Machine Learning Notes Thiago

Accessing Skylake

72 Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz cores!

ssh access.grid.unesp.br

ssh phi02.ncc.unesp.br

Installing ROOT

export PATH=/opt/miniconda3/bin/:$PATH

# Conda
conda remove --name btagging --all
conda create --name btagging python=3.6 numpy
source activate btagging

# ROOT
mkdir $ROOTPREFIX
cd $ROOTPREFIX
git clone http://github.com/root-project/root.git
cd root
git checkout -b v6-13-02 v6-13-02
cd ..
mkdir builddir
mkdir rootinstall
cd builddir
cmake3 -DCMAKE_INSTALL_PREFIX=$ROOTPREFIX/rootinstall \
-D tmva:BOOL=OFF \
-D x11:BOOL=OFF \
-D minuit2:BOOL=ON
../root
cmake3 --build . -- -j 96
cmake3 --build . --target install
source $ROOTPREFIX/rootinstall/bin/thisroot.sh

Javier Suggestions

Ntuple maker:
https://github.com/DeepDoubleB/DNNTuplesAK8/blob/94x/NtupleAK8/src/FatJetInfoFiller.cc

presentation:
https://indico.cern.ch/event/739204/contributions/3051141/attachments/1675866/2690557/BTV_doubleb_June272018.pdf

model:
https://github.com/jmduarte/DeepJet/blob/master/modules/models/convolutional.py#L442

training script:
https://github.com/jmduarte/DeepJet/blob/master/Train/deepDoubleB_reference.py


https://github.com/jmduarte/DeepJet
https://github.com/jmduarte/DeepJetCore

our ntuples (LPC):
/eos/uscms/store/group/lpchbb/20180401_ak8/

mixed+merged ntuples (LPC)
/eos/uscms/store/group/lpcjj/doubleb_merged_h_q_lessQCD/

data collection numpy arrays (LPC):
/eos/uscms/store/group/lpcjj/convert_20180401_ak8_deepDoubleB_db_cpf_sv_reduced_lessQCD_dl4jets_train_val
/eos/uscms/store/group/lpcjj/convert_20180401_ak8_deepDoubleB_db_cpf_sv_reduced_lessQCD_dl4jets_test

Some sites to read

• http://anotherdatum.com/vae.html
• https://jaan.io/what-is-variational-autoencoder-vae-tutorial/
• https://www.jeremyjordan.me/variational-autoencoders/

-- trtomei - 2018-09-28