2018

SIESTAをインストールの仕方、コンパイルと環境設定など。。。
How to install SIESTA (in Japanese)

2017

Install YCU "Security Check" on Ubuntu 16.04"

Install nupic on Ubuntu

1. NUPIC needs python (2.7), PIP (the python package manager), mysql (database server) and GIT (repository manager).
   sudo apt-get install python python-pip mysql-client mysql-server git
2. Then install nupic with
   pip install nupic (without sudo)
3. Verify what version of nupic you installed. You can use this command:
   python -c 'from pkg_resources import get_distribution; print "nupic:", get_distribution("nupic").version, "nupic.bindings:", get_distribution("nupic.bindings").version'
4. Create a directory and download there the folders with the source files from Numenta's GIT repository
   git clone https://github.com/numenta/nupic.git
   Enter the nupic folder and fetch the correct version (for example 0.6.0):
   cd nupic/
git fetch https://github.com/numenta/nupic
git checkout tags/0.6.0
   (you know your nupic's version from previous step)
5. Test your installation with
   py.test tests/unit (if it does not work, locate it with locate py.test and execute the test with /path/to/py.test /tests/unit)
6. If you do not have errors, you are OK (in case: read the error messages and use numenta wiki or google to fix them).
   

   ---

7. Nupic swarming uses your mysql database. Last step to be fully operational is to activate your mysql server. Use
   service mysql start

8. then veryfy if you can log in with
   mysql -u root -p
   by default you should be able to login with root and an empty (return) password.
   If you can't, but have another mysql account that works, you can tell nupic to use that. Just go and change the login information in the nupic files. Do this
   pip show nupic
   this will show you where is nupic-default.xml. Make a copy of this file and call it nupic-site.xml (should be in the same folder of nupic-default.xml). Now edit nupic-site.xml and change the database mysql informations (change root to your username and the empty password to your mysql password).
   Now you should be able to login in your database with
   mysql -u yourusername -p
   

Install Siesta 4.1 on Linux Ubuntu 16.04

for both Ubuntu 20.04 and 16.04
sudo apt-get install build-essential checkinstall
sudo apt-get install openmpi-common openmpi-bin libopenmpi-dev libnetcdf-dev netcdf-bin libnetcdf-dev libnetcdff-dev libscalapack-mpi-dev libblas-dev liblapack-dev
sudo apt-get install openmpi-doc libopenmpi-dev libmpich-dev
for Ubuntu 20.04 you need these:
sudo apt-get install bscalapack-openmpi-dev libscalapack-mpich-dev

1. download siesta 4.1 (link (or use the command: wget https://launchpad.net/siesta/4.1/4.1-b4/+download/siesta-4.1-b4.tar.gz)
2. extract it (tar -xvf siesta-4.1-b4.tar.gz). You will find a folder named "Obj/" inside the main siesta directory.
3. cd in that folder and then execute the setup script
   sh ../Src/obj_setup.sh
4. then you need an arch.make file. You can create one starting from the DOCUMENTED-TEMPLATE.make file. This file should be in your current folder Obj. Edit this file accordingly to your computer architecture.
5. Once you are ready, save it as arch.make. To compile just use the command
   make
6. (If it does not compile, check the examples gfortran.make and intel.make. If you are using ubuntu you can attempt to use our arch.make, made for Ubuntu 16.04 (it is commented for other versions. In Ubuntu 20.04 probably you need the correct lapack library, generally it is located in /usr/lib/x86_64-linux-gnu/lapack, the version that worked for our ubuntu 20.04 it is zipped here. Copy the contents to the lapack folder.
   If still you cannot compile you have to read the error messages and the manual).
7. Once the compiler finishes, you will have the executable "siesta" in the obj folder.
8. Now you can work. Create and go in another folder. For example
   mkdir ../myWork
cd ../myWork
9. put your .fdf and .psf files there. Make also a symbolic link to the siesta executable, with this command:
   ln -s ../Obj/siesta
10. then run your code. In parallel mode, the command is:
mpirun --host localhost:4 -np 4 siesta <gan.fdf> out.dat
(4 is the number of your processor threads)

Hope it helps !

2015

Multiprocessing for biological neuronal network model, by Sun Zhe

Multiprocessing is a standard library for Python and it is very easy to install into any systems (import multiprocessing). With this library, we can use multiple processors to calculate different processes at the same time.

As an example to demonstrate the use of this library we realized a small neural network, in which each neuron is calculated in one processor independently. Multiprocessing can not only improve the speed performance, but it is a more exact and reliable approach for realistic neurons.
Firstly, we use the Izhikevich model to emulate the individual neuron behavior. This model was proposed by Izhikevich and described in the paper.

We made a python class for an individual neuron (Izhikevich model) IzhikevichClass (by Sun Zhe). Based on this class, we linked a chattering neuron 'm' with a fast spiking neuron 'n'. And we used electric synaptic connection, the difference of the neuron's potential was used as the stimulus signal. In our simulation, we used Gaussian White noise to emulate realistic neural noise, synaptic delay is defined as 0.1 ms.

To simulate the neuron delay behavior and the stimulus from another neuron, we used module 'Value' and 'Array' in multiprocessing. With these two modules, numbers and arrays can be stored in shared memory. And in the modules, we can use 'd' and 'i' to indicate the double precision float and signed integer. For example, in our program we defined:

km=Value('i',0)

'km' is the loop index for the neuron 'm' that indicates how many loops have been processed. Then we create a process object for each neuron, target is a function to be invoked by start() method.

p1 = Process(target=neuronM, args=(VsignalM,UsignalM,VsignalN,UsignalN,ww,SNR))

The method start() will launch the process and the terminate() method will terminate it. Here is the example python file (MultiprocessingForNeuralNetwork.py), it uses the above class and the multiprocessing library. Resulting simulation is in the following figures, where we plotted variation of two coupled neurons. In figure 1 the synaptic strength is 0.1 and the in the figure 2 the value is 0.8. The red and green lines represent the first and second neuron action potentials calculated independently in the two processes.

2014

June 2014: LaTeX with Japanese fonts (linux)

We installed Latex in our Ubuntu 14.04. It was difficoult for us to find instructions for the correct production of Japanese text. So, here we list the things you need to install to write in Japanese with LaTeX.