Hi Johannes,
I'm in the debugging phase of porting Shujing's I-surprisingness code
https://wiki.opencog.org/w/Measuring_Surprisingness to the new pattern
miner (struggling against what seems to be a pattern matcher bug ATM).
In parallel I'm reviewing the literature on surprisingness and writing a
document to go beyond that code. It seems surprisingness is not only
relative to the knowledge available but also to how someone "thinks"
about something. So far I have come to the conclusion that, in its most
abstract form, surprisingness is merely the distance between the
outcomes of 2 inferences, and all specializations of surprisingness can
be derived from that abstraction. We'll see how that goes.
Porting Shujing's code will be for sure ready within 2 weeks. Going
beyond that code is gonna take a lot more time as it is research.
Nil
On 2/4/19 10:33 AM, Johannes Castner wrote:
Hi Nil,
Fascinating indeed!
What is the status of the pattern miner and the interestingness
measure? I'm not asking because I'm in a rush (I still have about two
weeks before I'm ready), before I can use it for my dissertation paper,
I still have work to do with respect to the AtomSpace representation of
the ontologies and also with auto-scaling of my system's memory so that
I can load in the Gazetteer, which is huge. I think that in about two
weeks, I'd like to give that interesting pattern mininer a spin on the
AtomSpace filled with two ontologies along with a huge, global
gazetteer, if I can. Gdelt might also be an interesting source of
knowledge to learn new things about Sustainable Development from
(https://www.gdeltproject.org/). In the meantime, I'm working on
encoding as much about those relationships as there is information for
in the ontology: if a relationship is transitive, symmetric, etc. I want
to at least encode that in addition to the Evaluation PredicateNode
construct, which I understand, but I think that coding more of the
correct structure will help to gain better insight. In the medium term,
I'd be interested in seeing if the full social-cultural-institutional
meaning of particular types of relations (what do they predict, etc.)
could emerge from intense online interactions between people and the
AtomSpace. That is what I'm working towards academically, while trying
to make use of it for business so that I can finance the operation.
Johannes
On Sunday, February 3, 2019 at 6:11:24 AM UTC, Nil wrote:
On 2/2/19 6:26 PM, Johannes Castner wrote:
> Let me ask you this: how much of the work in this book has been
> implemented as part of opencog, thus far:
> https://link.springer.com/book/10.2991/978-94-91216-11-4
<https://link.springer.com/book/10.2991/978-94-91216-11-4>
If you're talking about part III of the book, not much. However PLN
relies now on a much more powerful tool, the URE. At the time of
writing
the book, PLN was hard-coded in C++ and somewhat inflexible.
Back then it had
- Extensional and intensional reasoning
- A bit of contextual reasoning
- Maybe some embryonic causal/spatio/temporal reasoning
- No dynamic control mechanism
The new PLN has
- Extensional reasoning
- No/broken intensional reasoning
- No/broken contextual reasoning
- No causal/spatio-temporal reasoning
- Dynamic control mechanism, capable of meta-learning. :-)
What we have gained is that OpenCog is able to learn how to control its
own reasoning and even rewrite its own reasoning rules. The counterpart
is that we still need to port intensional, contextual, causal and
spatio-temporal reasoning.
If you're interested, there is an example of control meta-learning
https://github.com/opencog/opencog/tree/master/examples/pln/inference-control-learning
<https://github.com/opencog/opencog/tree/master/examples/pln/inference-control-learning>
Nil
> Is the work on this set of ideas dead in the water or is this
actively
> pursued and lastly, may we help with it? We're definitely eager
to help
> where we can, as we're trying to build on these ideas.
>
> Johannes
>
>
> On Saturday, February 2, 2019 at 3:09:54 PM UTC, linas wrote:
>
>
>
> On Sat, Feb 2, 2019 at 4:03 AM 'Nil Geisweiller' via opencog
> <[email protected] <javascript:>> wrote:
>
>
> I'm longing to go back to spatio-temporal reasoning but
I'm involved
> with other matters.
>
> Linas, Amen? Any idea of how this is going?
>
>
> Seems to be dead in the water. Anyway, all that I was
plumping for
> was not really "reasoning" but just simple predicates that
converted
> tow 3D points into "in front of", "behind", "next to" and if
the 3D
> points arrived at different times, "before", "after", just so
that
> natural language would get access to the 3D info. That's all;
its
> pretty minimal.
>
> Perhaps relevant for Johannes is the "electrical part of" demo,
> which shows how to convert a Concept into a Predicate (and
how to
> build complex relations out of simple ones)
>
> https://wiki.opencog.org/w/EquivalenceLink
<https://wiki.opencog.org/w/EquivalenceLink>
> <https://wiki.opencog.org/w/EquivalenceLink
<https://wiki.opencog.org/w/EquivalenceLink>>
>
> -- linas
> --
> cassette tapes - analog TV - film cameras - you
>
> --
> You received this message because you are subscribed to the Google
> Groups "opencog" group.
> To unsubscribe from this group and stop receiving emails from it,
send
> an email to [email protected] <javascript:>
> <mailto:[email protected] <javascript:>>.
> To post to this group, send email to [email protected]
<javascript:>
> <mailto:[email protected] <javascript:>>.
> Visit this group at https://groups.google.com/group/opencog
<https://groups.google.com/group/opencog>.
> To view this discussion on the web visit
>
https://groups.google.com/d/msgid/opencog/479c8040-a764-410d-b3f8-acb9967b357b%40googlegroups.com
<https://groups.google.com/d/msgid/opencog/479c8040-a764-410d-b3f8-acb9967b357b%40googlegroups.com>
>
<https://groups.google.com/d/msgid/opencog/479c8040-a764-410d-b3f8-acb9967b357b%40googlegroups.com?utm_medium=email&utm_source=footer
<https://groups.google.com/d/msgid/opencog/479c8040-a764-410d-b3f8-acb9967b357b%40googlegroups.com?utm_medium=email&utm_source=footer>>.
> For more options, visit https://groups.google.com/d/optout
<https://groups.google.com/d/optout>.
--
You received this message because you are subscribed to the Google
Groups "opencog" group.
To unsubscribe from this group and stop receiving emails from it, send
an email to [email protected]
<mailto:[email protected]>.
To post to this group, send email to [email protected]
<mailto:[email protected]>.
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit
https://groups.google.com/d/msgid/opencog/9f7b0614-8f4e-424f-8db8-5f1e45659cf9%40googlegroups.com
<https://groups.google.com/d/msgid/opencog/9f7b0614-8f4e-424f-8db8-5f1e45659cf9%40googlegroups.com?utm_medium=email&utm_source=footer>.
For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups
"opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit
https://groups.google.com/d/msgid/opencog/e166952f-2e3b-6c33-18b8-e920d1c964cf%40gmail.com.
For more options, visit https://groups.google.com/d/optout.
