house advertising warning! Just in case you guys e...
# datascience
h
house advertising warning! Just in case you guys ever feel the urge to discuss how to create a discrete event simulation with kotlin --> #kalasim
👀 1
👍 3
a
We are doing a lot simulations, but they are mostly continous.
h
What framework are you working with to do so?
a
Mostly naked kotlin coroutines. We plan to use Kmath for advanced things and geomtry, but so far we even managed without it.
h
Nice. Thanks for sharing.
a
here is the latest article: https://arxiv.org/abs/2011.07858 and the code for it: https://github.com/altavir/trapping And here is the simulation for not yet published work: https://github.com/mipt-npm/rl-tge-sim
h
Cool, i'll have a look even without knowing much about physics.😊
a
Also there is a video about very simplistic educational descreete Monte-Carlo epidemics model here:

https://youtu.be/dW6df7SdX-4

, but it is in Russian. You can see the code though
Actually, we are currently working on a coroutine-based Monte-Carlo engine. @Ролан will lead the project this year.
I think that the Flow itself is a very good model for even-based simulations.
Probably better to continue discussion in your channel
h
Cool, I actually have implemented a first epidemics model in https://github.com/holgerbrandl/kalasim/blob/master/simulations/covid19/src/main/kotlin/immune/covid19immune.kt as well and plan to continue this fun line of hobbyist research.
I heard about Flow but somehow have managed to avoid this tricky (for me) topic so far.
a
I've done more advanced model, which includes behavior patterns and probability distributions for all involved processes. It actuallly could produce realistic results, but I do not have the data for it of course.
If you want, we can do a master-class about using it in models sometimes soon.
h
Sure, diving deeper into DES is so much fun.
f
Very interesting project! We are doing discrete-event simulation of datacenters using Kotlin: https://github.com/atlarge-research/opendc By leveraging co-routines (using a custom, “time-traveling” dispatcher), we can separate most of the simulation logic from the actual business logic.