skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Explainable AI via learning to optimize
Abstract Indecipherable black boxes are common in machine learning (ML), but applications increasingly require explainable artificial intelligence (XAI). The core of XAI is to establish transparent and interpretable data-driven algorithms. This work provides concrete tools for XAI in situations where prior knowledge must be encoded and untrustworthy inferences flagged. We use the “learn to optimize” (L2O) methodology wherein each inference solves a data-driven optimization problem. Our L2O models are straightforward to implement, directly encode prior knowledge, and yield theoretical guarantees (e.g. satisfaction of constraints). We also propose use of interpretable certificates to verify whether model inferences are trustworthy. Numerical examples are provided in the applications of dictionary-based signal recovery, CT imaging, and arbitrage trading of cryptoassets. Code and additional documentation can be found athttps://xai-l2o.research.typal.academy.  more » « less
Award ID(s):
2309810
PAR ID:
10504901
Author(s) / Creator(s):
;
Publisher / Repository:
Springer Nature
Date Published:
Journal Name:
Scientific Reports
Volume:
13
Issue:
1
ISSN:
2045-2322
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; (, BMC Bioinformatics)
    Abstract We describe POInTbrowse, a web portal that gives access to the orthology inferences made for polyploid genomes with POInT, the Polyploidy Orthology Inference Tool. Ancient, or paleo-, polyploidy events are widely distributed across the eukaryotic phylogeny, and the combination of duplicated and lost duplicated genes that these polyploidies produce can confound the identification of orthologous genes between genomes. POInT uses conserved synteny and phylogenetic models to infer orthologous genes between genomes with a shared polyploidy. It also gives confidence estimates for those orthology inferences. POInTbrowsegives both graphical and query-based access to these inferences from 12 different polyploidy events, allowing users to visualize genomic regions produced by polyploidies and perform batch queries for each polyploidy event, downloading genes trees and coding sequences for orthologous genes meeting user-specified criteria. POInTbrowseand the associated data are online athttps://wgd.statgen.ncsu.edu. 
    more » « less
  2. ; ; ; ; ; ; ; ; (, Journal of Geophysical Research: Atmospheres)
    Abstract We report on the mountain top observation of three terrestrial gamma‐ray flashes (TGFs) that occurred during the summer storm season of 2021. To our knowledge, these are the first TGFs observed in a mountaintop environment and the first published European TGFs observed from the ground. A gamma‐ray sensitive detector was located at the base of the Säntis Tower in Switzerland and observed three unique TGF events with coincident radio sferic data characteristic of TGFs seen from space. We will show an example of a “slow pulse” radio signature (Cummer et al., 2011,https://doi.org/10.1029/2011GL048099; Lu et al., 2011,https://doi.org/10.1029/2010JA016141; Pu et al., 2019,https://doi.org/10.1029/2019GL082743; Pu et al., 2020,https://doi.org/10.1029/2020GL089427), a −EIP (Lyu et al., 2016,https://doi.org/10.1002/2016GL070154; Lyu et al., 2021,https://doi.org/10.1029/2021GL093627; Wada et al., 2020,https://doi.org/10.1029/2019JD031730), and a double peak TGF associated with an extraordinarily powerful and complicated positive‐polarity sferic, where each TGF peak is possibly preceded by a short burst of stepped leader emission. 
    more » « less
  3. ; (, Nature Methods)
    Abstract Nonnegative matrix factorization (NMF) is widely used to analyze high-dimensional count data because, in contrast to real-valued alternatives such as factor analysis, it produces an interpretable parts-based representation. However, in applications such as spatial transcriptomics, NMF fails to incorporate known structure between observations. Here, we present nonnegative spatial factorization (NSF), a spatially-aware probabilistic dimension reduction model based on transformed Gaussian processes that naturally encourages sparsity and scales to tens of thousands of observations. NSF recovers ground truth factors more accurately than real-valued alternatives such as MEFISTO in simulations, and has lower out-of-sample prediction error than probabilistic NMF on three spatial transcriptomics datasets from mouse brain and liver. Since not all patterns of gene expression have spatial correlations, we also propose a hybrid extension of NSF that combines spatial and nonspatial components, enabling quantification of spatial importance for both observations and features. A TensorFlow implementation of NSF is available fromhttps://github.com/willtownes/nsf-paper. 
    more » « less
  4. ; ; ; ; ; ; ; (, Journal of Geophysical Research: Space Physics)
    Abstract We compared the performance of DREAM3D simulations in reproducing the long‐term radiation belt dynamics observed by Van Allen Probes over the entire year of 2017 with various boundary conditions (BCs) and model inputs. Specifically, we investigated the effects of three different outer boundary conditions, two different low‐energy boundary conditions for seed electrons, four different radial diffusion (RD) coefficients (DLL), four hiss wave models, and two chorus wave models from the literature. Using the outer boundary condition driven by GOES data, our benchmark simulation generally well reproduces the observed radiation belt dynamics insideL* = 6, with a better model performance at lowerμthan higherμ, whereμis the first adiabatic invariant. By varying the boundary conditions and inputs, we find that: (a) The data‐driven outer boundary condition is critical to the model performance, while adding in the data‐driven seed population doesn't further improve the performance. (b) The model shows comparable performance withDLLfrom Brautigam and Albert (2000,https://doi.org/10.1029/1999ja900344), Ozeke et al. (2014,https://doi.org/10.1002/2013ja019204), and Liu et al. (2016,https://doi.org/10.1002/2015gl067398), while withDLLfrom Ali et al. (2016,https://doi.org/10.1002/2016ja023002) the model shows less RD compared to data. (c) The model performance is similar with data‐based hiss models, but the results show faster loss is still needed inside the plasmasphere. (d) The model performs similarly with the two different chorus models, but better capturing the electron enhancement at higherμusing the Wang et al. (2019,https://doi.org/10.1029/2018ja026183) model due to its stronger wave power, since local heating for higher energy electrons is under‐reproduced in the current model. 
    more » « less
  5. ; ; ; ; (, Journal of Geophysical Research: Solid Earth)
    Abstract A foundational assumption in paleomagnetism is that the Earth's magnetic field behaves as a geocentric axial dipole (GAD) when averaged over sufficient timescales. Compilations of directional data averaged over the past 5 Ma yield a distribution largely compatible with GAD, but the distribution of paleointensity data over this timescale is incompatible. Reasons for the failure of GAD include: (a) Arbitrary “selection criteria” to eliminate “unreliable” data vary among studies, so the paleointensity database may include biased results. (b) The age distribution of existing paleointensity data varies with latitude, so different latitudinal averages represent different time periods. (c) The time‐averaged field could be truly non‐dipolar. Here, we present a consistent methodology for analyzing paleointensity results and comparing time‐averaged paleointensities from different studies. We apply it to data from Plio/Pleistocene Hawai'ian igneous rocks, sampled from fine‐grained, quickly cooled material (lava flow tops, dike margins and scoria cones) and subjected to the IZZI‐Thellier technique; the data were analyzed using the Bias Corrected Estimation of Paleointensity method of Cych et al. (2021,https://doi.org/10.1029/2021GC009755), which produces accurate paleointensity estimates without arbitrarily excluding specimens from the analysis. We constructed a paleointensity curve for Hawai'i over the Plio/Pleistocene using the method of Livermore et al. (2018,https://doi.org/10.1093/gji/ggy383), which accounts for the age distribution of data. We demonstrate that even with the large uncertainties associated with obtaining a mean field from temporally sparse data, our average paleointensities obtained from Hawai'i and Antarctica (reanalyzed from Asefaw et al., 2021,https://doi.org/10.1029/2020JB020834) are not GAD‐like from 0 to 1.5 Ma but may be prior to that. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email