참고 :

LEARNING REDUCED SYSTEMS VIA DEEP NEURAL NETWORKS WITH MEMORY

1. 요약

2. Main Method

$$ \frac{dz(t)}{dt} = R(z(t)) + \int_{0}^{t}K(z(t-s),s)ds + F(x_0,t) $$

(1) Finite Memory Approximation

$$ \bigg| \int_{0}^{t}K(z(t-s),s)ds - \int_{0}^{T_M}K(z(t-s),s)ds \bigg| \leq \epsilon $$

(2) Discrete-Finite Memory Approximation

$$ \bigg| M(\hat z_{n-n_M}, ..., \hat z_{n-1}, \hat z_n) - \int_{0}^{T_M}K(\hat z(t_n-s),s)ds \bigg| \leq \eta(t_n;T_M,n_M) $$