FS-PIELM shifts the mean of Gaussian weights (variance fixed at 1) in PIELM to bound frequency variance and achieve 1-5 orders of magnitude better accuracy on high-frequency PDE benchmarks while retaining single linear solve efficiency.
Title resolution pending
9 Pith papers cite this work. Polarity classification is still indexing.
9
Pith papers citing it
years
2026 9representative citing papers
Quantum collocation framework for 1D linear and nonlinear BVPs that uses a residual-threshold oracle on joint spatial-parameter registers to produce coherent superposition of spatially conditioned amplitude amplification with poly-log gate complexity.
Lagrangian Gaussian Processes use discrete Euler-Lagrange equations to condition GPs, preserving geometric structure for stable dynamics learning from sparse position snapshots without velocities.
Dimensional types that persist through MLIR lowering jointly drive numeric representation selection and deterministic memory allocation as coeffects on a single program semantic graph.
A verification-first agentic workflow for SciML surrogate discovery adds per-candidate, machine-checkable physics audits that expose a causality failure an error-only baseline misses.
A fixed time-encoding, not barren-plateau gradient vanishing, sets a 33-dimensional capacity ceiling (11 accessible Fourier modes per observable) that prevents a four-qubit variational QPINN from fitting Lorenz-63.
Design-time Hindley-Milner unification over finitely generated abelian groups is claimed to verify AI model reliability properties and to compute a MAP hypothesis under a restricted Solomonoff prior.
An auxiliary finite-difference regularizer on residual gradients improves outer-wall flux and boundary-condition accuracy in a 3D annular heat-conduction PINN benchmark when aligned with the quantity of interest.
Adding a soft mass-conservation penalty to PINNs for the 1D advection-diffusion equation reduces long-term relative L2 error by 9–67× and mass error by 15–215× compared to vanilla PINNs across Peclet numbers 0.01–20.
citing papers explorer
-
Frequency Shift Physics-Informed Extreme Learning Machine for Solving High-Frequency Partial Differential Equations
FS-PIELM shifts the mean of Gaussian weights (variance fixed at 1) in PIELM to bound frequency variance and achieve 1-5 orders of magnitude better accuracy on high-frequency PDE benchmarks while retaining single linear solve efficiency.
-
A Quantum Collocation Approach to One-Dimensional Boundary Value Problems with Coherent Amplitude Amplification
Quantum collocation framework for 1D linear and nonlinear BVPs that uses a residual-threshold oracle on joint spatial-parameter registers to produce coherent superposition of spatially conditioned amplitude amplification with poly-log gate complexity.
-
Structure-Preserving Gaussian Processes Via Discrete Euler-Lagrange Equations
Lagrangian Gaussian Processes use discrete Euler-Lagrange equations to condition GPs, preserving geometric structure for stable dynamics learning from sparse position snapshots without velocities.
-
Dimensional Type Systems and Deterministic Memory Management: Design-Time Semantic Preservation in Native Compilation
Dimensional types that persist through MLIR lowering jointly drive numeric representation selection and deterministic memory allocation as coeffects on a single program semantic graph.
-
Physics-Audited Agentic Discovery in Scientific Machine Learning
A verification-first agentic workflow for SciML surrogate discovery adds per-candidate, machine-checkable physics audits that expose a causality failure an error-only baseline misses.
-
An architectural capacity ceiling, not a barren plateau: why a fixed-encoding variational quantum circuit cannot fit the Lorenz-63 attractor
A fixed time-encoding, not barren-plateau gradient vanishing, sets a 33-dimensional capacity ceiling (11 accessible Fourier modes per observable) that prevents a four-qubit variational QPINN from fitting Lorenz-63.
-
Decidable By Construction: Design-Time Verification for Trustworthy AI
Design-time Hindley-Milner unification over finitely generated abelian groups is claimed to verify AI model reliability properties and to compute a MAP hypothesis under a restricted Solomonoff prior.
-
Auxiliary Finite-Difference Residual-Gradient Regularization for PINNs
An auxiliary finite-difference regularizer on residual gradients improves outer-wall flux and boundary-condition accuracy in a 3D annular heat-conduction PINN benchmark when aligned with the quantity of interest.
-
Mass-Conserving Physics-Informed Neural Networks For The One-Dimensional Advection-Diffusion Equation
Adding a soft mass-conservation penalty to PINNs for the 1D advection-diffusion equation reduces long-term relative L2 error by 9–67× and mass error by 15–215× compared to vanilla PINNs across Peclet numbers 0.01–20.