mirror of
https://github.com/gsi-upm/soil
synced 2024-11-22 11:12:29 +00:00
feab0ba79e
The examples weren't being properly tested in the last commit. When we fixed that a lot of bugs in the new implementation of environment and agent were found, which accounts for most of these changes. The main difference is the mechanism to load simulations from a configuration file. For that to work, we had to rework our module loading code in `serialization` and add a `source_file` attribute to configurations (and simulations, for that matter).
134 lines
4.2 KiB
Python
134 lines
4.2 KiB
Python
from soil.agents import FSM, NetworkAgent, state, default_state, prob
|
|
from soil.parameters import *
|
|
import logging
|
|
|
|
from soil.environment import Environment
|
|
|
|
|
|
class DumbViewer(FSM, NetworkAgent):
|
|
"""
|
|
A viewer that gets infected via TV (if it has one) and tries to infect
|
|
its neighbors once it's infected.
|
|
"""
|
|
|
|
has_been_infected: bool = False
|
|
has_tv: bool = False
|
|
|
|
@default_state
|
|
@state
|
|
def neutral(self):
|
|
if self.has_tv:
|
|
if self.prob(self.get("prob_tv_spread")):
|
|
return self.infected
|
|
if self.has_been_infected:
|
|
return self.infected
|
|
|
|
@state
|
|
def infected(self):
|
|
for neighbor in self.get_neighbors(state_id=self.neutral.id):
|
|
if self.prob(self.get("prob_neighbor_spread")):
|
|
neighbor.infect()
|
|
|
|
def infect(self):
|
|
"""
|
|
This is not a state. It is a function that other agents can use to try to
|
|
infect this agent. DumbViewer always gets infected, but other agents like
|
|
HerdViewer might not become infected right away
|
|
"""
|
|
self.has_been_infected = True
|
|
|
|
|
|
class HerdViewer(DumbViewer):
|
|
"""
|
|
A viewer whose probability of infection depends on the state of its neighbors.
|
|
"""
|
|
|
|
def infect(self):
|
|
"""Notice again that this is NOT a state. See DumbViewer.infect for reference"""
|
|
infected = self.count_neighbors(state_id=self.infected.id)
|
|
total = self.count_neighbors()
|
|
prob_infect = self.get("prob_neighbor_spread") * infected / total
|
|
self.debug("prob_infect", prob_infect)
|
|
if self.prob(prob_infect):
|
|
self.has_been_infected = True
|
|
|
|
|
|
class WiseViewer(HerdViewer):
|
|
"""
|
|
A viewer that can change its mind.
|
|
"""
|
|
|
|
@state
|
|
def cured(self):
|
|
prob_cure = self.get("prob_neighbor_cure")
|
|
for neighbor in self.get_neighbors(state_id=self.infected.id):
|
|
if self.prob(prob_cure):
|
|
try:
|
|
neighbor.cure()
|
|
except AttributeError:
|
|
self.debug("Viewer {} cannot be cured".format(neighbor.id))
|
|
|
|
def cure(self):
|
|
self.has_been_cured = True
|
|
|
|
@state
|
|
def infected(self):
|
|
if self.has_been_cured:
|
|
return self.cured
|
|
cured = max(self.count_neighbors(self.cured.id), 1.0)
|
|
infected = max(self.count_neighbors(self.infected.id), 1.0)
|
|
prob_cure = self.get("prob_neighbor_cure") * (cured / infected)
|
|
if self.prob(prob_cure):
|
|
return self.cured
|
|
|
|
|
|
class NewsSpread(Environment):
|
|
ratio_dumb: probability = 1,
|
|
ratio_herd: probability = 0,
|
|
ratio_wise: probability = 0,
|
|
prob_tv_spread: probability = 0.1,
|
|
prob_neighbor_spread: probability = 0.1,
|
|
prob_neighbor_cure: probability = 0.05,
|
|
|
|
def init(self):
|
|
self.populate_network([DumbViewer, HerdViewer, WiseViewer],
|
|
[self.ratio_dumb, self.ratio_herd, self.ratio_wise])
|
|
|
|
|
|
from itertools import product
|
|
from soil import Simulation
|
|
|
|
|
|
# We want to investigate the effect of different agent distributions on the spread of news.
|
|
# To do that, we will run different simulations, with a varying ratio of DumbViewers, HerdViewers, and WiseViewers
|
|
# Because the effect of these agents might also depend on the network structure, we will run our simulations on two different networks:
|
|
# one with a small-world structure and one with a connected structure.
|
|
|
|
counter = 0
|
|
for [r1, r2] in product([0, 0.5, 1.0], repeat=2):
|
|
for (generator, netparams) in {
|
|
"barabasi_albert_graph": {"m": 5},
|
|
"erdos_renyi_graph": {"p": 0.1},
|
|
}.items():
|
|
print(r1, r2, 1-r1-r2, generator)
|
|
# Create new simulation
|
|
netparams["n"] = 500
|
|
Simulation(
|
|
name='newspread_sim',
|
|
model=NewsSpread,
|
|
model_params=dict(
|
|
ratio_dumb=r1,
|
|
ratio_herd=r2,
|
|
ratio_wise=1-r1-r2,
|
|
network_generator=generator,
|
|
network_params=netparams,
|
|
prob_neighbor_spread=0,
|
|
),
|
|
num_trials=5,
|
|
max_steps=300,
|
|
dump=False,
|
|
).run()
|
|
counter += 1
|
|
# Run all the necessary instances
|
|
|
|
print(f"A total of {counter} simulations were run.") |