def plot_swaption_volas2(dfInput, trade_dates, labels, optionPeriodSymbol,
swapPeriodSymbol):
dfMerged = dfInput['ST']
for trade_date in trade_dates:
dfMerged = dfMerged.merge(pd.DataFrame(
data=dfInput['data'].loc[dfInput['data'].index ==
trade_date].values[0],
columns=[trade_date],
index=dfInput['data'].loc[dfInput['data'].index ==
trade_date].columns),
left_index=True,
right_index=True)
fig, axs = plt.subplots(figsize=(10, 5))
for trade_date, label in zip(trade_dates, labels):
dfMerged.sort_values(by='strike').loc[
dfMerged.optionPeriodSymbol == optionPeriodSymbol][
dfMerged.swapPeriodSymbol == swapPeriodSymbol].plot(
ax=axs,
x='strike',
y=trade_date,
style='*-',
label=label,
sharex=True,
sharey=True)
save_fig(
str(ENCODER_BASE_NAME) + '_swaption_volas2_' +
str(optionPeriodSymbol) + '_' + str(swapPeriodSymbol))
def plot_loss(n_bottleneck, noise_stddev):
encoder_name = get_encoder_name(n_bottleneck, noise_stddev)
df_loss_train_test_epoch = pd.read_csv(os.path.join(
os.getcwd(), PATH_OUTPUT, encoder_name + ".csv"),
sep=';')
df_loss_train_test_epoch.plot(x='iEpoch',
y=['loss_train', 'loss_test'],
title='reconstruction loss')
save_fig(str(ENCODER_BASE_NAME) + '_loss_' + str(encoder_name))
def plot_swaption_volas3(dfInput, dfResults, n_bottlenecks, noise_stddevs,
trade_date, optionPeriodSymbol, swapPeriodSymbol):
dfMerged = dfInput['ST']
dfMerged = dfMerged.merge(pd.DataFrame(
data=dfInput['data'].loc[dfInput['data'].index ==
trade_date].values[0],
columns=['input'],
index=dfInput['data'].loc[dfInput['data'].index ==
trade_date].columns),
left_index=True,
right_index=True)
title = 'option maturity: ' + str(
optionPeriodSymbol) + ', swap maturity: ' + str(swapPeriodSymbol)
for n_bottleneck, noise_stddev in zip(n_bottlenecks, noise_stddevs):
encoder_name = get_encoder_name(n_bottleneck, noise_stddev)
dfMerged = dfMerged.merge(pd.DataFrame(
data=dfResults[encoder_name]['data'].loc[
dfResults[encoder_name]['data'].index == trade_date].values[0],
columns=[
'bl:' + str(n_bottleneck) + ', stdev:' + str(noise_stddev)
],
index=dfResults[encoder_name]['data'].loc[
dfResults[encoder_name]['data'].index == trade_date].columns),
left_index=True,
right_index=True)
fig, axs = plt.subplots(figsize=(10, 5))
dfMerged.sort_values(
by='strike').loc[dfMerged.optionPeriodSymbol == optionPeriodSymbol][
dfMerged.swapPeriodSymbol == swapPeriodSymbol].plot(ax=axs,
x='strike',
y='input',
style='*-',
sharex=True,
sharey=True,
title=title)
for n_bottleneck, noise_stddev in zip(n_bottlenecks, noise_stddevs):
dfMerged.sort_values(by='strike').loc[
dfMerged.optionPeriodSymbol == optionPeriodSymbol][
dfMerged.swapPeriodSymbol == swapPeriodSymbol].plot(
ax=axs,
x='strike',
y='bl:' + str(n_bottleneck) + ', stdev:' +
str(noise_stddev),
style='*-',
sharex=True,
sharey=True)
save_fig(
str(ENCODER_BASE_NAME) + '_swaption_volas3_' +
str(optionPeriodSymbol) + '_' + str(swapPeriodSymbol) + '_' +
str(trade_date))
def plot_swaption_volas4(dfInput, dfResults, n_bottleneck, noise_stddev,
trade_date):
encoder_name = get_encoder_name(n_bottleneck, noise_stddev)
dfInputMerged = pd.DataFrame(
data=dfInput['data'].loc[dfInput['data'].index ==
trade_date].values[0],
columns=['valueInput'],
index=dfInput['data'].loc[dfInput['data'].index == trade_date].columns)
dfInputMerged = dfInputMerged.merge(dfInput['ST'],
left_index=True,
right_index=True)
dfInputMerged = dfInputMerged.merge(pd.DataFrame(
data=dfResults[encoder_name]['data'].loc[
dfResults[encoder_name]['data'].index == trade_date].values[0],
columns=['valuePrediction'],
index=dfResults[encoder_name]['data'].loc[
dfResults[encoder_name]['data'].index == trade_date].columns),
left_index=True,
right_index=True)
nRows = len(dfInputMerged.optionPeriodSymbol.unique())
nColumns = len(dfInputMerged.swapPeriodSymbol.unique())
fig, axs = plt.subplots(nRows, nColumns, figsize=(30, 40))
for iRow, iOptionPeriodSymbol in enumerate(
dfInputMerged.optionPeriodSymbol.unique()):
for iColumn, iSwapPeriodSymbol in enumerate(
dfInputMerged.swapPeriodSymbol.unique()):
label = iOptionPeriodSymbol + '_' + iSwapPeriodSymbol
dfInputMerged.loc[
dfInputMerged.optionPeriodSymbol == iOptionPeriodSymbol][
dfInputMerged.swapPeriodSymbol == iSwapPeriodSymbol].plot(
ax=axs[iRow, iColumn],
x='strike',
y='valueInput',
style='*-',
label='In_' + label,
sharex=True,
sharey=True)
dfInputMerged.loc[
dfInputMerged.optionPeriodSymbol == iOptionPeriodSymbol][
dfInputMerged.swapPeriodSymbol == iSwapPeriodSymbol].plot(
ax=axs[iRow, iColumn],
x='strike',
y='valuePrediction',
style='--',
label='Pred_' + label,
sharex=True,
sharey=True)
save_fig(
str(ENCODER_BASE_NAME) + '_swaption_volas4_' + str(encoder_name) +
'_' + str(trade_date))
def plot_hist(dfResults, index_train, index_test, n_bottleneck, noise_stddev):
encoder_name = get_encoder_name(n_bottleneck, noise_stddev)
fig, axs = plt.subplots(figsize=(10, 5))
dfResults[encoder_name]['zValue'].loc[index_train].plot.hist(
y='zValue',
label='train',
ax=axs,
log=True,
title='bl: ' + str(n_bottleneck) + ', stdev: ' + str(noise_stddev))
dfResults[encoder_name]['zValue'].loc[index_test].plot.hist(y='zValue',
label='test',
ax=axs,
log=True)
dfResults[encoder_name]['zValue'].loc[
dfResults[encoder_name]['zValue'].index < '1900'].plot.hist(
y='zValue', label='injected points', ax=axs, log=True)
axs.set(xlabel="z-value")
save_fig(str(ENCODER_BASE_NAME) + '_hist_' + str(encoder_name))
def plot_swaption_volas(dfInput, trade_date):
df = pd.DataFrame(
data=dfInput['data'].loc[dfInput['data'].index ==
trade_date].values[0],
columns=['value'],
index=dfInput['data'].loc[dfInput['data'].index == trade_date].columns)
df = df.merge(dfInput['ST'], left_index=True, right_index=True)
fig, axs = plt.subplots(2, 3, figsize=(15, 10))
for idxPlot, iOptionPeriodSymbol in enumerate(
df.optionPeriodSymbol.unique()):
for iSwapPeriodSymbol in df.swapPeriodSymbol.unique():
iRow = int(idxPlot / 3)
iColumn = idxPlot % 3
label = iOptionPeriodSymbol + '_' + iSwapPeriodSymbol
df.loc[df.optionPeriodSymbol == iOptionPeriodSymbol][
df.swapPeriodSymbol == iSwapPeriodSymbol].plot(ax=axs[iRow,
iColumn],
x='strike',
y='value',
style='*-',
label=label,
sharex=True,
sharey=True)
save_fig(str(ENCODER_BASE_NAME) + '_swaption_volas_' + str(trade_date))
def plot_feature_vector(dfInput, trade_date):
from helper import save_fig
pd.DataFrame(dfInput['data'].loc[dfInput['data'].index == '20181228'].
reset_index().T.iloc[1:].values).plot(legend=False)
save_fig(str(ENCODER_BASE_NAME) + '_features_' + str(trade_date))
x = np.random.randn(2, N)
distrib0 = np.matmul(Sigma0, (x - np.expand_dims(mu[:,0], axis=1)))
distrib1 = np.matmul(Sigma1, (x - np.expand_dims(mu[:,1], axis=1)))
fig = plt.figure()
ax = fig.add_subplot(111)
ax.scatter(distrib0[0], distrib0[1], color = 'b', s = 10)
ax.scatter(distrib1[0], distrib1[1], color = 'y', s = 10)
ax.scatter(0,0,color = 'r', s = 100)
ax.set_xticks([])
ax.set_yticks([])
plt.tight_layout()
save_fig(fig, 'silly_example')
design = np.hstack((distrib1, distrib0)).transpose()
labels = np.hstack((np.ones(N), -np.ones(N)))
cl = MetricLearningClassifier(gamma=10**(-3), alpha=10**(-7), max_it=100)
cl.fit(design, labels, verbose = True)
S = cl.S_tot
cur_S = S[0]
x0 = np.matmul(cur_S, distrib0)
x1 = np.matmul(cur_S, distrib1)
fig = plt.figure()
ax = fig.add_subplot(111)