2012/learningchips12

Experimental setup

We will have three spiking multi-neuron chips stacked and connected in a loop with a mapper.

  • 2xIFSLWTA
    • 128 leaky linear Integrate & Fire neurons
    • Each neuron has 32 synapses
      • 2 excitatory synapses with short term depression and facilitation.
      • 2 inhibitory synapses
      • 28 excitatory bistable synapses with long-term plasticity.
  • IF2DWTA
    • 32x64 leaky Integrate & Fire neurons (arranged in a 2 dimensional grid)
    • Each neuron has 3 synapses
      • 2 excitatory synapses with short term depression and facilitation.
      • 2 inhibitory synapses
  • IFMEM
    • 32 adaptive exponential Integrate & Fire neurons
    • Each neuron has 32 synapses with 5bit programmable weights
    • The weights can be programmed by SW algorithms that use the timing of the chip input/output spikes (STDP)
  • AER mapper
    • Allows arbitrary connectivity in the network.
      • Has a limit of the connectivity fan-out (one to many connections), although considerably large enough for practical purposes. Exact number will be put here soon.

System requirements for users

The software we will be using is developed in python and the drivers are developed in Linux (tested on Ubuntu/Fedora?/openSUSE). Therefore in order to implement your own models on hardware via the provided software, you will need a PC with Linux installed on it. In case you do not have one, here are your alternatives.

  • PC users
    • Install Ubuntu next to your Windows operating system.
    • Run Linux from a USB pen-drive. If this is your preferred option, please bring with you atleast a 4GB (8GB preferred) USB stick. We can then install the operating system on the USB stick and boot from USB. We will show you how to do the installation at the workshop.
  • MAC users
    • Since MAC OS is UNIX based you can run most of the code we have developed. We have not done thorough testing, but our drivers are compatible with MAC, as long as the necessary packages are installed.
    • Is it possible to boot from a USB stick on a mac ? If yes, the I suppose you could do the same as PC users and run linux from a USB stick. Please bring a long a 4GB (8GB preferred) USB stick with you.We can then install the operating system on the USB stick and boot from USB. We will show you how to do the installation at the workshop.

Installation

Standard packages you require to install

python, python-dev, python-numpy, python-scipy, pydot, python-matplotlib, chaco, python-lxml

IPython : This is a very useful command-line tool to run python commands and scripts like you do in matlab.

Custom Python development packages related to the hardware can be found in attachment. Download and uncompress  ini_vlsi_software_complete.tar.gz:

$> tar zxvf ini_vlsi_software_complete.tar.gz

Download and install

pyNCS
python package for spike-based neural networks on hardware.
pyST
Spike Toolbox for pyNCS.
pyAex
The communication drivers to send and receive spikes to the hardware setup.
pyAMDA
The configuration drivers to configure the parameters of the neurons and synapses on the chips.

How to install.

One step installation.

$> cd extracted/folder/
$> sudo ./install_pyncs.sh

Alternatively you can install each package individually.

$> cd package/you/want/to/install
$> sudo python setup.py install

If you have Red Hat based distribution, like Fedora, you can install the packages by first becoming super-user:

$> cd package/you/want/to/install
$> su
Password:
$> python setup.py install

Check installatoin

To check that you have properly installed the required packages you can start python and import the package:

$> python
   
   >>> import pyNCS

If you get no errors the installation was successful. Try the same with all the required packages.

Documentation

A complete documentation of the software can be found  here.

A step by step tutorial can be found  here.

Start your experimentation

Please download the ExpSetup?.zip and extract the files. You can use these files to initialize the setup and start your experiments.

Attachments