This repository is a fork of the simulation code from the paper:
Ring Attractor Dynamics in the Drosophila Central Brain Authors: Sung Soo Kim*, Hervé Rouault*, Shaul Druckmann†, Vivek Jayaraman† * Equal contributors † Correspondence to: [email protected] or [email protected] Science, 2017, doi: 10.1126/science.aal4835
This repository was made to replicate and explore the model for COGSCI 318 - Theoretical Neuroscience at ENS. Specifcially, we perform an ablation study to generate model predictions of the impact of dead neurons on bump activity in the network.
This work is contained in the updated files in the folder ring_attractor. The updated files contain the author's code, as well as the necessary changes to conduct this study.
As in the original code, to run the parameter sweep, specify the type of connectivity model, delta or cosine, and the (integer) degree offset of the input stimulus with a cli command: ./simuself {model} {input_angle}.
However, we add the following functionality:
-
The ability to run a single simulation trial, specifying paramters. This is achieved by specifying the amp and width parameters in the cli command:
./simuself {model} {input_angle} {amp} {width}. -
The ability to run a single simulation trial with a range of neurons turned off. This is achieved by specifying amp, width, the --damaged flag, the degree offset from 0 of the center of the damaged neuron region, and the number of damaged neurons. i.e.,
./simuself {model} {input_angle} {amp} {width} --damaged {center_offset_angle} {num_neurons}. -
The ability to run a single simulation trial with a random number of neurons turned off. This is achieved by specifying amp, width, the --random_damage flag, the number of neurons you want to damage.
./simuself {model} {input_angle} {amp} {width} --random_damage {num_neurons}. -
The ability to run a parameter sweep with damaged neurons. This is achieved by setting the --damaged flag, the degree offset from 0 of the center of the damaged neurons, and the number of damaged neurons i.e.,
./simuself {model} {input_angle} --damaged {center_offset_angle} {num_neurons}. -
The ability to run a parameter sweep with randomly damaged neurons. This is achieved by setting the --random_damage flag, and the number of damaged neurons i.e.,
./simuself {model} {input_angle} --random_damage {num_neurons}.
The single trial simulations produce a file called single_run_activity.dat which contains the activity of each neuron at each time step during a trial. Single damaged runs will produce single_run_damage_activity.dat; single random damage runs will produce single_run_random_damage_activity.dat.
All parameter sweeps output to the terminal, however they can be saved via a > command, i.e., ./simuself {model} {input_angle} > filename.txt
plotting.ipynb contains the code to generate the plots used in our presentation. The videos of the activity one can generate are particularily fun.
To see the original code, checkout their repository, or the original-code branch of this repository.
A word about CMakeLists.txt
The version of this file contained in this branch has been jerry-rigged to run on my local machine. It will likely not work via a naive download. To return it to its original state, simply uncomment lines 35 and 36, and comment out lines 39-53 in ring_attractor/CMakeLists.txt. Also, comment out line 32 in ring_attractor/src/CMakeLists.txt.
A word about bump_sampling
The folder bump_sampling contains the code the authors used to fit multiple bumps on the ring. This was outside the scope of this project, and thus remains unchanged.