180 lines
4.7 KiB
Python
180 lines
4.7 KiB
Python
from engine.utils import bound_norm_random, ep_time_step, env_proc
|
|
|
|
import numpy as np
|
|
from decimal import Decimal
|
|
|
|
alpha = Decimal('.7') #forgetting param
|
|
theta = Decimal('.75') #weight param for rational price
|
|
beta = Decimal('0.5') #agant response gain
|
|
gamma = Decimal('.03') #action friction param
|
|
delta = Decimal('.3') #bounds on price change
|
|
omega = Decimal('.5') #bound on burn frac per period
|
|
|
|
seed = {
|
|
'z': np.random.RandomState(1),
|
|
'a': np.random.RandomState(2),
|
|
'b': np.random.RandomState(3),
|
|
'c': np.random.RandomState(3)
|
|
}
|
|
|
|
# Behaviors per Mechanism
|
|
|
|
#arbit X Bond
|
|
def b1m1(step, sL, s):
|
|
#returns "delta p"
|
|
if s['Price']< s['Pool']/s['Supply']-gamma:
|
|
#print('arbit bond')
|
|
#print((s['Pool']/s['Supply']-s['Price'])/s['Price']*s['Pool']*beta)
|
|
return (s['Pool']/s['Supply']-s['Price'])/s['Price']*s['Pool']*beta
|
|
else :
|
|
return 0
|
|
|
|
|
|
#invest X Bond
|
|
def b2m1(step, sL, s):
|
|
#returns "delta p"
|
|
if s['Price']< s['Belief']:
|
|
#print('invest bond')
|
|
#print((s['Belief']-s['Price'])/s['Price']*s['Pool']*beta)
|
|
return s['Pool']*(s['Belief']-s['Price'])/s['Price']*beta
|
|
else :
|
|
return 0
|
|
|
|
#arbit X Burn
|
|
def b1m2(step, sL, s):
|
|
#returns "delta s"
|
|
if Decimal('1')/s['Price']< s['Supply']/s['Pool']-gamma:
|
|
#print('arbit burn')
|
|
#print((s['Supply']/s['Pool']-Decimal('1')/s['Price'])*s['Price']*s['Supply']*beta)
|
|
return s['Price']*(s['Supply']/s['Pool']-Decimal('1')/s['Price'])*s['Supply']*beta
|
|
else :
|
|
return 0
|
|
|
|
#invest X Burn
|
|
def b2m2(step, sL, s):
|
|
#returns "delta s"
|
|
if Decimal('1')/s['Belief']< Decimal('1')/s['Price']:
|
|
#print('invest burn')
|
|
#print(np.min([ s['Pool']*(Decimal('1')/s['Price']-Decimal('1')/s['Belief'])*beta, omega*s['Supply']]))
|
|
return np.min([ s['Pool']*(Decimal('1')/s['Price']-Decimal('1')/s['Belief'])*beta, omega*s['Supply']])
|
|
else :
|
|
return 0
|
|
|
|
#
|
|
#def b1m3(step, sL, s):
|
|
# return s['s1']
|
|
#def b2m3(step, sL, s):
|
|
# return s['s2']
|
|
|
|
|
|
# Internal States per Mechanism
|
|
|
|
#Pool X Bond
|
|
def s1m1(step, sL, s, _input):
|
|
#_input = "delta p"
|
|
return ('Pool',s['Pool']+_input)
|
|
#print("Pool="+str(s['Pool']))
|
|
|
|
#Supply X Bond
|
|
def s2m1(step, sL, s, _input):
|
|
#_input = "delta p"
|
|
return ('Supply', s['Supply']+ _input*s['Supply']/s['Pool'] )
|
|
#print("Supply="+str(s['Supply']))
|
|
|
|
# Pool X Burn
|
|
def s1m2(step, sL, s, _input):
|
|
#_input is "delta s"
|
|
return ('Pool',s['Pool']- _input*s['Pool']/s['Supply'])
|
|
#print("Pool="+str(s['Pool']))
|
|
|
|
# Supply X Burn
|
|
def s2m2(step, sL, s, _input):
|
|
return ('Supply',s['Supply'] - _input)
|
|
#print("Supply="+str(s['Supply']))
|
|
|
|
#def s1m3(step, sL, s, _input):
|
|
# s['s1'] = s['s1']+Decimal(.25)*(s['s2']-s['s1']) + Decimal(.25)*(_input-s['s1'])
|
|
#
|
|
#def s2m3(step, sL, s, _input):
|
|
# s['s2'] = s['s2']+Decimal(.25)*(s['s1']-s['s2']) + Decimal(.25)*(_input-s['s2'])
|
|
|
|
# Exogenous States
|
|
proc_one_coef_A = -delta
|
|
proc_one_coef_B = delta
|
|
def es3p1(step, sL, s, _input):
|
|
rv = bound_norm_random(seed['a'], proc_one_coef_A, proc_one_coef_B)
|
|
return ('Price', theta*s['Price'] * (Decimal('1')+rv) +(Decimal('1')-theta)*s['Pool']/s['Supply'] )
|
|
def es4p2(step, sL, s, _input):
|
|
return ('Belief', alpha*s['Belief']+s['Pool']/s['Supply']*(Decimal('1')-alpha))
|
|
|
|
def es5p2(step, sL, s, _input): # accept timedelta instead of timedelta params
|
|
return ('timestamp', ep_time_step(s, s['timestamp'], seconds=1))
|
|
|
|
# Environment States
|
|
#from numpy.random import randn as rn
|
|
def env_a(x):
|
|
return 3
|
|
def env_b(x):
|
|
return 7
|
|
# def what_ever(x):
|
|
# return x + 1
|
|
|
|
# Genesis States
|
|
state_dict = {
|
|
'Pool': Decimal(10.0),
|
|
'Supply': Decimal(5.0),
|
|
'Price': Decimal(.01),
|
|
'Belief': Decimal(10.0),
|
|
'timestamp': '2018-10-01 15:16:24'
|
|
}
|
|
|
|
exogenous_states = {
|
|
"Price": es3p1,
|
|
"Belief": es4p2,
|
|
"timestamp": es5p2
|
|
}
|
|
|
|
env_processes = {
|
|
"Price": env_proc('2018-10-01 15:16:25', env_a),
|
|
"Belief": env_proc('2018-10-01 15:16:25', env_b)
|
|
}
|
|
|
|
# test return vs. non-return functions as lambdas
|
|
# test fully defined functions
|
|
mechanisms = {
|
|
"bond": {
|
|
"behaviors": {
|
|
"arbit": b1m1, # lambda step, sL, s: s['s1'] + 1,
|
|
"invest": b2m1
|
|
},
|
|
"states": {
|
|
"Pool": s1m1,
|
|
"Supply": s2m1,
|
|
}
|
|
},
|
|
"burn": {
|
|
"behaviors": {
|
|
"arbit": b1m2,
|
|
"invest": b2m2
|
|
},
|
|
"states": {
|
|
"Pool": s1m2,
|
|
"Supply": s2m2,
|
|
}
|
|
},
|
|
# "m3": {
|
|
# "behaviors": {
|
|
# "b1": b1m3,
|
|
# "b2": b2m3
|
|
# },
|
|
# "states": {
|
|
# "s1": s1m3,
|
|
# "s2": s2m3,
|
|
# }
|
|
# }
|
|
}
|
|
|
|
sim_config = {
|
|
"N": 1,
|
|
"R": 1000
|
|
} |