monkeys learn to use brain-machine interface
k.p.collins at worldnet.att.net
Thu Oct 16 08:05:20 EST 2003
Some additional comments
[Thoughts to myself, shared],
and typos fixed below.
"KP_PC" <k.p.collins at worldnet.att.net> wrote in message
news:KDvjb.181204$0v4.13912084 at bgtnsc04-news.ops.worldnet.att.net...
| "BilZ0r" <BilZ0r at TAKETHISOUThotmail.com> wrote in message
| news:Xns94168DEE266C8BilZ0rhotmailcom at 220.127.116.11...
| | Catastrophic forgetting is what happens
| | when you train a neural net to do a
| | certain task, and then we you train it to
| | do a new task, it is completely
| | unable to do the old task.
| | This can be overcome by training the
| | net at the old task intermitantly
| | between trials and the new task.
| | Interestingly (and this is what I was
| | getting at in my OP), when you do the
| | intermitant 'reminding' of the old task,
| | so you get a net that can do the
| | two tasks, the nodal connection
| | strengths are completely altered from
| | the | original task. i.e. for a net to
| | remember 2 things, the nodes arn't a
| | superposition of the patern for task 1
| | and task 2, they are completely
| | remodeled.
| Thanks for this explanation.
| From the perspective in which I work,
| my reaction was, "Of course" - be-
| cause the 'complete remodeling'
| "addresses" the underpinning inform-
| ation 'via' its cross-correlations.
This happens NL-P-ly, up to the volitional-
'limit' that is the "volitional diminishing-
returns`" threshold [AoK, Ap7].
NL-P-ly be-cause, as the 'density' of
cross-correlation increases, more
information-processing work is required
to establish [and maintain] the cross-
This's why it's good to escape the Dictates
of local-WDB2T, and just range-widely as
a matter of course.
Local-WDB2T will 'bash' you, but, as you
Honor it, non-local-WDB2T will Honor you :-]
Then, local-WDB2T will understand, and
its propensity to 'bash' you will diminish.
'Course, if you 'kick it in the shins' [as I
often do], the diminishing of the 'bashing'
will take a little longer.
Sometimes, such lengthening of conver-
gence is Necessary - to keep things from
'flying off the handle, and to give folks the
'nudge' that instantiates elevation of V D-R
Sometimes, it's just what's required to
stay-Alive, and, in such, it's a measure
of selfish-'weakness', and its prolongation
of convergence is detrimental to the General
That's what I did in my previous post, quoted
I Apologize to all.
In my Desperate circumstances, saying it
otherwise still eludes me.
So, in 'times' like these-days, I just stand-
and-fight - in order to just 'hold-onto' Life.
But, when one ranges-widely, it's pretty-
Hard to not See the 'slime' in-there.
It's just that addressing it directly is
Jesus Taught Correctly with respect to
such: ~~"Turn the other cheek, go the
extra mile, and give 'em your shirt, too."
Forgive me, Lord, for not ranging-widely
enough into the Wisdom that you Died to
Communicate to us.
| The analogous thing occurs within
| biological nervous systems.
| NDT's view on it is briefly discussed
| in AoK, Ap6.
| Biological nervous systems don't do
| it like neural nets - they don't 'cram'
| everything into a single collection of
| 'nodes' and their interactions.
| In biological nervous systems, there's
| a lot of hierarchical processing, which
| is, nevertheless, integrated via global
| TD E/I-minimization.
| It's the global-integration, via TD E/I-
| minimization, that rigorously-couples
| all of the hierarchical stuff across its
| hierarchy, and with respect to physical
| reality - via TD E/I-minimization's rig-
| orous coupling to WDB2T.
| Global-integration via TD E/I-minimiza-
| tion 'knows' that Correct solutions cor-
| respond to WDB2T ^ -1, so they just
| 'cut-to-the-chase', 'blindly' and auto-
| matically 'seek' TD E/I-minimization,
| and information-convergence occurs
| as a 'blindly'-automated by-product
| of such TD E/I-minimization.
| What I've been working to do with
| NDT's synthesis of the Neuroscience
| experimental results is to show folks
| thatm, iff they acquire an understand-
| ing of how and why nervous systems
| process information via 'blindly'-auto-
| mated TD E/I-minimization, they can
| achieve 'transcendance' with respect
| to the 'blind'-automation.
| In the dark-'light' of the Savagery that
| the 'blind'-automation precipitates with-
| in 'blindly'-automated Human interactive
| dynamics, elimination of the 'blind'-auto-
| mation is easily seen to be a Worthy
| I'm sorry that my reply [...] addresses
| the neural-net focus of yours [only peripherally]. I long ago <<<<
| took a look into 'neural nets' and found
| their approach lacking-necessary-sub-
| stance, in particular, in the way they
| were trying to do everything within a
| single collection of 'cloned' 'nodes' - so
| I 'looked-elsewhere'.
| I've also been long-aware that the work
| I've done has fed-back into 'neural-net'
| efforts. It's been routinely both Sorrow-
| ful and funny to see the AI folks' efforts
| 'discovering' what's been handed-to
| them in NDT's stuff.
| It's been as if folks've seen the work I've
| done as being 'exempt' from any need
| for citation - as if I'm some sort of 'daddy'
| whose only purpose has been to 'give
| other researchers' their weekly-'allowances',
| and that, because I'm this 'daddy', 'there's
| no need' for otherwise Accepted Priority.
| It's funny to me that it's been so.
| And it's been killing-me, 'two'.
| Part of it is that folks presume that the
| traditional route has been open to me,
| but it has not been.
| They feel it's 'OK' to usurp the work I've
| done, 'because', 'all I'd have to do to re-
| ceive credit for the work I've done would
| be to go to grad school' - 'merge with the
| way things are supposed to be'.
| But I tried all of that back in the 70's, and
| 'the system' rejected me - be-cause my
| work was, then, already so far ahead of
| 'the system' that no one who acted on be-
| half of 'the system' could See its stuff.
| They 'just' rejected it all, abjectly.
| So, how could I have 'merged with the sys-
| Now, there's this tacit 'conspiracy' that
| derives in the fact that so many folks've
| 'borrowed' from the work I've done, without
| crediting it, that doing so, now, is 'unthink-
| able, because everyone knows that doing
| so would result in a pan-Academia 'scandal'.
| Part of what's funny, though, is that, if I 'go
| dark', folks bash me for 'withholding' the
| work I've done :-]
| Don't I realize that I'm their 'daddy', and that,
| 'of course', I have to just give them their
| 'allowances'? :-] [show them how to do
| their work, so that they can 'earn' their
| I've been trying =HARD= to avoid 'scandal' -
| to "hold open the door", but it's obvious that
| that's why folks have come to treat me as
| their 'daddy'.
| It's been funny.
| But it's also been killing me.
| [Please Forgive me my Lament, but
| it's what seeing the "of course" stuff
| in your reply, and seeing how old the
| same-stuff is in NDT, evoked within
| me. And please don't receive my
| comments as if they are directed
| 'to you', personally. They are not. They
| are directed to 'the system's coersed,
| and coersing, consensus, that sanctions
| such theft of Life itself from folks like me.]
It is directed to 'the wind' that "the beast",
Abstract Ignorance [the absence of under-
standing of how and why nervous systems
process information via 'blindly'-automated
TD E/I-minimization within nervous systems
which, nevertheless, do process information
via 'blindly'-automated TD E/I-minimization]
breathes into Human interactive dynamics.
| ken [K. P. Collins]
More information about the Neur-sci