Running the pattern miner

[Nil Geisweiller]

Hi,

I'm trying to use the pattern miner. I'm posting here first before
creating a github issue in case it is a silly problem.

I'm loading the test in the cogserver

opencog> loadmodule
/home/nilg/OpenCog/opencog/build/opencog/learning/PatternMiner/libTestPatternMinerAgent.so

and getting

[2016-06-29 05:43:14:123] [WARN] Unable to load module
"/home/nilg/OpenCog/opencog/build/opencog/learning/PatternMiner/libTestPatternMinerAgent.so":
/home/nilg/OpenCog/opencog/build/opencog/learning/PatternMiner/libPatternMiner.so:
undefined symbol: _ZN7opencog11OBJECT_NODEE

Any idea?

Nil

[Nil Geisweiller]

Actually, I've moved past that issue (I didn't use the right conf), but
I've got another, perhaps more worrisome

[2016-06-29 06:58:08:109] [INFO] Loading module
"/home/nilg/OpenCog/opencog/build/opencog/learning/PatternMiner/libTestPatternMinerAgent.so"
[2016-06-29 06:58:08:109] [ERROR] Caught signal 11 (Segmentation fault)
on thread 140204635740224
Stack Trace:

I guess I'll have to fire up gdb. Meanwhile, any feedback is appreciated.

Nil

[Nil Geisweiller]

Well, I haven't moved past this issue, I think this other error is just
a different manifestation of the same problem...

Nil

[AmeBel]

Hi,

FYI, the in-development patterminer can be found on the PatternMinerEmbodiment branch @ opencog/opencog/ repo.

[Ben Goertzel]

True... though I wonder what incompatibilities that branch has, as
compared to master...

> --
> 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 opencog+u...@googlegroups.com.
> To post to this group, send email to ope...@googlegroups.com.
> 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/387b5bac-3b76-49c0-b650-40d6f8396ea2%40googlegroups.com.
>
> For more options, visit https://groups.google.com/d/optout.



--
Ben Goertzel, PhD
http://goertzel.org

Super-benevolent super-intelligence is the thought the Global Brain is
currently struggling to form...

[Nil Geisweiller]

Apparently this branch was started a year ago.

Shujing, what prevents you from merging it into the master?

Nil

[Ben Goertzel]

Nil, among other factors, she is using a lot of the old Embodiment
code that you deleted from the master branch...

However, the Pattern Matcher itself should not rely on any of this old
Embodiment code, so I think she could likely merge her improved PM
into master...

What she is doing (among other things) is using the PM to recognize
patterns in the Atomspace resulting from a game character's
perceptions and actions, where the game character is running around in
the old Unity3D game world, controlled by the old Embodiment

There is no conceptual reason not to just port all this to Minecraft,
but Shujing hasn't had time to do that, as she's been doing this work
in a hurry as it's part of what she needs to do to finish her PhD
thesis...

At least the above is my understanding; Shujing can correct me if I'm wrong!

> https://groups.google.com/d/msgid/opencog/57737E38.4010904%40gmail.com.

>
> For more options, visit https://groups.google.com/d/optout.

[Shujing Ke]

Hi,

The pure pattern miner test is a pure OpenCog thing, doesn't require the Unity3d game project. It has two modes, a single machine mode, and a distributed mode, see http://wiki.opencog.org/wikihome/index.php/Pattern_Miner   They both have been merged into the main Opencog branch before the end of last year. If you follow this wiki page, please use the one in the main branch. But because I haven't checked it again since this year, so maybe some parts in OpenCog it depends on have changed and the Pattern Miner is still last year's version. I will try to test it again later when I have time.

Yes, the current developing branch is integrating with the  Embodiment client to recognize patterns by unsupervised learning from other scripted NPCs' behaviour in the Unity3d game world for a demo in my PhD thesis. Because I got the impression that the new Embodiment system is in Ros now, and it is for connecting to the robots instead of Unity3d game world.  There are 3 reasons I chose to use the old embodiment + unity3d game world:

1. The demo I am making is to do unsupervised learning from a noisy and multi-contextual environment, needs to be easy to script the NPC agents to do a lot of random different high level actions, like walk to something, pick up something, eat something, drop something, even using magic to heal someone, open chests with a key....It will be hard to script a lot of NPC robots to do all of these.

2. If we want to connect the Pattern Miner to current robotic project, it will also require a computer vision processor to recognize low-level robotic actions first before all the perceived knowledge can all be turned into Atoms in the AtomSpace, which I assume won't become matured and work perfectly with the whole OpenCog in near future.  

3. For a more visible complete demo, after the Pattern Miner found a pattern it need a planner to work together to use the new found pattern in planning, and then a new plan can be sent to the client for execution. Since I am not there anymore, and to debugging the planner and the whole action execution pipeline from OpenCog embodiment to a robot and feedback to OpenCog is very difficult for me. And is probably also require a lot of complex robotic mechanism implementation which I am not good at. 

But once the perceiving modules and action execution modules are all implemented and integrated well with the robots and OpenCog, it should not be hard to integrate my current developing stuff with the new Embodiment system.

BTW, for anyone who is interested about what exactly my current demo can do:

The motivation:

Any other kinds of learning in AI all require manually pick the learning materials, examples for a certain task for the agent to learn from. But real human level intelligence is able to just observe the whole world, without pre-assigned any tasks, just learning every useful things from the world. And once a new task is given, a human can search their memory and find out a solution to solve it. It's also different from sequence recognize, a sequence for a task is not always continuous, it can be disturbed. So the Pattern Miner is aiming at figure out the cause and effect for every single action, and then using planning / reasoning to make them into a sequence when given a task.

The demo content: 

There are two different tasks in the game world: open a chest, heal an animal. 

There are  a lot of  NPCs doing random actions including the two tasks in the game world at the same time, for example: NPC A first pick up an apple; and then walk to a rabbit, apply healing magic to heal it; walk to a yellow key; walk around randomly; drop or eat the apple; walk to a yellow chest; open the yellow chest with the yellow key it is holding. 

At some point, give the AI agent a goal: open the black chest. Then it would be able to know from Pattern Miner that using the same color of key can open the same color of chest. If given a goal to heal the cat instead, the AI agent will also know from the Pattern Miner that applying healing magic to an animal can heal it. 

[Nil Geisweiller]

Hi Shujing,

are you saying that the "pure" pattern miner on the master is identical
to the "pure" pattern miner on the PatternMinerEmbodiment branch?

In other words, we may not need to ever merge this branch to the master?

I tried to compile the PatternMinerEmbodiment branch but got a lot of
troubles, first the following file PAI.MapInfo.h is missing, I tried to
disable the embodiment part but got a cascade of other problems.

Anyway, if the pattern miner on that branch isn't better I obviously
prefer to use the one on the master. Just haven't figured out what is
the missing function corresponding to that symbol
_ZN7opencog11OBJECT_NODEE, but I should soon.

Nil

> *The motivation:*

> Any other kinds of learning in AI all require manually pick the learning
> materials, examples for a certain task for the agent to learn from. But
> real human level intelligence is able to just observe the whole world,
> without pre-assigned any tasks, just learning every useful things from
> the world. And once a new task is given, a human can search their memory
> and find out a solution to solve it. It's also different from sequence
> recognize, a sequence for a task is not always continuous, it can be
> disturbed. So the Pattern Miner is aiming at figure out the cause and
> effect for every single action, and then using planning / reasoning to
> make them into a sequence when given a task.
>

> *The demo content: *

> <mailto:opencog%2Bunsu...@googlegroups.com>.

> >>> To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> >>> 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/387b5bac-3b76-49c0-b650-40d6f8396ea2%40googlegroups.com.
> >>>
> >>> 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 opencog+u...@googlegroups.com

> <mailto:opencog%2Bunsu...@googlegroups.com>.

> > To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> > 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/57737E38.4010904%40gmail.com.
> >

> > For more options, visithttps://groups.google.com/d/optout.

>
>
>
> --
> Ben Goertzel, PhD
> http://goertzel.org
>
> Super-benevolent super-intelligence is the thought the Global Brain is
> currently struggling to form...
>
>

> --
> 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 opencog+u...@googlegroups.com

> <mailto:opencog+u...@googlegroups.com>.

> To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> 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/CALpD4-L4Y21F7FLVO_o8TMxfzOPTSbR5rq6-RpOGpPOpCBeRvQ%40mail.gmail.com
> <https://groups.google.com/d/msgid/opencog/CALpD4-L4Y21F7FLVO_o8TMxfzOPTSbR5rq6-RpOGpPOpCBeRvQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.

[Nil Geisweiller]

On 06/29/2016 01:49 PM, Nil Geisweiller wrote:
> Anyway, if the pattern miner on that branch isn't better I obviously
> prefer to use the one on the master. Just haven't figured out what is
> the missing function corresponding to that symbol
> _ZN7opencog11OBJECT_NODEE, but I should soon.

I feel the dumbest person on earth... That object is

opencog::OBJECT_NODE

!!! Now I just need to find the missing linkage.

Nil

[Nil Geisweiller]

Just so you know I manage to run the pattern miner test!

Nil

[Nil Geisweiller]

Shujing,

I might start hacking the pattern miner on the master soon. Once I do
that there's is no coming back, I mean that merging the pattern miner
from the PatternMinerEmbodiment branch will be a huge pain, so if that's
the plan please let me know. I won't start touching it till I get your
feedback.

Thanks,

Nil

On 06/29/2016 01:20 PM, Shujing Ke wrote:

> *The motivation:*

> Any other kinds of learning in AI all require manually pick the learning
> materials, examples for a certain task for the agent to learn from. But
> real human level intelligence is able to just observe the whole world,
> without pre-assigned any tasks, just learning every useful things from
> the world. And once a new task is given, a human can search their memory
> and find out a solution to solve it. It's also different from sequence
> recognize, a sequence for a task is not always continuous, it can be
> disturbed. So the Pattern Miner is aiming at figure out the cause and
> effect for every single action, and then using planning / reasoning to
> make them into a sequence when given a task.
>

> *The demo content: *

> <mailto:opencog%2Bunsu...@googlegroups.com>.

> >>> To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> >>> 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/387b5bac-3b76-49c0-b650-40d6f8396ea2%40googlegroups.com.
> >>>
> >>> 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 opencog+u...@googlegroups.com

> <mailto:opencog%2Bunsu...@googlegroups.com>.

> > To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> > 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/57737E38.4010904%40gmail.com.
> >

> > For more options, visithttps://groups.google.com/d/optout.

>
>
>
> --
> Ben Goertzel, PhD
> http://goertzel.org
>
> Super-benevolent super-intelligence is the thought the Global Brain is
> currently struggling to form...
>
>

> --
> 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 opencog+u...@googlegroups.com

> <mailto:opencog+u...@googlegroups.com>.

> To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> 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/CALpD4-L4Y21F7FLVO_o8TMxfzOPTSbR5rq6-RpOGpPOpCBeRvQ%40mail.gmail.com
> <https://groups.google.com/d/msgid/opencog/CALpD4-L4Y21F7FLVO_o8TMxfzOPTSbR5rq6-RpOGpPOpCBeRvQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.

[Shujing Ke]

I would suggest not to merge the PatternMinerEmbodiment  into the master. The mining algorithm is not better. It's just integrated with planner and the old PAI. 

Is it urgent that you need to run the pure Pattern Miner (in the master)? I am already finishing the PatternMinerEmbodiment implementation and just need a few more days to make a pretty demo. I can switch back to the master to test the Pattern Miner on master probably next week if you can wait.

To unsubscribe from this group and stop receiving emails from it, send an email to opencog+u...@googlegroups.com.
To post to this group, send email to ope...@googlegroups.com.

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/5773CDDB.3050108%40gmail.com.

[Shujing Ke]

PAI.MapInfo.h is a auto generated file by protobuf I guess. Sometimes it has some compile order dependency problems that if this file is generated later then some part of the PAI will complain about missing it. It's many year ago, it did happen before but I haven't got this problem for a long time... I forgot how to solve it sorry..

[Nil Geisweiller]

On 06/29/2016 04:45 PM, Shujing Ke wrote:
> I would suggest not to merge the PatternMinerEmbodiment into the
> master. The mining algorithm is not better. It's just integrated with
> planner and the old PAI.

OK, thanks for the info.

>
> Is it urgent that you need to run the pure Pattern Miner (in the
> master)? I am already finishing the PatternMinerEmbodiment
> implementation and just need a few more days to make a pretty demo. I
> can switch back to the master to test the Pattern Miner on master
> probably next week if you can wait.

It wouldn't be necessary I can run the demo as expected. Now I just want
to do a bit of cleaning, then implement scheme bindings for it.

Nil

> <mailto:ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>>> wrote:
> > Apparently this branch was started a year ago.
> >
> > Shujing, what prevents you from merging it into the master?
> >
> > Nil
> >
> >
> > On 06/29/2016 10:45 AM, Ben Goertzel wrote:
> >>
> >> True... though I wonder what incompatibilities that
> branch has, as
> >> compared to master...
> >>
> >> On Wed, Jun 29, 2016 at 3:39 PM, AmeBel
> <am...@hansonrobotics.com <mailto:am...@hansonrobotics.com>

> <mailto:am...@hansonrobotics.com

> <mailto:opencog%2Bunsu...@googlegroups.com
> <mailto:opencog%252Buns...@googlegroups.com>>.

> >>> To post to this group, send email to
> ope...@googlegroups.com <mailto:ope...@googlegroups.com>

> <mailto:ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>>.
> >>> 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/387b5bac-3b76-49c0-b650-40d6f8396ea2%40googlegroups.com.
> >>>
> >>> 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 opencog+u...@googlegroups.com
> <mailto:opencog%2Bunsu...@googlegroups.com>

> <mailto:opencog%2Bunsu...@googlegroups.com
> <mailto:opencog%252Buns...@googlegroups.com>>.

> > To post to this group, send email to
> ope...@googlegroups.com <mailto:ope...@googlegroups.com>

> <mailto:ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>>.
> > 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/57737E38.4010904%40gmail.com.
> >
> > For more options, visithttps://groups.google.com/d/optout

> <http://groups.google.com/d/optout>.

>
>
>
> --
> Ben Goertzel, PhD
> http://goertzel.org
>
> Super-benevolent super-intelligence is the thought the
> Global Brain is
> currently struggling to form...
>
>
> --
> 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 opencog+u...@googlegroups.com

> <mailto:opencog%2Bunsu...@googlegroups.com>
> <mailto:opencog+u...@googlegroups.com

> <mailto:opencog%2Bunsu...@googlegroups.com>>.
> To post to this group, send email to ope...@googlegroups.com
> <mailto:ope...@googlegroups.com>

> <mailto:ope...@googlegroups.com <mailto:ope...@googlegroups.com>>.

> 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/CALpD4-L4Y21F7FLVO_o8TMxfzOPTSbR5rq6-RpOGpPOpCBeRvQ%40mail.gmail.com
> <https://groups.google.com/d/msgid/opencog/CALpD4-L4Y21F7FLVO_o8TMxfzOPTSbR5rq6-RpOGpPOpCBeRvQ%40mail.gmail.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 opencog+u...@googlegroups.com

> <mailto:opencog%2Bunsu...@googlegroups.com>.
> To post to this group, send email to ope...@googlegroups.com
> <mailto:ope...@googlegroups.com>.
> 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/5773CDDB.3050108%40gmail.com.
>
> 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 opencog+u...@googlegroups.com
> <mailto:opencog+u...@googlegroups.com>.
> To post to this group, send email to ope...@googlegroups.com
> <mailto:ope...@googlegroups.com>.
> 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/CALpD4-LU0KNqD0kv7hOdcU6ToT4gjJDWVs4eHhVgg_JWFmumVg%40mail.gmail.com
> <https://groups.google.com/d/msgid/opencog/CALpD4-LU0KNqD0kv7hOdcU6ToT4gjJDWVs4eHhVgg_JWFmumVg%40mail.gmail.com?utm_medium=email&utm_source=footer>.

[Linas Vepstas]

On Wed, Jun 29, 2016 at 8:32 AM, 'Nil Geisweiller' via opencog <ope...@googlegroups.com> wrote:

Shujing,

I might start hacking the pattern miner on the master soon. Once I do that there's is no coming back, I mean that merging the pattern miner from the PatternMinerEmbodiment branch will be a huge pain, so if that's the plan please let me know. I won't start touching it till I get your feedback.

It would be easiest, I think, to keep the master version in master, and pull (cherry-pick) changes from master back into embodiment.  Which is what Nil is saying, more or less.

--linas

[Linas Vepstas]

Re pattern mining in general:

The vision I had for pattern mining had more to do with observation, than with exploration.  So, for example, a child watches an adult perform some task, and tries to copy that performance.   Its easier to find out what's important by watching many performances:  one looks for the common, frequent, repeated parts of the performance, and ignores everything that varies or is different from one performance to the next.

A performance could be any physical or verbal task.   The idea is to observe common patterns that are already occurring in the environment.

So I was a little surprised to read that the actual demo does random exploration, and then tries to deduce "what worked".  Yes, clearly, if a task is successfully repeated multiple times in different ways, then one can look to see what is common and frequent in the successful runs.  That is one of the tricks that both scientists and comedians use to discover what works/what is funny.  But random exploration, de novo, is a LOT harder than simply watching someone who already knows how to do something, and imitating them.  So I am surprised a bit...

---linas 

To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CALpD4-L4Y21F7FLVO_o8TMxfzOPTSbR5rq6-RpOGpPOpCBeRvQ%40mail.gmail.com.

[Nil Geisweiller]

On 06/30/2016 02:17 AM, Linas Vepstas wrote:
> It would be easiest, I think, to keep the master version in master, and
> pull (cherry-pick) changes from master back into embodiment. Which is
> what Nil is saying, more or less.

Indeed. Although I should stress that after the changes I'm about to do
right on the pattern matcher master, cherry picking from the other
branch will be a pain. But apparently it's OK since Shujing said that
the one on the embodiment branch doesn't have any new features beside
embodiment specific things.

Nil

>
> --linas

>
>
> --
> 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 opencog+u...@googlegroups.com
> <mailto:opencog+u...@googlegroups.com>.
> To post to this group, send email to ope...@googlegroups.com
> <mailto:ope...@googlegroups.com>.
> 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/CAHrUA37wKEBqzrH5cdJzDXGE5akJ_h_L2_q-NoRsuqC629z9tw%40mail.gmail.com
> <https://groups.google.com/d/msgid/opencog/CAHrUA37wKEBqzrH5cdJzDXGE5akJ_h_L2_q-NoRsuqC629z9tw%40mail.gmail.com?utm_medium=email&utm_source=footer>.

[Shujing Ke]

Yes, after I finished developing the PatternMinerEmbodiment  branch, if we want to do a clean up , I can just move this branch to a private repository and delete it from Opencog repository; or wait longer to see if there is any chance in future to connect it to the new Embodiment with robot.

To unsubscribe from this group and stop receiving emails from it, send an email to opencog+u...@googlegroups.com.
To post to this group, send email to ope...@googlegroups.com.

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/5774DE81.90008%40gmail.com.

[Shujing Ke]

Hi Linas,

The AI agent in the demo doesn't really do exploring by itself. It still just observes other scripted NPC agents doing things. But it is more general, observing without any specific given goals. What makes the demo more special is: every scripted NPC agent is multiple-threading.  There are also a lot of random noises (meaningless actions) among the actions. 

e.g. if there are two tasks:

task A contains 3 actions :{A1,A2,A3}; 

task B contains 2 actions {B1, B2}; 

random meaningless actions: {R1,R2,R3,R4};

A scripted NPC may do actions in such an order: R1, A1,R2,R3,B1,A2,R4,B2,A3,A3. 

The AI doesn't only figure out the most frequent order of sequence, it will also figure out the common preconditions and effects for each action type. So that once given a new task but similar enough to what it has observed, it is able to figure out a new solution by planning / reason.

The original motivation to do this is,  in the OpenCog pet park demo many years ago, for learning by observations, the NPC or player always has to tell the AI agent that "watch, now I start 'fetching' ", and then conduct the sequence of actions to show the AI what is "fetching" and then tell it "now stop 'fetching' " when the sequence is finished. Ben and I really wanted to fix this problem since long ago because it is just too unnatural. In the real world, babies learn most of things by observing parents  without any particular purposes, and parents will not always tell the babies "now I start 'open the door' ", "now I stop 'open the door' " and so on...