Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

T h studies #20

Open
wants to merge 11 commits into
base: legacyAnalysis
Choose a base branch
from
+3,450 −216
Open

T h studies #20

Changes from 1 commit
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
49698f1
smaller changes
Nov 11, 2019
7a53b97
forward jet cut in generateJTcut
Nov 11, 2019
f328ad0
all vars for 2017 tH legacy
Nov 11, 2019
38fa853
colors & tH variable sets
Nov 13, 2019
de83543
new netConfig
Nov 13, 2019
3b3d67d
delete old training script
Nov 13, 2019
5ff3634
combined tH ttH training in central region
Nov 15, 2019
9e4a6d9
scripts for central tH training
Nov 15, 2019
39967ea
new tH varsets and STXS colours
Nov 19, 2019
6389b37
final trainings for strategy 1-4
Nov 25, 2019
ac92d25
netconfigs
Nov 26, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
new netConfig
pekramer committed Nov 13, 2019
commit de8354308af576e4be954f4a2066d04784394c98
137 changes: 130 additions & 7 deletions train_scripts/net_configs.py
View file
Original file line number Diff line number Diff line change
@@ -28,34 +28,47 @@
"earlystopping_epochs": 50,
}

config_dict["ttZ_2018_final"] = {
config_dict["ttZ_2018_4node"] = {
"layers": [50,50],
"loss_function": "categorical_crossentropy",
"Dropout": 0.4,
"L2_Norm": 1e-4,
"L2_Norm": 1e-5,
"batch_size": 200,
"optimizer": optimizers.Adagrad(),
"activation_function": "elu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.05,
"earlystopping_epochs": 10,
}


config_dict["ttZ_2018_4node_v4"] = {
"layers": [50,50],
"loss_function": "categorical_crossentropy",
"Dropout": 0.4,
"L2_Norm": 1e-4,
"batch_size": 100,
"optimizer": optimizers.Adagrad(),
"activation_function": "leakyrelu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.1,
"earlystopping_epochs": 50,
}


config_dict["ttH_2017"] = {
"layers": [100,100,100],
config_dict["STXS_2017"] = {
"layers": [500,250,100],
"loss_function": "categorical_crossentropy",
"Dropout": 0.50,
"L2_Norm": 1e-5,
"batch_size": 4096,
"batch_size": 100,
"optimizer": optimizers.Adam(1e-4),
"activation_function": "elu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.05,
"earlystopping_epochs": 100,
}

config_dict["Legacy_ttH_2017"] = {
config_dict["ttH_2017"] = {
"layers": [100,100,100],
"loss_function": "categorical_crossentropy",
"Dropout": 0.50,
@@ -94,6 +107,56 @@
"earlystopping_epochs": 100,
}

config_dict["ttZ_2018"] = {
"layers": [300,200,100],
"loss_function": "categorical_crossentropy",
"Dropout": 0.4,
"L2_Norm": 1e-5,
"batch_size": 4096,
"optimizer": optimizers.Adadelta(),
"activation_function": "selu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.05,
"earlystopping_epochs": 100,
}

config_dict["ttZ_2018_v2"] = {
"layers": [300,200,100,50],
"loss_function": "categorical_crossentropy",
"Dropout": 0.4,
"L2_Norm": 1e-5,
"batch_size": 5000,
"optimizer": optimizers.Adadelta(),
"activation_function": "elu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.05,
"earlystopping_epochs": 50,
}
config_dict["ttZ_2018_v3"] = {
"layers": [200,100,50],
"loss_function": "mean_squared_error",
"Dropout": 0.3,
"L2_Norm": 1e-5,
"batch_size": 5000,
"optimizer": optimizers.Adadelta(),
"activation_function": "elu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.05,
"earlystopping_epochs": 50,
}
config_dict["meme"] = {
"layers": [],
"loss_function": "mean_squared_error",
"Dropout": 0.,
"L2_Norm": 0.,
"batch_size": 5000,
"optimizer": optimizers.Adadelta(),
"activation_function": "elu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.05,
"earlystopping_epochs": 50,
}


config_dict["dnn_config"] = {
"layers": [20],
@@ -183,3 +246,63 @@
"earlystopping_percentage": 0.02,
"earlystopping_epochs": 100,
}
config_dict["tH"] = {
"layers": [100,100,100],
"loss_function": "categorical_crossentropy",
"Dropout": 0.50,
"L2_Norm": 1e-5,
"batch_size": 4096,
"optimizer": optimizers.Adam(1e-4),
"activation_function": "elu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.02,
"earlystopping_epochs": 100,
}
config_dict["ttH_STXS"] = {
"layers": [100,100,100,100],
"loss_function": "categorical_crossentropy",
"Dropout": 0.50,
"L2_Norm": 1e-5,
"batch_size": 4096,
"optimizer": optimizers.Adam(1e-4),
"activation_function": "elu",
"output_activation": "Softmax",
"earlystopping_percentage": 0.02,
"earlystopping_epochs": 100,
}
config_dict["regress_STXS"] = {
"layers": [2500],
"loss_function": "mean_squared_error",
"Dropout": 0.88,
"L2_Norm": 1e-7,
"batch_size": 100,
"optimizer": optimizers.Adam(1e-4),
"activation_function": "elu",
"output_activation": "linear",
"earlystopping_percentage": 0.002,
"earlystopping_epochs": 100,
}
config_dict["regress"] = {
"layers": [25],
"loss_function": "mean_squared_error",
"Dropout": 0.5,
"L2_Norm": 1e-5,
"batch_size": 100,
"optimizer": optimizers.Adam(1e-4),
"activation_function": "elu",
"output_activation": "linear",
"earlystopping_percentage": 0.002,
"earlystopping_epochs": 100,
}
config_dict["regression"] = {
"layers": [500,500],
"loss_function": "mean_squared_error",
"Dropout": 0.5,
"L2_Norm": 1e-5,
"batch_size": 100,
"optimizer": optimizers.Adam(1e-4),
"activation_function": "elu",
"output_activation": "linear",
"earlystopping_percentage": 0.002,
"earlystopping_epochs": 100,
}
21 changes: 15 additions & 6 deletions train_scripts/ttH17_train.py
View file
Original file line number Diff line number Diff line change
@@ -12,10 +12,12 @@
filedir = os.path.dirname(os.path.realpath(__file__))
basedir = os.path.dirname(filedir)
sys.path.append(basedir)
sys.path.append(basedir+"/preprocessing/root2pandas")

# import class for DNN training
import DRACO_Frameworks.DNN.DNN as DNN
import DRACO_Frameworks.DNN.data_frame as df
import preprocessing_utils as pputils

options.initArguments()

@@ -24,12 +26,19 @@

# define all samples
# only ttH sample needs even/odd splitting for 2017 MC
input_samples.addSample(options.getDefaultName("ttH"), label = "ttH", normalization_weight = options.getNomWeight())
input_samples.addSample(options.getDefaultName("ttbb"), label = "ttbb", normalization_weight = 1.)
input_samples.addSample(options.getDefaultName("tt2b"), label = "tt2b", normalization_weight = 1.)
input_samples.addSample(options.getDefaultName("ttb"), label = "ttb", normalization_weight = 1.)
input_samples.addSample(options.getDefaultName("ttcc"), label = "ttcc", normalization_weight = 1.)
input_samples.addSample(options.getDefaultName("ttlf"), label = "ttlf", normalization_weight = 1.)
#input_samples.addSample(options.getDefaultName("ttH"), label = "ttH", normalization_weight = options.getNomWeight())
#input_samples.addSample(options.getDefaultName("ttbb"), label = "ttbb", normalization_weight = 1.)
#input_samples.addSample(options.getDefaultName("tt2b"), label = "tt2b", normalization_weight = 1.)
#input_samples.addSample(options.getDefaultName("ttcc"), label = "ttcc", normalization_weight = 1.)
#input_samples.addSample(options.getDefaultName("ttlf"), label = "ttlf", normalization_weight = 1.)
#input_samples.addSample(options.getDefaultName("ttb"), label = "ttb", normalization_weight = 1.)
#input_samples.addSample(options.getDefaultName("ttb_bb"), label = "ttb_bb", normalization_weight = 1.)

sampleDict = pputils.readSampleFile(options.getInputDirectory())
print "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
print sampleDict
print "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
pputils.addToInputSamples(input_samples, sampleDict)

if options.isBinary():
input_samples.addBinaryLabel(options.getSignal(), options.getBinaryBkgTarget())