Neural-numerical-replicability

This project aims to provide tools to analyse the influence of numerical imprecision under different platforms in the context of neural computational replicability, taking as a case study a small neural network composed of Hodgkin-Huxley-type neurons. The simulated model and their parameters are described in JCN_2019_v4.0_appendix_Eqs_Parameters.pdf file. The project was implemented using C/C++ program language. The Ordinary Differential Equations (ODEs) were solved using Runge–Kutta fourth-order method (RK4), implemented by Boost C++ library, specifically Boost.Numeric.Odeint module.

Software requirements:

• C/C++ compiler (ex: gcc)
• Make builder
• Boost C++ libraries (http://www.boost.org)
  ODEint module, to solving ordinary differential equations, is included in the distribution

The project code includes the following files:

Project Tree

• SourceCode directory

  • checkActualPrecision.h
  • HH_BBT2017_allP.cpp
  • iappDist_allP.cpp
  • iappDist_allP.h
  • Makefile.win
  • SpikeTrain_allP.cpp
  • SpikeTrain_allP.h

• figures directory

  • Fig1_as.m
  • Fig2_as.m
  • Fig2_rp.m
  • Fig3_as.m
  • Fig3_rp.m
  • Figure1.png
  • Figure1_dt005.png
  • Figure2.png
  • Figure2_dt005.png
  • Figure3.png
  • Figure3_dt005.png

• JCN_2019_v4.0_appendix_Eqs_Parameters.pdf

How to compile the C++ source code

1. Edit the make file Makefile
   • path for the compiler and Boost library.
   • To use double, long double or Boost precision, you should comment/uncomment the lines (9-16) based on your choice.
2. Execute the make command.
3. Make sure that your HH_BBT2017_allP.exe file was created
4. For simulations with 100% excitatorys neurons with vInh = 70 mV (Figure 1 and Figure 2B) type the following line code: HH_BBT2017_allP.exe -pExcN 1.0 -vInh 70

• Then, case double precision was seleted, It will create four files in the results directory, :
  • HH_BBT_rk4_dt0100_100,0,vI70,t=8s_double_IappDES,Epis.txt
  • HH_BBT_rk4_dt0100_100,0,vI70,t=8s_double_IappDES,Iapp.txt
  • HH_BBT_rk4_dt0100_100,0,vI70,t=8s_double_IappDES.txt
  • HH_BBT_rk4_dt0100_100_0_vI_70_t8s_double_IappDES_Spikes.m

5. For simulations with 80/20% Excitatory/Inhibitory with vInh = 70 mV (Figure 2C) type the following line code: HH_BBT2017_allP.exe -pExcN 0.8 -vInh 70

• Then, It will create four files inside the results directory:
  • HH_BBT_rk4_dt0100_100,20,vI70,t=8s_double_IappDES,Epis.txt
  • HH_BBT_rk4_dt0100_100,20,vI70,t=8s_double_IappDES,Iapp.txt
  • HH_BBT_rk4_dt0100_100,20,vI70,t=8s_double_IappDES.txt
  • HH_BBT_rk4_dt0100_100_20_vI_70_t8s_double_IappDES_Spikes.m

Note: Case you set long double or Boost precision (check step 1), the name of your files will be:
- *t=8s_long_double_*.txt
- *t=8s_boost_double_*.txt

How generate the figures

After generated the files for each platform and each precision

• Figure 1

1. Open the Fig1_as.m file from figures directory on Matlab software.
2. Set the variables 'path_to_directory' on the lines 5 (for Windows file), 18 (for MacOs file) and 31 (for Linux file) for each platform.
3. Verify that the names of the files read are the same within the directories. Line 6 for Windows, line 19 for MacOs and line 32 for Linux file.
4. Compile the file Fig1_as.m to generated the Figure 1.

• Figure 2

1. Open the Fig2_as.m file from figures directory on Matlab software.
2. Set the variables 'path_to_directory' on the lines 5 (for Windows file with neurons 100% Excitatory) and line 18 (for Windows file with neurons 80/20% Excitatory/Inhibitory).
3. Verify that the names of the files read are the same within the directories. Line 6 for neurons 100% Excitatory on Windows platform and line 19 for neurons 80/20% Excitatory/Inhibitory on Windows platform.
4. Compile the file Fig2_as.m to generated the Figure 2 (Without the flowchart and the Raster Plot).
5. Change the file name (line 6 and line 19) to files with name dt = 0.05; and the variable dt (line 31) for 0.05 to generate the figure with dt = 0.05.

• Figure 2 - Raster Plot

1. Open the Fig2_rp.m file from figures directory on Matlab software.
2. Set the variables 'path_to_directory' on the lines 4 (for Windows file with neurons 100% Excitatory) and line 10 (for Windows file with neurons 80/20% Excitatory/Inhibitory).
3. Verify that the names of the files read are the same within the directories. Line 5 for neurons 100% Excitatory on Windows platform and line 11 for neurons 80/20% Excitatory/Inhibitory on Windows platform.
4. Compile the file Fig2_rp.m to generated the Figure 2 (Without the A,S plot and the flowchart).

• Figure 3

1. Open the Fig3_as.m file from figures directory on Matlab software.
2. Set the variables 'path_to_directory' on the lines 5 (for Windows file with neurons 100% Excitatory with Long Double precision), line 18 (for Windows file with neurons 80/20% Excitatory/Inhibitory with Long Double precision), line 31 (for Windows file with neurons 100% Excitatory with Boost precision) and line 44 (for Windows file with neurons 80/20% Excitatory/Inhibitory with Boost precision).
3. Verify that the names of the files read are the same within the directories. Line 6 for neurons 100% Excitatory on Windows platform with Long double precision, line 19 for neurons 80/20% Excitatory/Inhibitory on Windows platform with Long double precision, line 32 for neurons 100% Excitatory on Windows platform with Boost precision, line 45 for neurons 80/20% Excitatory/Inhibitory on Windows platform with Boost precision.
4. Compile the file Fig3_as.m to generated the Figure 3 (Without the Raster Plot).
5. Change the file name (line 6, line 19, 32 and 45) to files with name dt = 0.05; and the variable dt (line 57) for 0.05 to generate the figure with dt = 0.05.

• Figure 3 - Raster Plot

1. Open the Fig3_rp.m file from figures directory on Matlab software.
2. Set the variables 'path_to_directory' on the lines 3 (for Windows file with neurons 100% Excitatory with Long Double precision), line 9 (for Windows file with neurons 80/20% Excitatory/Inhibitory with Long Double precision), line 15 (for Windows file with neurons 100% Excitatory with Boost precision) and line 21 (for Windows file with neurons 80/20% Excitatory/Inhibitory with Boost precision).
3. Verify that the names of the files read are the same within the directories. Line 4 for neurons 100% Excitatory on Windows platform with Long double precision, line 10 for neurons 80/20% Excitatory/Inhibitory on Windows platform with Long double precision, line 16 for neurons 100% Excitatory on Windows platform with Boost precision, line 22 for neurons 80/20% Excitatory/Inhibitory on Windows platform with Boost precision.
4. Compile the file Fig3_rp.m to generated the Figure 3 (Without the A,S plot).