def do_variable_plots ():
def do_variable_plot (name, cut_level, outdir):
filename = '{0}/{1}.png'.format (outdir, name)
# note that both the thing to plot (name, in this case) and the cut can
# be specified in a very flexible way; see Validator.eval()
# documentation for more info.
objs = v.create_plot (name, 'dist',
lines,
bins=50,
dual=True,
data_mc=True,
left_ylabel='Hz per bin',
xlabel=name,
log_kwargs=dict (
legend=dict (loc='best')
),
title='{0} | bdt score > {1:.3f}'.format (name, cut_level),
cut='scores > {0}'.format (cut_level)
)
objs['fig'].savefig (filename)
input_vars_dir = 'output/vars_at_input'
print ('Plotting input vars in {0} ...'.format (input_vars_dir))
mkdir (input_vars_dir)
do_variable_plot ('a', -1, input_vars_dir)
do_variable_plot ('b', -1, input_vars_dir)
do_variable_plot ('c', -1, input_vars_dir)
postcut_vars_dir = 'output/vars_after_cut'
print ('Plotting vars after score > {0} cut in {1} ...'.format (
bdt_cut, postcut_vars_dir))
mkdir (postcut_vars_dir)
do_variable_plot ('a', bdt_cut, postcut_vars_dir)
do_variable_plot ('b', bdt_cut, postcut_vars_dir)
do_variable_plot ('c', bdt_cut, postcut_vars_dir)
fig = v.create_correlation_matrix_plot ('train_sig')
fig.savefig ('output/correlation_matrix-signal.png')
fig = v.create_correlation_matrix_plot ('train_data')
fig.savefig ('output/correlation_matrix-background.png')
def do_eff_plots ():
eff_plot_dir = 'output/eff_vs_vars'
mkdir (eff_plot_dir)
print ('Plotting efficiency vs training variables...')
def do_eff_plot (name, cut_level):
# Validator doesn't know how to make efficiency plots, but it can
# provide the histogram objects with which to do so.
bins = 25
vals = v.data['test_sig'].eval (name)
range = (vals.min (), vals.max ())
h_input = v.get_Hist ('test_sig', name,
bins=bins, range=range,
)
h_cut = v.get_Hist ('test_sig', name,
bins=bins, range=range,
cut='scores > {0}'.format (cut_level)
)
eff = h_cut / h_input
fig = plt.figure ()
ax = fig.add_subplot (111)
plotter = Plotter (ax)
plotter.add (eff)
plotter.finish ()
ax.set_xlabel (name)
ax.set_xlim (range)
ax.set_ylabel ('efficiency')
ax.set_ylim (0,1.05)
ax.set_title ('efficiency vs {0}'.format (name))
fig.savefig ('{0}/eff_vs_{1}.png'.format (eff_plot_dir, name))
plt.close (fig)
do_eff_plot ('a', bdt_cut)
do_eff_plot ('b', bdt_cut)
do_eff_plot ('c', bdt_cut)
### Make Training Samples ###
train_sig_sim = make_sig_sim ()
test_sig_sim = make_sig_sim ()
bg_sim = make_bgsim ()
### Datamerize ###
train_data = make_data ()
test_data = make_data ()
### Make them Sets ###
train_sig = DataSet (train_sig_sim)
train_data = DataSet (train_data)
test_sig = DataSet (test_sig_sim)
test_data = DataSet (test_data)
bg = DataSet (bg_sim)
# create a data directory, if one does not already exist
mkdir ('hesdata')
print ('Saving data in hesdata/ ...')
# save these files
save (train_sig, 'hesdata/train_sig.ds')
save (train_data, 'hesdata/train_data.ds')
save (test_sig, 'hesdata/test_sig.ds')
save (test_data, 'hesdata/test_data.ds')
save (bg, 'hesdata/bg.ds')
bg_sim = make_bgsim ()
### Datamerize ###
train_data = make_data ()
test_data = make_data ()
### Make them Sets ###
train_sig = DataSet (train_sig_sim)
train_data = DataSet (train_data)
test_sig = DataSet (test_sig_sim)
test_data = DataSet (test_data)
bg = DataSet (bg_sim)
# create a data directory, if one does not already exist
mkdir ('data')
print ('Saving data in data/ ...')
# save these files
save (train_sig, 'data/train_sig_woreco.ds')
save (train_data, 'data/train_data_woreco.ds')
save (test_sig, 'data/test_sig_woreco.ds')
save (test_data, 'data/test_data_woreco.ds')
save (bg, 'data/bg_woreco.ds')
