derivative = BachelierBinary(maturity, strike, sigma)
numeraire_simulator = simulators.ConstantBankAccount(0.0)
book = books.DerivativeBook(maturity, instrument_simulator,
numeraire_simulator)
book.add_derivative(derivative, 0, 1.0)
init_instruments = tf.constant([spot], FLOAT_DTYPE)
init_numeraire = tf.constant([1.0], FLOAT_DTYPE)
driver = utils.HedgeDriver(
timesteps=timesteps,
frequency=0, # no need for frequency for non-path dependent derivatives.
init_instruments=init_instruments,
init_numeraire=init_numeraire,
book=book,
cost=1 / 100 if cost else None,
risk_neutral=True,
learning_rate=1e-1)
driver.verbose = 2
risklevels = [0.05, 0.5, 0.95] if not cost else [0.95]
for alpha in risklevels:
driver.add_testcase(f"deep network {alpha}",
hedge_models.NeuralHedge(
timesteps=timesteps,
instrument_dim=book.instrument_dim,
internal_dim=0,
num_layers=4,
num_units=5,
risk_measure = partial(hedge_models.MeanVariance, aversion=1.0)
else:
risk_measure = partial(hedge_models.ExpectedShortfall, alpha=alpha)
for num in range(num_trials):
print(f"dimension {dimension} at test {num + 1} ".ljust(80, "="), end="")
start = perf_counter()
init_instruments, init_numeraire, book = random_books.random_empty_book(
timesteps / 250, dimension, rate, drift, volatility, num)
random_books.add_rko(init_instruments, book, spread)
driver = utils.HedgeDriver(timesteps=timesteps,
frequency=frequency,
init_instruments=init_instruments,
init_numeraire=init_numeraire,
book=book,
cost=1 / 100 if cost else None,
risk_neutral=not cost,
learning_rate=1e-1)
driver.add_testcase(name="continuous-time",
model=hedge_models.FeatureHedge(),
risk_measure=risk_measure(),
normaliser=None,
feature_function="delta",
price_type="arbitrage")
driver.add_testcase("deep network",
hedge_models.NeuralHedge(
timesteps=timesteps,
instrument_dim=book.instrument_dim,
case_name = "call"
# ==============================================================================
# === sample data
init_instruments, init_numeraire, book = random_books.random_empty_book(
timesteps / 250, 1, 0.0, 0.05, 0.2, seed=69) # NOTE timesteps / 52 for DGA
# random_books.add_calls(init_instruments, book)
random_books.add_butterfly(init_instruments, book, 10)
# random_books.add_dga_calls(init_instruments, book)
# random_books.add_rko(init_instruments, book, 10.0)
driver = utils.HedgeDriver(timesteps=timesteps * hedge_multiplier,
frequency=frequency,
init_instruments=init_instruments,
init_numeraire=init_numeraire,
book=book,
cost=None,
risk_neutral=True,
learning_rate=1e-1)
driver.add_testcase(case_name,
hedge_models.FeatureHedge(),
risk_measure=hedge_models.ExpectedShortfall(alpha),
normaliser=None,
feature_function="delta",
price_type="arbitrage")
driver.train(sample_size=train_size, epochs=0, batch_size=2**2)
driver.test(test_size)
print(
risk_measure = partial(hedge_models.ExpectedShortfall, alpha=alpha)
for num in range(num_trials):
print(f"dimension {dimension} at test {num + 1} ".ljust(80, "="), end="")
start = perf_counter()
init_instruments, init_numeraire, book = random_books.random_empty_book(
timesteps / 12, dimension, rate, drift, volatility, num)
random_books.add_dga_calls(init_instruments, book)
driver = utils.HedgeDriver(
timesteps=timesteps * hedge_multiplier,
frequency=0, # no need for frequency
init_instruments=init_instruments,
init_numeraire=init_numeraire,
book=book,
cost=1/100 if cost else None,
risk_neutral=not cost,
learning_rate=1e-1
)
driver.add_testcase(
name="continuous-time",
model=hedge_models.FeatureHedge(),
risk_measure=risk_measure(),
normaliser=None,
feature_function="delta",
price_type="arbitrage")
driver.add_testcase(
"deep network",
alpha = 0.95
# ==============================================================================
# === sample data
init_instruments, init_numeraire, book = random_books.random_empty_book(
1.0, 1, 0.05, 0.1, 0.2, seed=69)
derivative = derivatives.PutCall(book.maturity, 105.0,
book.numeraire_simulator.rate,
book.instrument_simulator.volatility, 1)
book.add_derivative(derivative, 0, 1.0)
driver = utils.HedgeDriver(
timesteps=timesteps,
frequency=0, # no need for frequency
init_instruments=init_instruments,
init_numeraire=init_numeraire,
book=book,
cost=None,
risk_neutral=False,
learning_rate=1e-1)
driver.add_testcase("delta",
hedge_models.FeatureHedge(),
risk_measure=hedge_models.ExpectedShortfall(alpha),
normaliser=None,
feature_function="delta",
price_type="arbitrage")
driver.train(train_size, 1, int(2**10))
driver.test(test_size)
