Hello everyone,
Happy New Year! I want to create a simple neural network consisting of a single neuron connected by multiple inputs (spike generators). I am aware that I can create single spike generators and then connect them but it becomes infeasible as the size of input grows. Is there a way to create a network of maybe 100 spike generators connected to a single LIF Neuron without explicitly creating 100 of them? Once the connections are made is there a way that the network is updated when the weights are updated. I want the network to get updated as I update the weights. Now, I update the weights and reset the network to create it again with new weights. Any help would be appreciated. Thanks in advance!
Hi!
There is no way to get 100 spike generators without creating them individually. But as they are pretty efficient with respect to memory and performance, I also don't see a problem with that.
Other than that, the description of what you want to achieve is far too vague to give you more detailed advice. Please consider sending a figure of your planned network topology and/or (pseudo) code that shows what you want to do. Also have a look at the example networks for NEST, as those contain a lot of useful design patterns for all kinds of simulation scenarios:
https://nest-simulator.readthedocs.io/en/latest/examples/index.html
Cheers, Jochen!
On 07.01.20 16:46, shanka19@purdue.edu wrote:
Hello everyone,
Happy New Year! I want to create a simple neural network consisting of a single neuron connected by multiple inputs (spike generators). I am aware that I can create single spike generators and then connect them but it becomes infeasible as the size of input grows. Is there a way to create a network of maybe 100 spike generators connected to a single LIF Neuron without explicitly creating 100 of them? Once the connections are made is there a way that the network is updated when the weights are updated. I want the network to get updated as I update the weights. Now, I update the weights and reset the network to create it again with new weights. Any help would be appreciated. Thanks in advance! _______________________________________________ NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
-- Dr. Jochen Martin Eppler Phone: +49(2461)61-96653 ---------------------------------- Simulation Laboratory Neuroscience Jülich Supercomputing Centre Institute for Advanced Simulation
------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------ Forschungszentrum Juelich GmbH 52425 Juelich Sitz der Gesellschaft: Juelich Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzender des Aufsichtsrats: MinDir Volker Rieke Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender), Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, Prof. Dr. Sebastian M. Schmidt ------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------
Hi,
Perhaps it helps to note that 100 Poisson generators of rate, say, 10 Hz, is equivalent to one generator with rate 1000 Hz.
Kind regards, Charl
On Thu, Jan 16, 2020, at 11:13, Jochen Martin Eppler wrote:
Hi!
There is no way to get 100 spike generators without creating them individually. But as they are pretty efficient with respect to memory and performance, I also don't see a problem with that.
Other than that, the description of what you want to achieve is far too vague to give you more detailed advice. Please consider sending a figure of your planned network topology and/or (pseudo) code that shows what you want to do. Also have a look at the example networks for NEST, as those contain a lot of useful design patterns for all kinds of simulation scenarios:
https://nest-simulator.readthedocs.io/en/latest/examples/index.html
Cheers, Jochen!
On 07.01.20 16:46, shanka19@purdue.edu wrote:
Hello everyone,
Happy New Year! I want to create a simple neural network consisting of a single neuron connected by multiple inputs (spike generators). I am aware that I can create single spike generators and then connect them but it becomes infeasible as the size of input grows. Is there a way to create a network of maybe 100 spike generators connected to a single LIF Neuron without explicitly creating 100 of them? Once the connections are made is there a way that the network is updated when the weights are updated. I want the network to get updated as I update the weights. Now, I update the weights and reset the network to create it again with new weights. Any help would be appreciated. Thanks in advance! _______________________________________________ NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
-- Dr. Jochen Martin Eppler Phone: +49(2461)61-96653
Simulation Laboratory Neuroscience Jülich Supercomputing Centre Institute for Advanced Simulation
Forschungszentrum Juelich GmbH 52425 Juelich Sitz der Gesellschaft: Juelich Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzender des Aufsichtsrats: MinDir Volker Rieke Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender), Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, Prof. Dr. Sebastian M. Schmidt
NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
Hi,
as long as I didn't miss a significant change in the way Poisson generators are implemented, please be extremely careful with this! Yes, Charl is correct, in terms of statistics it does not make a difference, _but_ in terms of performance it makes a big difference due to the way Poisson generators are implemented. Without going into the details (with which other are much more familiar anyway), always try to use the least possible number of Poisson generators (1x1000Hz >> 100x 10Hz)! Otherwise your simulations might be terribly slow.
Note however, that for spike generators, the above does not apply, as far as I know.
Best, Jakob
On 1/16/20 11:43 AM, Charl Linssen wrote:
Hi,
Perhaps it helps to note that 100 Poisson generators of rate, say, 10 Hz, is equivalent to one generator with rate 1000 Hz.
Kind regards, Charl
On Thu, Jan 16, 2020, at 11:13, Jochen Martin Eppler wrote:
Hi!
There is no way to get 100 spike generators without creating them individually. But as they are pretty efficient with respect to memory and performance, I also don't see a problem with that.
Other than that, the description of what you want to achieve is far too vague to give you more detailed advice. Please consider sending a figure of your planned network topology and/or (pseudo) code that shows what you want to do. Also have a look at the example networks for NEST, as those contain a lot of useful design patterns for all kinds of simulation scenarios:
https://nest-simulator.readthedocs.io/en/latest/examples/index.html
Cheers, Jochen!
On 07.01.20 16:46, shanka19@purdue.edu wrote:
Hello everyone,
Happy New Year! I want to create a simple neural network consisting of a single neuron connected by multiple inputs (spike generators). I am aware that I can create single spike generators and then connect them but it becomes infeasible as the size of input grows. Is there a way to create a network of maybe 100 spike generators connected to a single LIF Neuron without explicitly creating 100 of them? Once the connections are made is there a way that the network is updated when the weights are updated. I want the network to get updated as I update the weights. Now, I update the weights and reset the network to create it again with new weights. Any help would be appreciated. Thanks in advance! _______________________________________________ NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
-- Dr. Jochen Martin Eppler Phone: +49(2461)61-96653
Simulation Laboratory Neuroscience Jülich Supercomputing Centre Institute for Advanced Simulation
Forschungszentrum Juelich GmbH 52425 Juelich Sitz der Gesellschaft: Juelich Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 Vorsitzender des Aufsichtsrats: MinDir Volker Rieke Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender), Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, Prof. Dr. Sebastian M. Schmidt
NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
NEST Users mailing list -- users@nest-simulator.org To unsubscribe send an email to users-leave@nest-simulator.org
-
Charl Linssen -
Jakob Jordan -
Jochen Martin Eppler -
shanka19＠purdue.edu