notesum.ai

Published at November 4

ControlSynth Neural ODEs: Modeling Dynamical Systems with Guaranteed Convergence

cs.LG

cs.AI

cs.SY

eess.SY

Released Date: November 4, 2024

Authors: Wenjie Mei¹, Dongzhe Zheng¹, Shihua Li¹

Aff.: ¹Institution not specified

Arxiv: http://arxiv.org/abs/2411.02292v1

Refer to caption

Group	Model	Hindmarsh-Rose			Reaction-Diffusion			Shallow Water
Group	Model	MSE	MAE	$\mathbf{R}^{\textbf{2}}$	MSE	MAE	CD	MSE	MAE	CD
A	MLP	2.394	1.033	0.241	7.431e-2	0.398	7.745	0.135	0.35	3.17
A	RNN	1.975	0.871	0.356	2.376e-2	0.143	3.115	0.103	0.26	3.06
B	NODE	1.551	0.682	0.515	1.134e-2	0.081	1.311	0.071	0.22	2.89
	ANODE	0.745	0.586	0.637	1.095e-2	0.076	1.392	0.067	0.19	2.88
	SONODE	0.739	0.561	0.611	8.145e-3	0.056	1.315	0.065	0.19	2.83
	CSODE (Ours)	0.783	0.470	0.758	6.365e-3	0.058	0.939	0.041	0.16	2.27
C	CSODE-Adapt (Ours)	0.408	0.370	0.887	5.635e-3	0.035	0.837	0.031	0.15	1.89