def main():
data_location = './data/lorenz_96.npy'
data = generate(N=40, F=8.0)
utils.save_as_numpy(data, data_location)
data = utils.load_to_numpy(data_location)
utils.plot(data)
def gen_spatial_heat_maps(q_spatial, decomposed_channel_num, spatial_factors, save_directory,
attr_class, factorization_method, no_slash_layer_name, img, AM, model):
if q_spatial is not None:
for i_decomposed_channel_num in range(decomposed_channel_num):
spatial_factors[i_decomposed_channel_num, ...] = spatial_factors[i_decomposed_channel_num, ...] * (
spatial_factors[i_decomposed_channel_num, ...] > np.quantile(spatial_factors[i_decomposed_channel_num, ...],
q_spatial / 100))
index_saveimg = 0
for i_factor in range(spatial_factors.shape[0]):
factor_resized = resize(spatial_factors[i_factor], (model.image_shape[0], model.image_shape[1]), order=1,
mode='constant', anti_aliasing=False)
imgtype_name1 = 'SpatialHM'
plot(factor_resized, save_directory, attr_class, factorization_method, no_slash_layer_name,
imgtype_name1, index_saveimg, xi=img, cmap2='seismic', alpha=0.3)
index_saveimg = index_saveimg + 1
print('heat maps have been saved')
def main():
# numeric greeks example
# default market environment
market_env = MarketEnvironment()
print(market_env)
# define option style and type
opt_style = "plain_vanilla" # "digital"
opt_type = "call" # "call"
option = option_factory(market_env, opt_style, opt_type)
print(option)
# numeric greeks instance
NumGreeks = NumericGreeks(option)
# underlying range at which compute greeks
S_range = np.linspace(50, 150, 2000)
# time-to-maturity range at which compute greeks
tau_range = np.linspace(1e-4, 1.0, 1000)
tau_range = homogenize(tau_range, reverse_order=True)
# select greek
for greek_type in ["delta", "theta", "gamma", "vega", "rho"]:
#
# greek Vs Underlying level S
#
# numeric greek calculation
greek_numeric_Vs_S = greeks_factory(NumGreeks, greek_type)(S=S_range)
# labels
label_numeric_S = greeks_label_factory(greek_type,
opt_type,
kind="num")
# plot title
plot_title_S = greeks_title_factory(option, greek_type)
# plot
plot(x=S_range,
f=greek_numeric_Vs_S,
x_label=r"$S$",
f_label=label_numeric_S,
title=plot_title_S)
#
# greek Vs residual time to maturity tau
#
# numeric greek calculation
greek_numeric_Vs_tau = greeks_factory(NumGreeks,
greek_type)(tau=tau_range)
# labels
label_numeric_tau = greeks_label_factory(greek_type,
opt_type,
kind="num",
underlying=r"\tau")
# plot title
plot_title_tau = greeks_title_factory(option,
greek_type,
underlying=r"\tau")
# plot
plot(x=tau_range,
f=greek_numeric_Vs_tau,
x_label=r"$\tau$",
f_label=label_numeric_tau,
title=plot_title_tau)
marker_size=9,
symbol='square',
))
trace_data.append(
utils.scatter_trace(
x,
n_15,
'A',
symbol='triangle-up',
marker_size=9,
))
utils.plot(
trace_data,
'',
'Kshitij',
x_title=x_title,
y_title='Test Error',
x_scale='log',
x_tickvals=x_tickvals,
x_ticktext=x_ticktext,
# y_tickvals=[0, 0.01, 0.02, 0.03, 0.4],
# y_ticktext=y_ticktexts,
# x_range=x_range,
# y_range=[0, 0.04],
# legend_x=legend_x,
# legend_xanchor=legend_xanchor,
legend_y=0.8,
# legend_orientation='h',
)
np.mean(episode_length_history[max(0,
len(episode_length_history) -
100):]))
if not episode % 20 and episode:
print(
"Episode {} over. Broken WR: {:.3f}. AVG reward: {:.3f}. Episode legth: {:.2f}."
.format(episode, wr_array[-1], reward_history_avg[-1],
episode_length_avg[-1]))
if not episode % 1000 and episode:
# Save model
player.save_model(MODELS_DIR, episode)
print("Model saved")
# Update plot of the Winning Rate
plot(wr_array, wr_array_avg, "WR history", "WR_history_training",
PLOTS_DIR, ["WR", "100-episode average"])
# Update plot of the reward
plot(reward_history, reward_history_avg, "Reward history",
"reward_history_training", PLOTS_DIR,
["Reward", "100-episode average"])
# Update plot of the episode length
plot(episode_length_history, episode_length_avg, "Episode length",
"episode_length_training", PLOTS_DIR,
["Episode length", "100-episode average"])
episode += 1