IDS-ALGES Collaboration

Information for Decision Making

At ALGES (Advanced Laboratory for Geostatistical Supercomputing), we have built a longstanding and fruitful collaboration with the Information and Decision Systems Group (IDS) of the Department of Electrical Engineering at the University of Chile.

Together they have developed research in statistical learning, in generative models for sensory data analysis and in Bayesian and uncertainty modelling for decision making. This partnership has allowed us to jointly explore and develop advanced methodologies at the intersection of statistical learning, signal processing, and geoscientific data analysis.

A core focus of our collaboration has been on hyperspectral image modeling and classification, where we have combined generative and discriminative techniques to better understand and process sensory data. One of our joint contributions in this area proposed a discriminative generative framework for hyperspectral image classification, demonstrating strong performance in complex material datasets (Egaña et al., 2024).

Together, we have also addressed challenges in understanding deep learning architectures, using information-theoretic tools to study the structure and performance of encoder–decoder models in machine learning (Silva et al., 2025; Faraggi et al., 2024).

In the field of geometallurgy, we introduced a method that integrates hyperspectral imaging with statistical topic modeling to estimate mineralogical characteristics, offering a data-driven approach to mineral sample analysis (Santibáñez-Leal et al., 2022).

Our collaboration has extended to the classification and analysis of geological textures, including the use of variogram-based descriptors for rock image comparison (Díaz et al., 2020) and transform-based features for classifying natural rock textures (Lobos et al., 2015).

We have also co-developed innovative strategies for reconstructing channelized geological facies using compressed sensing techniques, improving the efficiency and accuracy of geospatial reconstruction methods (Calderon et al., 2015; Calderon et al., 2016).

Our earliest joint efforts included contributions to signal separation techniques for hyperspectral imagery, pioneered in the thesis work of Sergio Liberman, conducted under the guidance of researchers from both ALGES and IDS (Liberman et al., 2016).

This collaboration continues to be a key pillar of our research ecosystem, allowing us to bridge foundational signal processing theory with real-world mining and geoscientific applications.