comp_preds_bkg.extend(output.T)
comp_true_bkg.extend(batchY)
# plot the comparison to the truth
scatter(comp_preds, comp_true, [0.0, 1.00], [0.0, 1.00], "Prediction", "Truth",
"Approximation comparison", "plots/approx_vs_truth_deep_fromLoad.pdf")
comp_preds = [d.item(0) for d in comp_preds]
difflist = [(p - t) for p, t in zip(comp_preds, comp_true)
if (math.fabs(p - t) < 0.0001)]
comp_preds_bkg = [d.item(0) for d in comp_preds_bkg]
difflist_bkg = [(p - t) for p, t in zip(comp_preds_bkg, comp_true_bkg)
if (math.fabs(p - t) < 0.0001)]
hd_hist([difflist, difflist_bkg], 'plots/approx_vs_truth_diff.pdf',
[-0.00005, 0.00005], [0.0, 1100.0], "Approx. difference", "Events",
np.arange(-0.00005, 0.00005, 0.0001 / 350), ['signal', 'background'])
# Training analysis
f_in = open('training.pkl', 'rb')
training = pickle.load(f_in)
count = [n for n in range(len(training[0]))]
# Plot loss history
fig = plt.figure()
plt.plot(count, training[0], '-')
fig.suptitle("")
plt.xlabel("Epoch")
plt.ylabel("Loss")
plt.savefig("plots/loss_history.pdf")
plt.clf()
etabins = np.linspace(-4.0, 4.0, num=100)
mbins = np.linspace(0.0, 200.0, num=100)
ktbins = np.linspace(0.0, 100000.0, num=100)
sig_data = joblib.load('../data/signal_data_gpd.p')
bkg_data = joblib.load('../data/background_data_gpd.p')
sig_data = sig_data[:2000]
bkg_data = bkg_data[:2000]
sig_eta = [e[0] for e in sig_data]
sig_et = [e[1] for e in sig_data]
sig_m = [e[2] for e in sig_data]
sig_kt = [e[3] for e in sig_data]
bkg_eta = [e[0] for e in bkg_data]
bkg_et = [e[1] for e in bkg_data]
bkg_m = [e[2] for e in bkg_data]
bkg_kt = [e[3] for e in bkg_data]
hd_hist([sig_et, bkg_et], 'plots_gpd/et_comp_gpd.pdf', [20.0, 400.0],
[0.0, 600.0], "$E_{T}$ GeV", "Events", etbins,
['signal', 'background'])
hd_hist([sig_eta, bkg_eta], 'plots_gpd/eta_comp_gpd.pdf', [-4.0, 4.0],
[0.0, 100.0], "$\eta$", "Events", etabins, ['signal', 'background'])
hd_hist([sig_m, bkg_m], 'plots_gpd/m_comp_gpd.pdf', [0.0, 100.0], [0.0, 600.0],
"Mass [GeV]", "Events", mbins, ['signal', 'background'])
hd_hist([sig_kt, bkg_kt], 'plots_gpd/kt_comp_gpd.pdf', [0.0, 100000.0],
[0.0, 1000.0], "$K_{T}$", "Events", ktbins, ['signal', 'background'])
es, rb, nSig, nBKG = scanPoint(s, comp_true, comp_true_bkg)
xvals_orig.append(rb)
yvals_orig.append(es)
xvals_drone.append(xvals_drone_inner)
yvals_drone.append(yvals_drone_inner)
# Plot
fig = plt.figure()
for p, q in zip(xvals_drone, yvals_drone):
plt.plot(p, q, '-b')
plt.plot(xvals_orig, yvals_orig, '-r')
fig.suptitle("")
plt.ylabel("Signal efficiency")
plt.xlabel("Background rejection")
plt.savefig("plots_gpd/rocs.pdf")
plt.clf()
'''
comp_preds = [d.item(0) for d in comp_preds]
difflist = [(p-t) for p, t in zip(comp_preds, comp_true) if (math.fabs(p-t) < 0.0001)]
comp_preds_bkg = [d.item(0) for d in comp_preds_bkg]
difflist_bkg = [(p-t) for p, t in zip(comp_preds_bkg, comp_true_bkg) if (math.fabs(p-t) < 0.0001)]
print len(difflist)
print len(difflist_bkg)
hd_hist([difflist, difflist_bkg], 'plots_gpd/approx_vs_truth_diff.pdf'
, [-0.00005, 0.00005], [0.0, 1100.0]
, "Approx. difference", "Events", np.arange(-0.00005, 0.00005, 0.0001/350)
, ['signal', 'background'])
# Training analysis **************************************************************
f_in = open('training_gpd.pkl', 'rb')
training = pickle.load(f_in)
count = [n for n in range(len(training[0]))]
sig_pt = [e[0] for e in sig_data]
sig_eta = [e[1] for e in sig_data]
sig_minPT = [e[2] for e in sig_data]
sig_maxPT = [e[3] for e in sig_data]
sig_minETA = [e[4] for e in sig_data]
sig_maxETA = [e[5] for e in sig_data]
bkg_pt = [e[0] for e in bkg_data]
bkg_eta = [e[1] for e in bkg_data]
bkg_minPT = [e[2] for e in bkg_data]
bkg_maxPT = [e[3] for e in bkg_data]
bkg_minETA = [e[4] for e in bkg_data]
bkg_maxETA = [e[5] for e in bkg_data]
hd_hist([sig_pt, bkg_pt], 'plots/pt_comp.pdf'
, [0.0, 10.0], [0.0, 1000.0]
, "Mother $p_{T}$ GeV", "Events", ptbins
, ['signal', 'background'])
hd_hist([sig_eta, bkg_eta], 'plots/eta_comp.pdf'
, [1.0, 6.0], [0.0, 400.0]
, "Mother $\eta$", "Events", etabins
, ['signal', 'background'])
hd_hist([sig_minPT, bkg_minPT], 'plots/minpt_comp.pdf'
, [0.0, 10.0], [0.0, 5000.0]
, "min. $p_{T}$ GeV", "Events", ptbins
, ['signal', 'background'])
hd_hist([sig_minETA, bkg_minETA], 'plots/mineta_comp.pdf'
, [1.0, 6.0], [0.0, 400.0]
, "min. $\eta$", "Events", etabins