>>>>> On 09 Feb 1997, KAJ at gnn.com (Kenneth Jenrow) wrote:
KAJ> I am requesting that anyone with some experience with presently
KAJ> available neural network software packages please provide reviews of
KAJ> each. Alternatively, if there is a single package which dominates
KAJ> please provide information about it. I am primarily interested in
KAJ> modelling hippocampal and neocortical networks. Thanks to all who
KAJ> respond.
Whenever you ask for information about any software, please remember to tell
people what kind of computer platform you're using, whether it's a UNIX
workstation, or a PC, or a Mac, or something else.
I have only minor experience with the following, so I can't provide actual
reviews, but in case you don't get any other information, here's a summary of
a few major packages. A longer list of general neural network software and
additional tips and information can be found in the comp.ai.neural-nets FAQ
(frequently-asked questions) list, available from
ftp://ftp.sas.com/pub/neural/FAQ.html. The items below are free software
specifically geared towards biologically-motivated simulations.
GENESIS 2.0.2
A general-purpose neural simulation system. It is useful for a range of
types of simulations, from detailed simulations of parts of single
neurons, to large networks of many abstract neurons. It is used in the
computational neuroscience course offered at the Marine Biological
Laboratory at Woods Hole, MA.
Platforms: UNIX + X Windows
More info: http://www.bbb.caltech.edu/GENESIS/.
NEURON 3.2.2
Another general-purpose simulation system, providing somewhat more
elaborate features (IMHO) in the area of detailed neural simulation than
GENESIS. Like GENESIS, this has been used at the MBL computational
neuroscience course.
Platforms: UNIX + X Windows, NeXT, 486PC's with Linux or MS Windows
More info: ftp://ftp.nnc.yale.edu/neuron/
BIOSIM
A biologically-oriented neural network simulator. From the README file:
"Four neuron models are implemented: a simple model only switching ion
channels on and off, the Hodgkin-Huxley model, the SWIM model and the
Golowasch-Buchholz model as the most enhanced model. Dendrites consist
of a chain of segments without bifurcation. A neural network can be
created by using the interactive network editor which is part of
BIOSIM. Parameters could be changed via context sensitive menus and the
results of the simulation can be visualized in observation windows for
neurons and synapses. Stochastic processes such as noise can be
included. In addition, biologically orientied learning and forgetting
processes are modeled, e.g. sensitization, habituation, conditioning,
hebbian learning and competitive learning. Three synaptic types are
predefined (an excitatatory synapse type, an inhibitory synapse type and
an electrical synapse). Additional synaptic types can be created
interactively as much as desired."
Platforms: UNIX + X Windows
More info: ftp://ftp.uni-kl.de/pub/bio/neurobio/biosim.readme
Many other packages exist that are better-suited to simulating abstract
networks of the back-propagation or self-organizing map variety. Again, the
comp.ai.neural-nets FAQ has a good list of these.
--
Mike Hucka hucka at umich.eduhttp://www.eecs.umich.edu/~hucka University
PhD to be, computational models of human visual processing (AI Lab) of
UNIX systems administrator & programmer/analyst (EECS DCO) Michigan
