This repo is forked from https://github.com/VeriDeep/DLV
The goal is to reproduce the results presented in CAV 2017 paper.
The following instructions are tested on Mac OS X and Debian Linux.
(0) assume miniconda or anaconda is used to setup python environment
(1) create an virtual environment (called DLV or any other name you like) using conda
conda create -n DLV python=2.7 theano=0.9
(2) activate the virtual environment
source activate DLV
(3) install packages available in conda
conda install matplotlib cvxopt scikit-image h5py
(4) install packages available in pip
pythonp install stopit keras==1.2.2
(5) revise default keras config
"image_dim_ordering": "tf", ==> "image_dim_ordering": "th",
"backend": "tensorflow" ==> "backend": "theano"
(6) export z3 python environment
export PYTHONPATH=~/Projects/z3/build/python/
(1) Installation:
To run the program, one needs to install the following packages:
Python 2.7
Theano 0.9
Keras 1.2.2 (Note: the software currently does not work well with Keras 2.x because of image dimension ordering problems, please use a previous 1.x version)
z3
(2) Usage:
Use the following command to call the program:
python DLV.py
Please use the file ''configuration.py'' to set the parameters for the system to run.