Dear Björn and Frank,
could you reproduce the results of the following script:
https://github.com/electronicvisions/spikey_demo/blob/master/networks/stdp.py
I can follow your arguments, but, on first sight, I cannot identify
obvious reasons for this unexpected behavior.
Due to the unknown initialization of correlations stored in the STDP
synapses, please consider sequentially repeating your stimulation
pattern within a single elongated emulation run?
The temporal interval between repetitions should be large enough to
avoid learning between them.
This allows to investigate the long-term limit of weight changes.
If these results are difficult to interpret, you may consider plotting
the synaptic weight over the length of your emulation (here, number of
repeated identical stimulation patterns).
Cheers,
Thomas
Zitat von "Deiseroth, Bjoern" <[log in to unmask]>:
> Hello there, us again :)
>
>
>
> so we understood, that a nearest-neighbor, pair-based STDP is
> running on spikey.
>
> Lets say a neuron "pre" is connected to a neuron "post" via a STDP synapse.
>
>
>
> Attached, we built a network such that the following spike train occurs:
>
>
>
> Pre:
>
> 0.0,15.850000381469727
>
> 0.0,20.149999618530273
>
> 0.0,24.700000762939453
>
> 0.0,29.5
>
>
>
> Post:
>
> 5.0,28.850000381469727
>
>
>
> So changes to the STDP stack should be: a_c for Dt(28,85-24,7) and
> a_a for Dt(28,85-29,5), according to your survey "is 4-bit
> synaptic...".
>
>
>
> Starting at weight 3, it should therefore decreased to 2, or maybe
> remain equal if the difference is not enough, for a single iteration.
>
> Instead, it is always rises to 4.
>
>
>
> Am I missing something?
>
>
>
> Thank you :)
>
> Björn and Frank
>
>
> ########################################################################
>
> To unsubscribe from the KIP-SPIKEY-USERS list, click the following link:
> http://listserv.uni-heidelberg.de/cgi-bin/wa?TICKET=NzM2MDAxIHRob21hcy5wZmVpbEBLSVAuVU5JLUhFSURFTEJFUkcuREUgS0lQLVNQSUtFWS1VU0VSU3LRa7VD1lHq&c=SIGNOFF
########################################################################
To unsubscribe from the KIP-SPIKEY-USERS list, click the following link:
https://listserv.uni-heidelberg.de/cgi-bin/wa?SUBED1=KIP-SPIKEY-USERS
|