Multi-Modal Scientific Foundation Models and Sampling Multiplex Networks
9/12/2024 3:10pm
Derek Jollie
Consistent Symbolic Representation in Multi-Modal Scientific Foundation Models
Abstract:
Scientific foundation models necessitate the ability to maintain accuracy and precision
for prediction. This work implements an automatic process as a means of creating a
consistent standard tree for implementation in a scientific foundation model. We compared
against previous trees and tested stability against randomly added terms and noise
on the PROSE-PDE architecture. In addition, we will implement a Bayesian particle
filter to refine symbolic out-
puts utilizing the output data of PROSE.
Alec Fluer
Sampling Multiplex Networks Under the Constraints of Correlated Mean Degree Sequences and Social Distance Attachment
Abstract:
The presence of overlap between disparate layers of multiplex social networks motivates the need to develop a sampling algorithm that better accounts for survey data. We develop such an algorithm for sampling an ensemble of multiplex networks by extending the configuration model to leverage both correlated mean degree sequences and the notion of social distance as a probabilistic factor for edge formation. We examine the convergence of a sampled ensemble to estimate the representative accuracy of the proposed algorithm. This work will allow us to facilitate realistic simulation studies, and we present a discussion on future directions and applications to epidemiology.