Pycabnn (Python tool for Constructing an Anatomical Basis of a Neural Network; pronounce it as "pie cabin") is a Python 3 package to aid building a physiologically realistic neural network model. Pycabnn is for setting up a structural basis of a model, such as placing neurons in space and determine their connectivity, based on anatomical constraints. It is primarily developed for a model of the granular layer in the cerebellar cortex [1]. However, we tried to make it as adaptable as possible to other circuit models.

For a detailed explanation about algorithms used/implemented, please check out our paper [2].

Pycabnn is written in pure Python 3 and depends on the following packages:

• numpy
• scikit-learn
• joblib
• tqdm (for progress bars)
• pandas and pytables (for saving results in HDF5 files)
• ipyparallel and cloudpickle (for utilizing multiple CPUs)

You will also need to install the followings to run example scripts:

• matplotlib (for plotting)
• NEURON (for reading a parameter file)
• Jupyter notebook (for reading notebooks)

We do not have proper setup.py for installation yet. For usage, please check out example scripts:

Run generate_cell_position.py as:

python generate_cell_position.py -p PARAM_PATH -o OUTPUT_PATH all

We included some parameter data in test_data/params for PARAM_PATH. The results will be save in a ".npz" file, cell_positions.npz, within OUTPUT_PATH, along with text files.

Run run_connector.py as:

python run_connector.py -i INPUT_PATH -o OUTPUT_PATH -p PARAM_PATH all

We included some test data in the test_data directory: Use test_data/cell_position for INPUT_PATH and test_data/params for PARAM_PATH. OUTPUT_PATH can be anywhere. This script will generate the connectivity data as tables in HDF5 and text files.

• Ines Wichert - Connectivity generation

• Sanghun Jee - Cell position generation

• Sungho Hong - Project conception and supervision

• Erik De Schutter - Project supervision

This project is licensed under the MIT License - see the LICENSE.txt file for details

• Ivan Raikov
• Peter Bratby

1. Sudhakar, S.K., Hong, S., Raikov, I., Publio, R., Lang, C., Close, T., Guo, D., Negrello, M., and De Schutter, E. (2017). Spatiotemporal network coding of physiological mossy fiber inputs by the cerebellar granular layer. PLoS Comput. Biol. 13, e1005754.
2. Wichert I., Jee S., De Schutter, E., and Hong S. (2020) Pycabnn: Efficient and extensible software to construct an anatomical basis for a physiologically realistic neural network model. Front. Neuroinform. 14, 31.

Written by Sungho Hong, Computational Neuroscience Unit, Okinawa Institute of Science Technology

March, 2020