Kendrick Simulator can not work in large population scale free network

[thanhbka85dn]

Implementing Bansal's idea for Kendrick simulator works well in scale-free network with SMALL population.
I run the test case with 100 individuals and it prints a nice graph:

But as trying larger population, such as 1000 or 10000 individuals, the simulator leaks memory and not responding at all.

I will take care of this case.

[thanhbka85dn]

After fixing, now IBM Simulator can work with 1000 individuals of Scale-free network (k mean degree is 10), but it still takes too long to finish (about 1 hour)

[thanhbka85dn]

In case 10000 individuals, I reduce the maximum of time to 10 days and change the step to 0.5; the simulator takes 2 hours to complete the graph:

[thanhbka85dn]

I have fixed this issue, the main problem is - I guess - the time complexity of saveCompartmentsValueAtTime: O(N*N) - order N square.
As looking at the current version of this method:

model population compartment keysAndValuesDo will loop N times (N is number of compartment). In case 10000 individuals and three status #(S, I, R), N = 10000 x 3 = 30000.
The timeSeriesOutputs detect will loop n times with n ranges from 1 to N.
We have: 1 + 2 + ... + N = N*(N+1)/2 about N*N/2 times. If N = 30000, the simulator has no choice but to loop 450 million times. The situation will become worse as epidemiologist uses other compartments (like ages, genders, countries...).

Therefore, I suggest the code above is replaced by:

Assuming that (model population compartments) at: (e compartment) requires a constant time, the new version will take O(N) because timeSeriesOutputs do loops N times.

Last but not least, I increase the step of simulator to 0.5 (instead of 0.1), and it takes about 3 hours to finish.

Finally, I got this result:

[SergeStinckwich]

Integrate code from : master...thanhbka85dn:master