Sherlock is a tool for output range anaylsis of Deep Neural Networks with ReLU activation units.
Sherlock is a tool for output range anaylsis of Deep Neural Networks with ReLU activation units.
Sherlock is an output range analysis tool for deep neural networks. The current version can handle only feedforward neural networks, with ReLU activation units.
The only library that is needed is the MILP solver Gurobi. It's free for academic purposes and can be downloaded from here : http://www.gurobi.com/resources/getting-started/mip-basics
Please modify the file Makefile.locale to help us find Gurobi.
For a Mac with the latest gurobi7.52 installed your likely settings will be:
HOST_ARCH=mac64 GUROBI_PATH=/Library/gurobi752
For a linux box, your settins will be:
ARCH=linux64 # if you are using a linux box GUROBI_PATH=/opt/gurobi752
You should feel free to modify these two variables. The Makefile will look for Gurobi headers under
$(GUROBI_PATH)/$(HOST_ARCH)/include
and libraries under
$(GUROBI_PATH)/$(HOST_ARCH)/include
Once these are set, you should type
make
to compile. It should work out of the box.
Please use the command line run_file to run all the benchmarks.
./run_file all
Enjoy the output as it scrolls on your screen :-)
Alternatively, you can run a specific benchmark using the following command :
./run_file <benchmark_no>