Learning Through Target Bursts (LTTB)

This is the accompanying source code of the paper: "Beyond spiking networks: The computational advantages of dendritic amplification and input segregation", PNAS, Vol. 120 (2023) e2220743120, by Cristiano Capone^*, Cosimo Lupo^*, Paolo Muratore, Pier Stanislao Paolucci.

Please give credit to this paper if you use or modify the code in a derivative work. This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.

This work is licensed under a Creative Commons Attribution 4.0 International License.

The code is written in Python and organized as follows. The network model itself is defined into the lttb.py module (which defines the LTTB model object). The model parameters used for the different experiments are organized in different sections of a unique json configuration file.

To easily replicate the experiments presented in the paper we provide Jupyter Notebooks that can be run on their own.

External Dependencies

To run the different experiments the following external libraries should be installed on the machine:

jupyter notebook    # for an easier user interface
numpy               # Basic numerical array library
matplotlib          # Used for producing the various visualizations
json                # Use to parse the model configuration file
tqdm

Bursts & Context Switch

To reproduce the results presented in Figure 3 and Figure 4 of the paper, simply run the associated notebooks: Figure_3.ipynb and Figure_4.ipynb.

Button & Food Task

To reproduce the results presented in Figure 5 of the paper, run the associated notebook Figure_5.ipynb.

To reproduce the results presented in Figure 6 of the paper, for the spiking network, run the associated notebook Figure_6.ipynb, in the LTTS folder.