neural-sde

Parallel Computing and Scientific Machine Learning (SciML): Methods and Applications (MIT 18.337J/6.338J)

Pre-built implicit layer architectures with O(1) backprop, GPUs, and stiff+non-stiff DE solvers, demonstrating scientific machine learning (SciML) and physics-informed machine learning methods

A component of the DiffEq ecosystem for enabling sensitivity analysis for scientific machine learning (SciML). Optimize-then-discretize, discretize-then-optimize, adjoint methods, and more for ODEs, SDEs, DDEs, DAEs, etc.

Code for "Infinitely Deep Bayesian Neural Networks with Stochastic Differential Equations"

A library of noise processes for stochastic systems like stochastic differential equations (SDEs) and other systems that are present in scientific machine learning (SciML)

A component of the SciML scientific machine learning ecosystem for optimal control

Implementation of "Neural Jump-Diffusion Temporal Point Processes" (ICML 2024 Spotlight)

Repository for the Control of Stochastic Quantum Dynamics with Differentiable Programming paper.

A thin wrapper over Bridge.jl for the SciML scientific machine learning common interface, enabling new methods for neural stochastic differential equations (neural SDEs)

Work in progress of Finite Dimensional Matching for Neural SDEs implementation.

Independent R&D bridging classical financial econometrics and modern continuous-time deep learning. Projects on PINNs for Value-at-Risk and Neural SDEs for density forecasting. "Complexity must earn its place."

Neural SDE Climate Risk Framework A PyTorch implementation using Variational Autoencoders (VAEs) and Neural Stochastic Differential Equations (SDEs) to model continuous-time corporate default risk and simulate the impact of climate transition shocks on high-carbon sectors.

A PyTorch framework for training neural stochastic differential equations (SDEs) to model dynamical systems with both deterministic and stochastic components. This framework enables the learning of dynamical systems through a combination of neural networks representing drift and diffusion terms, with adversarial training. Docs ⬇️

基于 Neural SDE（神经网络随机微分方程）的中证1000股指期权定价框架：MLE 训练 + Girsanov 测度变换 + MC 定价，与 BS/GBM/GARCH 对比

Sky-camera solar power nowcasting. One small latent stochastic differential equation turns a sky image into a calibrated forecast distribution for every horizon from 1 to 30 minutes.

Physics-informed latent neural SDEs for surrogate modelling of industrial process systems, in JAX.

Geometric neuroplasticity for LLMs: a closed self-regulation loop — Fisher (K-FAC) -> thermodynamic gating -> liquid-time SDE -> natural-gradient adaptation -> drift detection. Honest, measured research.

Neural SDE framework for rough volatility modeling (H ≈ 0.1) with deep hedging. Implements Davies-Harte fBM, signature-based losses, and convergence analysis.

Latent neural SDEs with exact first-passage likelihoods via Kolmogorov PDEs (JAX/Diffrax)