Closed nijawa closed 2 months ago
The values of step_size
and dt
are the same during simulation, at least when using a mio::ssirs::(Flow)Simulation
. See models/sde_sirs/simulation.h:63
and :124
.
I think the problem is that with the current parameters the flows have pretty extreme values, and can sometimes even get negative values. Clamping to 0 fixes the NaNs at least, here is some output:
0.81800 0.56708 8442.04686 1557.38607
0.81900 0.34570 8442.05194 1557.60237
0.82000 0.37208 8441.09590 1558.53202
0.82100 0.40253 8440.00211 1559.59536
0.82200 0.57070 8439.48523 1559.94408
0.82300 0.68348 8439.53834 1559.77818
0.82400 0.44854 8438.97716 1560.57430
0.82500 0.43399 8438.46640 1561.09961
0.82600 0.42257 8438.52051 1561.05692
0.82700 0.26341 8438.06443 1561.67215
0.82800 0.33228 8438.03298 1561.63474
0.82900 0.31843 8436.41828 1563.26329
0.83000 0.23948 8435.94474 1563.81578
0.83100 0.19645 8433.88955 1565.91400
0.83200 0.19920 8433.33963 1566.46117
0.83300 0.19920 8431.30414 1568.49666
Bug description
The SDE Models have a member variable step_size controlling the step size of the stochastic process. The function simulate also takes a function parameter dt which controls the step size of the integrator core. Those two can be different. This can result in NAN output (and the result is not what you would expect from the SDE model)
Version
Windows
To reproduce
Relevant log output
No response
Add any relevant information, e.g. used compiler, screenshots.
No response
Checklist