Proportional weight change

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Proportional weight change

Rath
Nice job on the software. It looks helpful for learning about neural networks and testing ideas.  I’ve been fiddling with it.  You have a number of synapse types to choose from as well as neuron types.  Is there a synapse type where the connection strength changes proportionally to the signal it carries?  Is there a neuron type where the threshold changes proportionally to the signal it transmits?  Thanks.
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Proportional weight change

jyoshimi
Administrator
Hi Rath,

It looks like the  main discussion is happening over at youtube.   My notifications in nabble still don't seem to work....

Basically there is nothing that does this directly, but a lot of things that are kind of in the ballpark.   Oja's rule for example takes in to account other incoming weights and so it normalizes.  Decay rule will decay activation more depending on how active the neuron is.   I think Grossberg's ART networks had some kind of adaptive threshold.    Homeostatic neural mechanisms also come to mind.  These are all cases where different values change proportionally to current state.

As Zach said, your best bet is to  build custom things to do this sort of thing, e.g. the adaptive threshold idea.  And as I said in another thread, I'm working on a system to make customization even easier than it is now.     Eventually something you do could make it in the main release. We're always happy to add more stuff to Simbrain!

Cheers,

- Jeff
Loading...