BUG: run value for Genesis State always last run for large datasets
This commit is contained in:
Joshua E. Jodesty
parent
4cc180b9d4
commit
21f1155ae7
|
|
@ -9,4 +9,5 @@ SimCAD.egg-info
|
|||
__pycache__
|
||||
Pipfile
|
||||
Pipfile.lock
|
||||
scrapbox/
|
||||
scrapbox/
|
||||
results/
|
||||
|
|
@ -63,8 +63,9 @@ class Executor:
|
|||
# Dimensions: N x r x mechs
|
||||
|
||||
def single_proc_exec(simulation_execs, states_lists, configs_structs, env_processes_list, Ts, Ns):
|
||||
simulation, states_list, config = simulation_execs.pop(), states_lists.pop(), configs_structs.pop()
|
||||
env_processes, T, N = env_processes_list.pop(), Ts.pop(), Ns.pop()
|
||||
l = [simulation_execs, states_lists, configs_structs, env_processes_list, Ts, Ns]
|
||||
simulation, states_list, config, env_processes, T, N = list(map(lambda x: x.pop(), l))
|
||||
# print(states_list)
|
||||
result = simulation(states_list, config, env_processes, T, N)
|
||||
return flatten(result)
|
||||
|
||||
|
|
|
|||
|
|
@ -1,6 +1,7 @@
|
|||
from copy import deepcopy
|
||||
from fn.op import foldr, call
|
||||
|
||||
import pprint
|
||||
pp = pprint.PrettyPrinter(indent=4)
|
||||
|
||||
class Executor:
|
||||
def __init__(self, behavior_ops):
|
||||
|
|
@ -61,6 +62,7 @@ class Executor:
|
|||
def block_gen(self, states_list, configs, env_processes, t_step, run):
|
||||
m_step = 0
|
||||
states_list_copy = deepcopy(states_list)
|
||||
# print(states_list_copy)
|
||||
# remove copy
|
||||
genesis_states = states_list_copy[-1]
|
||||
genesis_states['mech_step'], genesis_states['time_step'] = m_step, t_step
|
||||
|
|
@ -82,6 +84,7 @@ class Executor:
|
|||
time_seq = [x + 1 for x in time_seq]
|
||||
simulation_list = [states_list]
|
||||
for time_step in time_seq:
|
||||
# print(run)
|
||||
pipe_run = self.block_gen(simulation_list[-1], configs, env_processes, time_step, run)
|
||||
_, *pipe_run = pipe_run
|
||||
simulation_list.append(pipe_run)
|
||||
|
|
@ -94,9 +97,13 @@ class Executor:
|
|||
pipe_run = []
|
||||
for run in range(runs):
|
||||
run += 1
|
||||
head, *tail = self.pipe(states_list, configs, env_processes, time_seq, run)
|
||||
head[-1]['mech_step'], head[-1]['time_step'], head[-1]['run'] = 0, 0, run
|
||||
simulation_list = [head] + tail
|
||||
pipe_run += simulation_list
|
||||
# print("Run: "+str(run))
|
||||
states_list_copy = deepcopy(states_list) # WHY ???
|
||||
head, *tail = self.pipe(states_list_copy, configs, env_processes, time_seq, run)
|
||||
genesis = head.pop()
|
||||
genesis['mech_step'], genesis['time_step'], genesis['run'] = 0, 0, run
|
||||
first_timestep = [genesis] + tail.pop(0)
|
||||
pipe_run += [first_timestep] + tail
|
||||
del states_list_copy
|
||||
|
||||
return pipe_run
|
||||
|
|
@ -79,11 +79,6 @@ def sum_dict_values():
|
|||
# config7c
|
||||
@curried
|
||||
def dict_op(f, d1, d2):
|
||||
print('d1')
|
||||
print(d1)
|
||||
print('d2')
|
||||
print(d2)
|
||||
print()
|
||||
def set_base_value(target_dict, source_dict, key):
|
||||
if key not in target_dict:
|
||||
return get_base_value(type(source_dict[key]))
|
||||
|
|
|
|||
|
|
@ -6,7 +6,7 @@ from SimCAD.utils.configuration import exo_update_per_ts, proc_trigger, bound_no
|
|||
ep_time_step
|
||||
|
||||
seed = {
|
||||
'z': np.random.RandomState(1)
|
||||
'z': np.random.RandomState(1)
|
||||
}
|
||||
|
||||
# Signals
|
||||
|
|
@ -19,7 +19,7 @@ external_draw = Decimal('0.01') # between 0 and 1 to draw Buy_Log to external
|
|||
|
||||
# Stochastic process factors
|
||||
correction_factor = Decimal('0.01')
|
||||
volatility = Decimal('5.0')
|
||||
volatility = Decimal('5.0')
|
||||
|
||||
# Buy_Log_signal =
|
||||
# Z_signal =
|
||||
|
|
|
|||
|
|
@ -0,0 +1,243 @@
|
|||
from decimal import Decimal
|
||||
import numpy as np
|
||||
|
||||
from SimCAD import Configuration, configs
|
||||
from SimCAD.utils.configuration import exo_update_per_ts, proc_trigger, bound_norm_random, \
|
||||
ep_time_step
|
||||
|
||||
seed = {
|
||||
'z': np.random.RandomState(1)
|
||||
}
|
||||
|
||||
# Signals
|
||||
# Pr_signal
|
||||
beta = Decimal('0.25') # agent response gain
|
||||
beta_LT = Decimal('0.1') # LT agent response gain
|
||||
# alpha = .67, 2 block moving average
|
||||
alpha = Decimal('0.67') # 21 day EMA forgetfullness between 0 and 1, closer to 1 discounts older obs quicker, should be 2/(N+1)
|
||||
max_withdraw_factor = Decimal('0.9')
|
||||
external_draw = Decimal('0.01') # between 0 and 1 to draw Buy_Log to external
|
||||
|
||||
|
||||
#alpha * s['Zeus_ST'] + (1 - alpha)*s['Zeus_LT']
|
||||
|
||||
# Stochastic process factors
|
||||
correction_factor = Decimal('0.01')
|
||||
volatility = Decimal('5.0')
|
||||
|
||||
# Buy_Log_signal =
|
||||
# Z_signal =
|
||||
# Price_signal =
|
||||
# TDR_draw_signal =
|
||||
# P_Ext_Markets_signal =
|
||||
|
||||
# Behaviors per Mechanism
|
||||
|
||||
# BEHAVIOR 1: EMH Trader
|
||||
EMH_portion = Decimal('0.250')
|
||||
EMH_Ext_Hold = Decimal('42000.0')
|
||||
|
||||
|
||||
def b1m1(step, sL, s):
|
||||
# print('b1m1')
|
||||
theta = (s['Z']*EMH_portion*s['Price'])/(s['Z']*EMH_portion*s['Price'] + EMH_Ext_Hold * s['P_Ext_Markets'])
|
||||
if s['Price'] < (theta*EMH_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*EMH_portion*(1-theta)):
|
||||
buy = beta * theta*EMH_Ext_Hold * s['P_Ext_Markets']/(s['Price']*EMH_portion*(1-theta))
|
||||
return {'buy_order1': buy}
|
||||
elif s['Price'] > (theta*EMH_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*EMH_portion*(1-theta)):
|
||||
return {'buy_order1': 0}
|
||||
else:
|
||||
return {'buy_order1': 0}
|
||||
|
||||
|
||||
def b1m2(step, sL, s):
|
||||
# print('b1m2')
|
||||
theta = (s['Z']*EMH_portion*s['Price'])/(s['Z']*EMH_portion*s['Price'] + EMH_Ext_Hold * s['P_Ext_Markets'])
|
||||
if s['Price'] < (theta*EMH_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*EMH_portion*(1-theta)):
|
||||
return {'sell_order1': 0}
|
||||
elif s['Price'] > (theta*EMH_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*EMH_portion*(1-theta)):
|
||||
sell = beta * theta*EMH_Ext_Hold * s['P_Ext_Markets']/(s['Price']*EMH_portion*(1-theta))
|
||||
return {'sell_order1': sell}
|
||||
else:
|
||||
return {'sell_order1': 0}
|
||||
|
||||
# BEHAVIOR 3: Herding
|
||||
Herd_portion = Decimal('0.250')
|
||||
Herd_Ext_Hold = Decimal('42000.0')
|
||||
Herd_UB = Decimal('0.10') # UPPER BOUND
|
||||
Herd_LB = Decimal('0.10') # LOWER BOUND
|
||||
def b3m2(step, sL, s):
|
||||
theta = (s['Z']*Herd_portion*s['Price'])/(s['Z']*Herd_portion*s['Price'] + Herd_Ext_Hold * s['P_Ext_Markets'])
|
||||
# if s['Price'] - s['Price_Signal'] < (theta*Herd_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*Herd_portion*(1-theta)) - Herd_LB:
|
||||
if (s['Price'] - s['Price_Signal']) < - Herd_LB:
|
||||
|
||||
sell = beta * theta*Herd_Ext_Hold * s['P_Ext_Markets']/(s['Price']*Herd_portion*(1-theta))
|
||||
return {'herd_sell': sell, 'herd_buy': 0}
|
||||
# elif s['Price'] > Herd_UB - (theta*Herd_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*Herd_portion*(1-theta)):
|
||||
elif (s['Price'] - s['Price_Signal']) > Herd_UB:
|
||||
buy = beta * theta*Herd_Ext_Hold * s['P_Ext_Markets']/(s['Price']*Herd_portion*(1-theta))
|
||||
return {'herd_sell': 0, 'herd_buy': buy}
|
||||
else:
|
||||
return {'herd_sell': 0, 'herd_buy': 0}
|
||||
|
||||
# BEHAVIOR 4: HODLers
|
||||
HODL_belief = Decimal('10.0')
|
||||
HODL_portion = Decimal('0.250')
|
||||
HODL_Ext_Hold = Decimal('4200.0')
|
||||
|
||||
|
||||
def b4m2(step, sL, s):
|
||||
# print('b4m2')
|
||||
theta = (s['Z']*HODL_portion*s['Price'])/(s['Z']*HODL_portion*s['Price'] + HODL_Ext_Hold * s['P_Ext_Markets'])
|
||||
if s['Price'] < 1/HODL_belief*(theta*HODL_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*HODL_portion*(1-theta)):
|
||||
sell = beta * theta*HODL_Ext_Hold * s['P_Ext_Markets']/(s['Price']*HODL_portion*(1-theta))
|
||||
return {'sell_order2': sell}
|
||||
elif s['Price'] > (theta*HODL_Ext_Hold * s['P_Ext_Markets'])/(s['Z']*HODL_portion*(1-theta)):
|
||||
return {'sell_order2': 0}
|
||||
else:
|
||||
return {'sell_order2': 0}
|
||||
|
||||
|
||||
# STATES
|
||||
# ZEUS Fixed Supply
|
||||
def s1m1(step, sL, s, _input):
|
||||
y = 'Z'
|
||||
x = s['Z'] #+ _input # / Psignal_int
|
||||
return (y, x)
|
||||
|
||||
|
||||
# def s2m1(step, sL, s, _input):
|
||||
# y = 'Price'
|
||||
# x = (s['P_Ext_Markets'] - _input['buy_order1']) / s['Z'] * 10000
|
||||
# #x= alpha * s['Z'] + (1 - alpha)*s['Price']
|
||||
# return (y, x)
|
||||
|
||||
|
||||
def s3m1(step, sL, s, _input):
|
||||
y = 'Buy_Log'
|
||||
x = _input['buy_order1'] + _input['herd_buy'] # / Psignal_int
|
||||
return (y, x)
|
||||
|
||||
|
||||
def s4m2(step, sL, s, _input):
|
||||
y = 'Sell_Log'
|
||||
x = _input['sell_order1'] + _input['sell_order2'] + _input['herd_sell'] # / Psignal_int
|
||||
return (y, x)
|
||||
|
||||
|
||||
def s3m3(step, sL, s, _input):
|
||||
y = 'Buy_Log'
|
||||
x = s['Buy_Log'] + _input # / Psignal_int
|
||||
return (y, x)
|
||||
|
||||
|
||||
# Price Update
|
||||
def s2m3(step, sL, s, _input):
|
||||
|
||||
y = 'Price'
|
||||
#var1 = Decimal.from_float(s['Buy_Log'])
|
||||
x = s['Price'] + s['Buy_Log'] /s['Z'] - s['Sell_Log']/s['Z']
|
||||
#+ np.divide(s['Buy_Log'],s['Z']) - np.divide() # / Psignal_int
|
||||
return (y, x)
|
||||
|
||||
def s5m3(step, sL, s, _input):
|
||||
y = 'Price_Signal'
|
||||
x = alpha * s['Price'] + (1 - alpha)*s['Price_Signal']
|
||||
return (y, x)
|
||||
|
||||
def s6m1(step, sL, s, _input):
|
||||
y = 'P_Ext_Markets'
|
||||
x = s['P_Ext_Markets'] - _input
|
||||
#x= alpha * s['Z'] + (1 - alpha)*s['Price']
|
||||
return (y, x)
|
||||
|
||||
|
||||
def s2m2(step, sL, s, _input):
|
||||
y = 'Price'
|
||||
x = (s['P_Ext_Markets'] - _input) /s['Z'] *10000
|
||||
#x= alpha * s['Z'] + (1 - alpha)*s['Price']
|
||||
return (y, x)
|
||||
|
||||
# Exogenous States
|
||||
proc_one_coef_A = -125
|
||||
proc_one_coef_B = 125
|
||||
|
||||
# A change in belief of actual price, passed onto behaviors to make action
|
||||
def es4p2(step, sL, s, _input):
|
||||
y = 'P_Ext_Markets'
|
||||
x = s['P_Ext_Markets'] + bound_norm_random(seed['z'], proc_one_coef_A, proc_one_coef_B)
|
||||
|
||||
return (y,x)
|
||||
|
||||
|
||||
def es5p2(step, sL, s, _input): # accept timedelta instead of timedelta params
|
||||
y = 'timestamp'
|
||||
x = ep_time_step(s, s['timestamp'], seconds=1)
|
||||
return (y, x)
|
||||
|
||||
#Environment States
|
||||
# NONE
|
||||
|
||||
# Genesis States
|
||||
state_dict = {
|
||||
'Z': Decimal(21000000.0),
|
||||
'Price': Decimal(100.0), # Initialize = Z for EMA
|
||||
'Buy_Log': Decimal(0.0),
|
||||
'Sell_Log': Decimal(0.0),
|
||||
'Price_Signal': Decimal(100.0),
|
||||
'Trans': Decimal(0.0),
|
||||
'P_Ext_Markets': Decimal(25000.0),
|
||||
'timestamp': '2018-10-01 15:16:24'
|
||||
}
|
||||
|
||||
def env_proc_id(x):
|
||||
return x
|
||||
|
||||
env_processes = {}
|
||||
|
||||
exogenous_states = exo_update_per_ts(
|
||||
{
|
||||
"P_Ext_Markets": es4p2,
|
||||
"timestamp": es5p2
|
||||
}
|
||||
)
|
||||
|
||||
sim_config = {
|
||||
"N": 20,
|
||||
"T": range(1000)
|
||||
}
|
||||
|
||||
# test return vs. non-return functions as lambdas
|
||||
# test fully defined functions
|
||||
mechanisms = {
|
||||
"m1": {
|
||||
"behaviors": {
|
||||
"b1": b1m1,
|
||||
"b3": b3m2
|
||||
},
|
||||
"states": {
|
||||
"Z": s1m1,
|
||||
"Buy_Log": s3m1
|
||||
}
|
||||
},
|
||||
"m2": {
|
||||
"behaviors": {
|
||||
"b1": b1m2,
|
||||
"b3": b3m2,
|
||||
"b4": b4m2
|
||||
},
|
||||
"states": {
|
||||
"Sell_Log": s4m2
|
||||
}
|
||||
},
|
||||
"m3": {
|
||||
"behaviors": {
|
||||
},
|
||||
"states": {
|
||||
"Price": s2m3,
|
||||
"Price_Signal": s5m3
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
configs.append(Configuration(sim_config, state_dict, seed, exogenous_states, env_processes, mechanisms))
|
||||
|
|
@ -2,8 +2,8 @@ import pandas as pd
|
|||
from tabulate import tabulate
|
||||
|
||||
from SimCAD.engine import ExecutionMode, ExecutionContext, Executor
|
||||
# from sandboxUX import config1, config2
|
||||
from sandboxUX import config3
|
||||
from sandboxUX import config1, config2
|
||||
# from sandboxUX import config4
|
||||
from SimCAD import configs
|
||||
|
||||
# ToDo: pass ExecutionContext with execution method as ExecutionContext input
|
||||
|
|
@ -18,15 +18,16 @@ single_proc_ctx = ExecutionContext(exec_mode.single_proc)
|
|||
run1 = Executor(single_proc_ctx, single_config)
|
||||
run1_raw_result = run1.main()
|
||||
result = pd.DataFrame(run1_raw_result)
|
||||
# result.to_csv('~/Projects/DiffyQ-SimCAD/results/config4.csv', sep=',')
|
||||
print(tabulate(result, headers='keys', tablefmt='psql'))
|
||||
print()
|
||||
#
|
||||
# print("Simulation Run 2: Pairwise Execution")
|
||||
# print()
|
||||
# multi_proc_ctx = ExecutionContext(exec_mode.multi_proc)
|
||||
# run2 = Executor(multi_proc_ctx, configs)
|
||||
# run2_raw_results = run2.main()
|
||||
# for raw_result in run2_raw_results:
|
||||
# result = pd.DataFrame(raw_result)
|
||||
# print(tabulate(result, headers='keys', tablefmt='psql'))
|
||||
# print()
|
||||
|
||||
print("Simulation Run 2: Pairwise Execution")
|
||||
print()
|
||||
multi_proc_ctx = ExecutionContext(exec_mode.multi_proc)
|
||||
run2 = Executor(multi_proc_ctx, configs)
|
||||
run2_raw_results = run2.main()
|
||||
for raw_result in run2_raw_results:
|
||||
result = pd.DataFrame(raw_result)
|
||||
print(tabulate(result, headers='keys', tablefmt='psql'))
|
||||
print()
|
||||
Loading…
Reference in New Issue