Abstract:With the advent of artificial intelligence (AI) and machine learning (ML), various domains of science and engineering communites has leveraged data-driven surrogates to model complex systems from numerous sources of information (data). The proliferation has led to significant reduction in cost and time involved in development of superior systems designed to perform specific functionalities. A high proposition of such surrogates are built extensively fusing multiple sources of data, may it be published papers, patents, open repositories, or other resources. However, not much attention has been paid to the differences in quality and comprehensiveness of the known and unknown underlying physical parameters of the information sources that could have downstream implications during system optimization. Towards resolving this issue, a multi-source data fusion framework based on Latent Variable Gaussian Process (LVGP) is proposed. The individual data sources are tagged as a characteristic categorical variable that are mapped into a physically interpretable latent space, allowing the development of source-aware data fusion modeling. Additionally, a dissimilarity metric based on the latent variables of LVGP is introduced to study and understand the differences in the sources of data. The proposed approach is demonstrated on and analyzed through two mathematical (representative parabola problem, 2D Ackley function) and two materials science (design of FeCrAl and SmCoFe alloys) case studies. From the case studies, it is observed that compared to using single-source and source unaware ML models, the proposed multi-source data fusion framework can provide better predictions for sparse-data problems, interpretability regarding the sources, and enhanced modeling capabilities by taking advantage of the correlations and relationships among different sources.

B-scaling: A Novel Nonparametric Data Fusion Method

Scalable data fusion via a scale-based hierarchical framework: Adapting to multi-source and multi-scale scenarios

Similarity network fusion based on local scaling affinity construction

Data fusion using weakly aligned sources

A Novel Metabolomic Data Scaling Method Based on K-L Divergence

Biclustering Analysis of Functionals Via Penalized Fusion

Integrative Factorization of Bidimensionally Linked Matrices

Nonparametric fusion learning: synthesize inferences from diverse sources using depth confidence distribution

A novel multi-fidelity surrogate modeling method for non-hierarchical data fusion

Towards a Unified Theory for Semiparametric Data Fusion with Individual-Level Data

Calibrated regression estimation using empirical likelihood under data fusion

SCALE Method for Single-Cell ATAC-seq Analysis Via Latent Feature Extraction

Integrative analysis of individual-level data and high-dimensional summary statistics

A Multiscale Image Fusion Algorithm Based on Joint Distribution of Source Coefficients

A statistical framework for genomic data fusion

Data Fusion With Latent Map Gaussian Processes

A Novel Method for Fusion Operators Evaluating at Score-Level Fusion in Biometrie Authentication

A New Fusion Approach Based on Distance of Evidences

Interpretable Multi-Source Data Fusion Through Latent Variable Gaussian Process

Functional Integrative Bayesian Analysis of High-dimensional Multiplatform Genomic Data

A Novel Multi-Sensor Hybrid Fusion Framework