A key concern for building your own environment model is how to define your own capacity growth model in Environment.
To do so, a simple way is to override the learn function in Learner
and rewrite the generate_learners function in Environment.
For Example, Learner of KSS/Learner is:
from EduSim.Envs.Learner import Learner as BaseLearner
class Learner(BaseLearner):
def __init__(*args, **kwargs):
...
def learn(self, learning_item: int):
structure = self.structure
a = self._state
if self.learning_history:
if learning_item not in influence_control(
structure, a, self.learning_history[-1], allow_shortcut=False, target=self._target,
)[0]:
return
assert isinstance(learning_item, int), learning_item
self.learning_history.append(learning_item)
# capacity growth function
discount = math.exp(sum([(5 - a[node]) for node in structure.predecessors(learning_item)] + [0]))
ratio = 1 / discount
inc = (5 - a[learning_item]) * ratio * 0.5
def _promote(_ind, _inc):
a[_ind] += _inc
if a[_ind] > 5:
a[_ind] = 5
for node in structure.successors(_ind):
_promote(node, _inc * 0.5)
_promote(learning_item, inc)
...And you can implement a new Learner class inheriting KSS/Learner and override the learn function:
from EduSim.Envs.KSS.meta.Learner import Learner
class NewLearner(Learner):
def learn(self, learning_item: int):
...After that, you need to override generate_learners in Environment, that is
from EduSim.Envs.KSS import KSS
class newKSS(KSS):
def generate_learners(self, learner_num, step=20):
learner = NewLearner(...)
...