Dandekar, Raj, and George Barbastathis. “Neural Network Aided Quarantine Control Model Estimation of Global Covid-19 Spread.” ArXiv:2004.02752 [Physics, q-Bio], Apr. 2020. arXiv.org, http://arxiv.org/abs/2004.02752.

Used a neural network to augment a SEIR model (pg 2)

Quarantine Strength as a function of time

SEIR/SIR models assume “homogenous mixing among the subpopulations” (pg 8)

Model 1: W/O quarantine control (pg 8-9)

Initial conditions:

I_0=500 cases

S_0=11,000,000 (Pop. of Wuhan)

E_0=20 x I_0

R_0≈10

, , and were optimized based off of available data

Model 2: With quarantine control (pg 9)

Initial Conditions:

I_0=500 cases

S_0=11,000,000 (Pop. of Wuhan)

E_0=20 x I_0

R_0≈10

This model used a multilayer neural network to adjust quarantine strength over time.