misleading interpretations.more » « less
-
Abstract Prediction of the spatial‐temporal dynamics of the fluid flow in complex subsurface systems, such as geologic storage, is typically performed using advanced numerical simulation methods that solve the underlying governing physical equations. However, numerical simulation is computationally demanding and can limit the implementation of standard field management workflows, such as model calibration and optimization. Standard deep learning models, such as RUNET, have recently been proposed to alleviate the computational burden of physics‐based simulation models. Despite their powerful learning capabilities and computational appeal, deep learning models have important limitations, including lack of interpretability, extensive data needs, weak extrapolation capacity, and physical inconsistency that can affect their adoption in practical applications. We develop a Fluid Flow‐based Deep Learning (FFDL) architecture for spatial‐temporal prediction of important state variables in subsurface flow systems. The new architecture consists of a physics‐based encoder to construct physically meaningful latent variables, and a residual‐based processor to predict the evolution of the state variables. It uses physical operators that serve as nonlinear activation functions and imposes the general structure of the fluid flow equations to facilitate its training with data pertaining to the specific subsurface flow application of interest. A comprehensive investigation of FFDL, based on a field‐scale geologic storage model, is used to demonstrate the superior performance of FFDL compared to RUNET as a standard deep learning model. The results show that FFDL outperforms RUNET in terms of prediction accuracy, extrapolation power, and training data needs.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1093/qje/qjac015
