meditech.py

meditech.py

import jax
import jax.numpy as jnp
import matplotlib.pyplot as plt
from collections import namedtuple

Decision = namedtuple('Decision', ['cable', 'assembly', 'pack'])
EquityRate = namedtuple('EquityRate', ['CEO', 'VC', 'Emp', 'Supp'])

INITIAL_MONEY = 760_000

PROTOTYPE_COST = {
    Decision(cable=True, assembly=True, pack=True): 1_990_165,
    Decision(cable=True, assembly=True, pack=False): 2_142_000,
    Decision(cable=True, assembly=False, pack=True): 2_282_156,
    Decision(cable=True, assembly=False, pack=False): 2_433_991,
    Decision(cable=False, assembly=True, pack=True): 2_457_350,
    Decision(cable=False, assembly=True, pack=False): 2_609_185,
    Decision(cable=False, assembly=False, pack=True): 2_749_341,
    Decision(cable=False, assembly=False, pack=False): 2_901_176
}

ITERATION_TIME_RATE = {
    Decision(cable=True, assembly=True, pack=True): 1/(7+2+0),
    Decision(cable=True, assembly=True, pack=False): 1/(7+2+3),
    Decision(cable=True, assembly=False, pack=True): 2_282_156,
    Decision(cable=True, assembly=False, pack=False): 2_433_991,
    Decision(cable=False, assembly=True, pack=True): 2_457_350,
    Decision(cable=False, assembly=True, pack=False): 2_609_185,
    Decision(cable=False, assembly=False, pack=True): 2_749_341,
    Decision(cable=False, assembly=False, pack=False): 2_901_176
}

EQUITY_RATE = {
    Decision(cable=True, assembly=True, pack=True): EquityRate(CEO=0.226, VC=0.539, Emp=0.100, Supp=0.136),
    Decision(cable=True, assembly=True, pack=False): EquityRate(CEO=0.226, VC=0.539, Emp=0.100, Supp=0.136),
    Decision(cable=True, assembly=False, pack=True): EquityRate(CEO=0.226, VC=0.539, Emp=0.100, Supp=0.136),
    Decision(cable=True, assembly=False, pack=False): EquityRate(CEO=0.226, VC=0.539, Emp=0.100, Supp=0.136),
    Decision(cable=False, assembly=True, pack=True): EquityRate(CEO=0.272, VC=0.608, Emp=0.120, Supp=0.000),
    Decision(cable=False, assembly=True, pack=False): EquityRate(CEO=0.272, VC=0.608, Emp=0.120, Supp=0.000),
    Decision(cable=False, assembly=False, pack=True): EquityRate(CEO=0.272, VC=0.608, Emp=0.120, Supp=0.000),
    Decision(cable=False, assembly=False, pack=False): EquityRate(CEO=0.272, VC=0.608, Emp=0.120, Supp=0.000)
}

def iteration_time(key, decision: Decision):
    return ITERATION_TIME_RATE[decision] * jax.random.exponential(key)


def fda_approved(iter_number):
    if iter_number <= 4:
            return 0.1
    if iter_number <= 5:
        return 0.2 
    return 0.35 
        

def model(key, decision: Decision):
    equity = EQUITY_RATE[decision]
    evaluation = 20_000_000 + EVALUATION_CHANGE[decision]
    cash = INITIAL_MONEY

    iter_number, time = 1, 0
    while cash > 0:
        cash = cash - PROTOTYPE_COST[decision]
        time = time + iteration_time(key, decision)
        iter_number = iter_number + 1
        key = jax.random.split(key)[0]

    approval_rate = fda_approved(iter_number)

    # XXX TODO

    return approval_rate, time, equity

#### optimization
 N = 1000

for decision in PROTOTYPE_COST.keys():
    ceo_costs = []
    for _ in range(N):
        approval_rate, time, equity = model(decision)
        ceo_costs.append(ceo_cost_fn(approval_rate, equity.CEO, time))

    print('average cost for ceo', sum(ceo_costs) / N)