Skip to content
Predicting atomization energies of organic molecules with DNN
Branch: training
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
__pycache__
checkpoints
data
.gitattributes
NN_Input_Handler.py
README.txt
bibliography.rtf
example_input.txt
qchem_nn.py
run_qchem_calculations.py
testing.tfrecords
training.tfrecords
validation.tfrecords

README.txt

The purpose of this project is to create a deep neural network for the calculation of the 
atomization energies of molecules. This project is to be implemented into the QCHEM interface
for rapid, accurate calculations of atomization energies. 

NN SPECIFICATIONS:
Input: Binarized coulomb matrices. User only has to provide a .xyz file

Output: QM calculation. The result is a much quicker way of approximating the QM value. 
All initial energy calculations done with QCHEM.

TODO:
TRAIN the DNN
	Trained to ~2% error over validation set
IMPROVE the DNN
	Add dropout
	Utilize testing dataset
IMPLEMENT the network into the QCHEM package
You can’t perform that action at this time.