Dr Bubacarr Bah’s research spans foundational mathematical theory and practical data science methods, with applications in machine learning, signal processing, and public health. His work combines rigorous analysis with real-world impact, particularly in low- and middle-income settings, and has led to contributions across multiple interconnected research areas outlined below.
DATA SCIENCE
Recent methodological advances in federated learning for healthcare [LINK]
A protocol for a scoping study of economic and data system considerations for climate [LINK] change and pandemic preparedness in Africa
Gaps and opportunities for data systems and economics to support priority setting for climate-sensitive infectious diseases in sub-Saharan Africa [LINK]
Study protocol for an observational cohort study of heat stress impacts in pregnancy in The Gambia [LINK]
ARTIFICIAL INTELLIGENCE
Learning deep linear neural networks: Riemannian gradient flows and convergence to global minimizers [LINK]
Efficient and robust mixed-integer optimization methods for training binarized deep neural networks [LINK]
An integer programming approach to deep neural networks with binary activation functions [LINK]
Proto-BagNets for local and global interpretability-by-design [LINK]
COMPRESSED SENSING
Vanishingly sparse matrices and expander graphs, with application to compressed sensing [LINK]
Improved bounds on restricted isometry constants for Gaussian matrices [LINK]
Bounds of restricted isometry constants in extreme asymptotics [LINK]
Compressed sensing-based SARS-CoV-2 pool testing [LINK]
SIGNAL PROCESSING
The sample complexity of weighted sparse approximation [LINK]
Efficient noise-blind ℓ1-regression of nonnegative compressible signals [LINK]
Efficient tuning-free ℓ1-regression of nonnegative compressible signal [LINK]
Recent developments in signal approximation and reconstruction [LINK]
NUMERICAL LINEAR ALGEBRA
Improved bounds on restricted isometry constants for Gaussian matrices [LINK]
Bounds of restricted isometry constants in extreme asymptotics [LINK]
Low-rank matrix approximation for imputing missing categorical data [LINK]
RANDOM MATRIX THEORY
Improved bounds on restricted isometry constants for Gaussian matrices [LINK]
Restricted isometry constants in compressed sensing [LINK]
Restricted isometry property (RIP) [LINK]
SCIENTIFIC COMPUTING
Diffusion maps: analysis and applications [LINK]
Physics-informed neural networks for Timoshenko systems with thermoelasticity [LINK]
A physics-informed neural network framework for modeling obstacle-related equations [LINK]
INVERSE PROBLEMS
Towards the localisation of lesions in diabetic retinopathy [LINK]
Improving identification accuracy in equation learning via comprehensive R²-elimination and Bayesian model selection [LINK]
Outcome prediction in disorders of consciousness using machine learning [LINK]
SPARSE APPROXIMATION
Weighted sparse recovery with expanders [LINK]
Convex block-sparse linear regression with expanders [LINK]
Model-based sketching and recovery with expanders [LINK]
Discrete optimization methods for group model selection in compressed sensing [LINK]