Skip to content

Latest commit

 

History

History
51 lines (38 loc) · 1.74 KB

README.md

File metadata and controls

51 lines (38 loc) · 1.74 KB
Raw

Soruce code of DeepDendrite (built up based on CoreNeuron) and the demos in the paper.

src: source code of DeepDendrite and the modified NEURON simulator.
NOTICE: To perform AI tasks with DeepDendrite, we should use our modified NEURON simulator.
Figure5: Model used in Figure5 in the paper, i.e. spine-level human pyramidal models.
Figure6: Model used in Figure6 in the paper, i.e. image classification on MNIST dataset.

1. Compile NEURON simulator (modified from NEURON 7.6):

Dependencies:

CMake >= 3.10.2
MPI >= 3.0
Python >= 3.6

Compile commands:

cd src/nrn_modify   
./configure --prefix ../../install --without-iv --with-paranrn --with-nrnpython=`which python`  
make -j8  
make install

2. Compile DeepDendrite:

Dependencies (the same as CoreNEURON):

CMake >= 3.10.2
MPI >= 3.0
PGI OpenACC Compiler or NVIDIA HPC SDK
CUDA Toolkit >= 10.1

Compile commands:

cd src/DeepDendrite  
modify the install.sh file to load correct modulefile of PGI or NVHPC SDK (starting with "/path/to/your") and set the install path and mod file path in cmake command (defaults are "install" and "src/all_mechanisms")
source install.sh

3. Evaluate the run time of solving equations with various methods:

Follow the README.md in "Figure_3_4"

4. Run the demo of spine-level model:

Commands:

cd Figure5/run_samples
python3 run.py

For more details please see Figure5/run_samples/README.md

5. Run the demo of image classification task:

Commands:

cd Figure6/run_samples
python3 run.py (for training and testing)
python3 run_test.py (only for testing, with pre-trained model weights)

For more details please see Figure6/run_samples/README.md