Research map / Optimization / Optimizers

Accelerating Neural Network Training with Augmented Koopman Dynamics

NeurIPS 2026 · Optimization › Optimizers · approach: Constrained and non-backprop training

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a training acceleration method that learns and extrapolates the neural network optimisation trajectory using augmented Koopman dynamics, reducing the cost of repeated backpropagation.

Problem

Neural network training is costly due to repeated backpropagation; the optimisation trajectory may be learnable and extrapolable to reduce this cost.

Approach

Views training as a discrete-time dynamical system and uses Koopman operator theory to represent its nonlinear evolution as linear dynamics in a lifted space, with augmented Koopman dynamics used for trajectory extrapolation instead of directly approximating the nonlinear trajectory.

Experimental setup

Builds on: Koopman operator theory

Closest papers by meaning

• Deep-Koopman-KANDy: Dictionary Discovery for Deep-Koopman Operators with Kolmogorov-Arnold Networks for Dynamics (NeurIPS 2026 · similarity 0.88)
• Efficient Parametric SVD of Koopman Operator for Stochastic Dynamical Systems (NeurIPS 2025 · similarity 0.87)
• SKOLR: Structured Koopman Operator Linear RNN for Time-Series Forecasting (ICML 2025 · similarity 0.87)
• Identifying Equivalent Training Dynamics (NeurIPS 2024 · similarity 0.87)
• KoNODE: Koopman-Driven Neural Ordinary Differential Equations with Evolving Parameters for Time Series Analysis (ICML 2025 · similarity 0.86)
• ResKoopNet: Learning Koopman Representations for Complex Dynamics with Spectral Residuals (ICML 2025 · similarity 0.86)