9 papers
Scalable Uncertainty Quantification for Extreme Weather Forecasting via Empirical Neural Tangent Kernels
Jose Marie Antonio Miñoza, Rex Gregor Laylo, Sebastian C. Ibañez
Deep learning weather models now match numerical weather prediction accuracy while running orders of magnitude faster, but produce deterministic forecasts without uncertainty estim…
The Hamilton-Jacobi Theory of Deep Learning
Jose Marie Antonio Miñoza, Jose Marie Antonio Miñoza, Erika Fille T. Legara +1
In this paper, training a neural network is identified, exactly, as a search through Hamilton--Jacobi initial-value problems: each gradient step selects the initial data of a visco…
eXplaining to Learn (eX2L): Regularization Using Contrastive Visual Explanation Pairs for Distribution Shifts
Paulo Mario P. Medina, Jose Marie Antonio Miñoza, Sebastian C. Ibañez
Despite extensive research into mitigating distribution shifts, many existing algorithms yield inconsistent performance, often failing to outperform baseline Empirical Risk Minimiz…
Linearized Attention Cannot Enter the Kernel Regime at Any Practical Width
Jose Marie Antonio Miñoza, Paulo Mario P. Medina, Sebastian C. Ibañez
Understanding whether attention mechanisms converge to the kernel regime is foundational to the validity of influence functions for transformer accountability. Exact NTK characteri…
UltraLIF: Fully Differentiable Spiking Neural Networks via Ultradiscretization and Max-Plus Algebra
Jose Marie Antonio Miñoza
Spiking Neural Networks (SNNs) offer energy-efficient, biologically plausible computation but suffer from non-differentiable spike generation, necessitating reliance on heuristic s…
SPIKE: Sparse Koopman Regularization for Physics-Informed Neural Networks
Jose Marie Antonio Miñoza
Physics-Informed Neural Networks (PINNs) provide a mesh-free approach for solving differential equations by embedding physical constraints into neural network training. However, PI…