🏢 Max Planck Institute for Dynamics and Self-Organization

Deep learning models for neural activity lack reproducibility; this paper introduces adaptive regularization and iterative feature pruning to improve embedding consistency and predictive performance.