Re: I think that relational DBs are dead. See link to my article inside
Date: Thu, 06 Jul 2006 16:14:07 +0200
Message-ID: <e8j5r4$cp4$1_at_ss408.t-com.hr>
Dmitry Shuklin wrote:
> Hi Josip,
>
>>But how about weighted graphs/synapses? >>Pointers/references don't have weights...
>
> You are absolutelly right. Good question.
>
> I have described logical model of OODB. It is not neural model. Neural
> model should be emulated. Of course it can be done by many ways.
>
> If needed to emulate whole bilogical neuron I prefer this:
> Split one neuron into atomic pieces. Synapses, axons, dendrites, ...
... sure, I usually do it somewhat like this:)
(In fact I've started writing an essay about that but didn't get much interest, http://www.vrspace.org/docs/zen_of_self-adaptive_code-jitsu.html)
And I use soft references so I don't need all objects in memory but keep
some MRU cached etc etc.
However, when working with say 10000000 'neurons', I don't even have
enough RAM for references.
It occurred to me that such a reference 'weight' would be quite usefull
hint for decision what to keep in memory.
In OO model such 'weight' could turn to i.e. simple integer referrer
count divided by class tree depth. Or maybe more simillar to synapse
weight - object access count (handle with care).
Or something; I'm improvising:)
FTR simple MRU gives me about 97% cache hit rate.
Well, my 2c, keep up the good work.
Regards... Received on Thu Jul 06 2006 - 16:14:07 CEST