home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!BEN.DCIEM.DND.CA!MMT
- Message-ID: <9212182052.AA04259@chroma.dciem.dnd.ca>
- Newsgroups: bit.listserv.csg-l
- Date: Fri, 18 Dec 1992 15:52:13 EST
- Sender: "Control Systems Group Network (CSGnet)" <CSG-L@UIUCVMD.BITNET>
- From: mmt@BEN.DCIEM.DND.CA
- Subject: Re: Memory in the model
- Lines: 79
-
- [Martin Taylor 921218 15:30]
- (Bill Powers 921218.1030)
-
- I have a problem with the specific detail of Bill's proposal. Not with
- the principle.
-
- >There is one version of the HPCT model that we have tossed around
- >for some time without really deciding about it, in which the
- >world we experience is ALWAYS the modeled world, PLUS whatever
- >error comes from lower systems. It works like this:
- >
- > |
- > ref|sig
- > |
- > --percept-->COMPARE-->error--
- > | |
- > | |
- > INPUT FUNC OUTPUT FUNC
- > | -| |
- > | ------- MODEL ------------
- > | ->correction-/ |
- > | /
- > **P** **R**--<--------- (P and R added by MMT)
- > | |
- > | ref|sig
- > | |
- > \ --error<--COMP<--percept<----
- > | |
- > | |
- >
- >
- >The bottom system is mirror-reversed to avoid crossing signals.
- >
- >The higher system controls a perception made of the sum of a
- >MODEL output and the lower-level error signal.
- >
- >The basic idea is that when the higher system acts to control the
- >model, its output enters both the model and the lower systems (as
- >reference signals). If the lower systems succeed in bringing
- >their individual perceptual signals to the demanded reference level,
- >they will send no error signals to the higher system. That
- >means that the model works; the outputs that control it have the
- >correct effects when also used as reference signals for the lower
- >systems.
-
- This seems to work when there is a one-to-one relationship between lower
- and higher systems, but will it work with many-to-many? The reference
- signal in the lower ECS is not the output from the higher, but a function
- of a vector of such outputs from higher ECSs. So when the lower perceptual
- signal matches its reference, the perceptual signal returned may well be
- quite different from the output that is sent (Points R and P that I added
- to Bill's diagram. Likewise the perceptual signal in the higher ECS is
- a function of a vector of signals like P from many lower ECSs.
-
- If all the lower error signals are zero, then the higher one will presumably
- also be zero, but the reverse is not true. The higher ECS can have a zero
- error when none of the lower ones are satisfied. I'm not sure what this
- would do to the correction aspect of the memory module. Should there be
- a correction or not?
-
- And what evokes a specific memory? In the BCP diagram, we mentally inserted
- "reference vector" as an addressing code for a scalar memory output--a
- table lookup function. That seems reasonable, but we were not at all clear
- how it would or could work. And in the context of the planning exercise
- that I described the other day (shopping robots in the supermarket), we
- could not see how to incorporate the real-world constraints presented
- verbally and never before experienced by any ECS operating through a real
- world CEV. That question, also, still hangs in the air.
-
- Basically, I like the concept that the ECS deals mostly with model predictions,
- corrected by error. But I'm not sure what error, or what is the memory.
- Is it a vector of sensory inputs, to be corrected by the vector of incoming
- error signals (or more probably the incoming perceptual signals--more robust)?
- And what evokes one memory rather than another?
-
- Remember my point about the reduction of information rate as we go up the
- hierarchy. It is this that gives play to predictive memory models.
-
- Martin
-