0:00:14a low uh i'm going to
0:00:16but then you
0:00:18uh
0:00:19our
0:00:20the localisation
0:00:21based is not that are really a where we mainly focus on
0:00:26but an empirical facts for optimal
0:00:28and to a really
0:00:30are
0:00:31the the
0:00:32uh
0:00:33yeah for part right is the from times
0:00:36no
0:00:36the right panel
0:00:38then
0:00:38three and uh and of that
0:00:40probably most of the my
0:00:45that
0:00:46with the soft introduction
0:00:48uh
0:00:49i i don't model for
0:00:51oh
0:00:52that is uh
0:00:54introduced to a it not by really for is uh
0:00:57and then main part of a
0:01:00it is
0:01:01really
0:01:01as we do the using monte carlo simulation
0:01:04and
0:01:05and
0:01:05loose
0:01:09so a some real problem do not not farm the hz now
0:01:13but but is that for
0:01:15a message passing method for vol
0:01:18for a number of the case of to think of
0:01:20there is a
0:01:22a a this is a a a a a a bit is
0:01:24based on that are used
0:01:27i no
0:01:28uh can don't know number of the condition
0:01:30like a piece without didn't
0:01:32uh
0:01:32the the the yeah
0:01:34and many more
0:01:38however the main problem of this method is
0:01:41uh
0:01:42or or or or that really
0:01:44this means that there are quite a them is really that
0:01:47that
0:01:48so
0:01:49and also possible not or to score
0:01:51and
0:01:52not covered
0:01:53you know
0:01:54not problem for of like this at least we have not
0:01:57this brown
0:01:59uh
0:02:01there are a lot of solutions
0:02:02course
0:02:03a standard was that a to partition
0:02:06uh
0:02:07is also possible to sort by splitting the finding season
0:02:10and
0:02:11uh
0:02:13so our research and or
0:02:15to our work
0:02:17some kind of a
0:02:18oh submission now
0:02:20mean
0:02:21simplest method
0:02:23so
0:02:24that are is to it
0:02:26yeah
0:02:27iteration
0:02:27not part of
0:02:28are
0:02:28however
0:02:30if does not very
0:02:31serious problem because
0:02:33oh or anything
0:02:34and it is
0:02:37the the parameter
0:02:39sure
0:02:39or as the probability that a
0:02:41and
0:02:42years
0:02:43and and with a three
0:02:46yeah be a best
0:02:48oh
0:02:48so
0:02:50right
0:02:54and so we propose a one of of
0:02:57this method for
0:02:58as it is
0:02:59same
0:02:59and
0:03:00really
0:03:01for
0:03:03we
0:03:04find a a possible
0:03:07and minimize the error
0:03:09mean
0:03:09to really an estimate
0:03:13so
0:03:15i real star
0:03:16uh
0:03:16simple
0:03:18a a a a a little additional
0:03:20uh uh
0:03:22yeah you or or or time you know
0:03:25lee
0:03:26yeah
0:03:27the the a distance measure
0:03:29we you know probably or not not well
0:03:31all
0:03:33uh this is for this problem so that the and it's no
0:03:37yeah it's on
0:03:38so
0:03:40a a you know the two
0:03:42oh
0:03:43and and they are of this these they need to
0:03:46um are at each other but are not that money
0:03:50are or not
0:03:51and
0:03:52except increased
0:03:53speech and i of this we make this method that
0:03:59or got this this person you not but those are some known
0:04:03and the
0:04:04uh
0:04:07the noise be an four and five years used
0:04:10or
0:04:11we
0:04:11uh a you know the a samples of then
0:04:14yeah
0:04:15there's not sound
0:04:16no for this problem we can use was
0:04:18we don't want
0:04:19or and uh
0:04:21the problem with north wind
0:04:24for these problem
0:04:25recall this
0:04:28um and you also need to find probability of detection probability is the no D
0:04:33when you you know you
0:04:36this is a i don't not know the one
0:04:39no no you
0:04:40only you
0:04:41uh there there is the
0:04:43the distance is smaller than speech right a
0:04:47so and have a model
0:04:49a a right form a very
0:04:52uh
0:04:53i defined also indicate to the action of by it
0:04:57right a
0:04:59there is additional but
0:05:01yeah can like
0:05:02i
0:05:02yeah
0:05:04uh
0:05:05a of all the right or and all the light
0:05:13or are we are going to different problem in
0:05:15and that all models
0:05:16we are going to use
0:05:17mark that one hundred know one
0:05:20just more suitable for this problem
0:05:22uh that means that the
0:05:24uh B
0:05:25for it's no design right prior
0:05:29a not that things you know
0:05:31i are from that one
0:05:33i i i don't is but then she of three
0:05:35like
0:05:36a function of one mission
0:05:39so
0:05:41and define this
0:05:42we can see that for a attention there everything things that and want to know and
0:05:46i the best one oh
0:05:48the prior
0:05:49is
0:05:50still
0:05:51so
0:05:53for we can i
0:05:54the posterior here
0:05:56right
0:05:56everything of all single not your
0:05:59there is potentially a
0:06:01so we can going to push a button or well that's fine that your money you know
0:06:06uh
0:06:08or as it is
0:06:10oh the position and
0:06:11so this is usually that it was a really use less passing met
0:06:16where there
0:06:17and uh also a mission
0:06:20then show that they have these time so i think that that's
0:06:23writer
0:06:24and here is but is so is the line
0:06:26function
0:06:28this probably in his probably due to case
0:06:31the
0:06:32a probability of the detection and
0:06:34a likely with the position of the northeast data
0:06:37you
0:06:38yes i don't no
0:06:39this is not a fixed and just might be but
0:06:43and you are not going to use a a is this you like this only the graph
0:06:48and uh
0:06:49uh
0:06:50this this right and
0:06:51means means that we need more
0:06:53communication
0:06:54not
0:06:55smart minute
0:06:56range
0:06:57so
0:07:00i mentions so that they is that in this addition the joint for really this is not a
0:07:05uh that that was so
0:07:06we need to use message passing that that
0:07:08and
0:07:09the main problem
0:07:10it is
0:07:12uh
0:07:13but this is not a scene
0:07:15little loss
0:07:16so it that if you can actually you can have like that but
0:07:20you know yeah that are there is no
0:07:24i don't see that it
0:07:25cover and that
0:07:26provide
0:07:27a X can also
0:07:29do you are not
0:07:31the
0:07:32and
0:07:34use right of to your rate partition
0:07:37a given by uh
0:07:39really
0:07:40a and
0:07:42a a message data
0:07:44first i
0:07:45i i start at the location of a special case of given to this and to be they are really
0:07:50here
0:07:52you to
0:07:53well
0:07:53so that i
0:07:56really locations so
0:07:58was that you're but you know it i we you really is
0:08:01right of all the right or on incoming that
0:08:06and the
0:08:08message from node you know T is the
0:08:12we found out
0:08:14you about position of T
0:08:16so
0:08:17i want you know
0:08:18so really
0:08:19know
0:08:20information about these that
0:08:22so we make this problem may right
0:08:26to know read out
0:08:28no
0:08:28the value that you don't know so
0:08:31you only information about the about
0:08:34engine out
0:08:36and you also you need to divide
0:08:38uh
0:08:41really
0:08:41a uh with the
0:08:43be having events
0:08:44we use the variation because
0:08:46uh the
0:08:48you need to double count the election
0:08:51we we direction
0:08:52used
0:08:53is
0:08:53one
0:08:57oh what does not lot to lose this is not a wiener problem so
0:09:00and this is also a
0:09:01you are a um very big that's so we need to use
0:09:04uh uh part
0:09:06dimensions
0:09:07are we don't know what that the you can take the reference this this that the same as not are
0:09:12really relation
0:09:14so are final really
0:09:17are a set of five
0:09:19uh
0:09:22a part of that and all that these are all
0:09:24is that right
0:09:26that i mean
0:09:27as we doesn't right race
0:09:29i this power so
0:09:30there's is no problem with this
0:09:34uh kind of the main vector
0:09:36are that are to the
0:09:37and you didn't
0:09:39really do
0:09:40is the probability that a a it is not a
0:09:42i'm straight
0:09:44uh for example
0:09:45that that is do you have just last time three so
0:09:48uh well
0:09:49equal to one
0:09:52and that's another stop
0:09:54is
0:09:55for no please
0:09:56so here we can sixteen and say
0:09:59so that is
0:10:00yeah about it
0:10:01you seen eight of them
0:10:02so
0:10:03um
0:10:05the
0:10:06and you know for the zero or is a equal to zero point five
0:10:11you do not assume at this
0:10:13a for all that
0:10:15and
0:10:17and so
0:10:21as you can see that a lot of spongy you for small that so it is not so
0:10:26uh
0:10:27possible was able to find a for each edge
0:10:29that this
0:10:31yeah and probability
0:10:32set
0:10:33it not seen cases
0:10:34so you are going to
0:10:37remove this restriction and to my is uh it is you
0:10:41or not
0:10:42like there's and okay
0:10:43like
0:10:44but
0:10:45and the location is also used in values and
0:10:49or are all equal to one
0:10:50so
0:10:51it
0:10:52instead of the here by two
0:10:54lower an optimal all minimizing kl divergence
0:10:57with
0:10:59to really find a a set and the
0:11:02the
0:11:03also a are was there a what of the problem is that
0:11:06uh finding can be really and the
0:11:10to do it
0:11:12not i
0:11:13but can simply is
0:11:14in
0:11:16so we we want to use
0:11:18R S I error
0:11:19but means that all
0:11:21uh
0:11:25or you start for not only
0:11:27but in one thing you one
0:11:29find him that
0:11:32is that it
0:11:33okay a six it is one dimensional case
0:11:35and if you have a lot of but is not relevant for this
0:11:40problem
0:11:41so we but here
0:11:43yeah were there
0:11:45and it is probability
0:11:47and
0:11:48mean square error line
0:11:50so we can see that
0:11:52yeah
0:11:53optimal it
0:11:54or or a zero point five
0:11:56and also
0:11:57saying that the remaining lines
0:11:59root square means or or or or or uh
0:12:02yeah yeah
0:12:04a
0:12:05you you part of getting what is special case
0:12:08and it is that a to i
0:12:10you can see
0:12:11okay
0:12:12yeah can the egg
0:12:14okay they were does not
0:12:16very
0:12:17uh
0:12:18see again
0:12:19from
0:12:20root mean and uh uh
0:12:22uh
0:12:23so maybe this the main problem by people use your to do not part
0:12:27really evaluation for any
0:12:29there is usually the only take up one point estimate you
0:12:32yeah a very similar problem
0:12:34but
0:12:34the leave will be over all and you can see this
0:12:38a in this example here we can see a go was you know and
0:12:41at really more
0:12:43uh
0:12:46i really don't have a really but it's really and
0:12:50and
0:12:51to read and
0:12:52the the location of that as you can see
0:12:54uh and it really
0:12:57is
0:12:57very you know that one
0:12:58it's very tight
0:13:02not call a presentation of or is it
0:13:05and to the right don't time really propagation
0:13:08something between
0:13:10a and the
0:13:11uh
0:13:12then be be really so it is that
0:13:15it's not the best
0:13:15because it one is that
0:13:17estimate
0:13:18we you need to a apply for example do to be read
0:13:21then
0:13:23because you that we can never make
0:13:25yeah
0:13:26i zero
0:13:28using this approach
0:13:31so that would use you is that we can minimize
0:13:34yeah that the
0:13:36yeah they were
0:13:38as you need to begin yeah because
0:13:40we're is you a so if you have that one believes that
0:13:44uh
0:13:45make
0:13:46from
0:13:47so are we
0:13:48move to
0:13:49then to find a that for
0:13:50in the politics
0:13:52uh
0:13:54the same or
0:13:55you you know noise
0:13:58oh
0:13:59a not anybody know
0:14:01but
0:14:02i
0:14:03this
0:14:03a is not but then
0:14:05so not a lot
0:14:06we are using vectors
0:14:08okay so i
0:14:09of course
0:14:10and i the
0:14:13for them very problem for each
0:14:15not legally
0:14:17so
0:14:18a lot
0:14:19oh the role of that part of that one of these
0:14:22is
0:14:23a national
0:14:25so
0:14:27yeah
0:14:29this all
0:14:29S or at least you one or at
0:14:32but
0:14:33this is on a segment code
0:14:36so you also sort of think that or
0:14:39uh
0:14:43the that
0:14:44everything
0:14:44is that
0:14:45okay
0:14:45you it's life
0:14:46not difficult but i don't where
0:14:49so but um it's
0:14:51but uh
0:14:53oh
0:14:54similar or of the like that that
0:14:56that different
0:14:58firemen
0:14:59these
0:15:00i
0:15:03so that you can say that a a a a a lot of wants to be the fence on are
0:15:08originally
0:15:09not on some either
0:15:11that
0:15:12the net
0:15:13is
0:15:14and not
0:15:14we didn't so everything
0:15:17paper
0:15:20with the but was also a little
0:15:23a little bit
0:15:24uh so we propose by
0:15:26so
0:15:28uh not part
0:15:30partition
0:15:30and i four for i of a show
0:15:34oh
0:15:35one also has it's
0:15:37very it performance
0:15:39because for a
0:15:40so i
0:15:41error that
0:15:42yeah the for example
0:15:44case
0:15:46uh
0:15:49on the the to model depends on you was not it's not really
0:15:54uh
0:15:55a a net
0:15:56duration does not have
0:15:58but on the network
0:16:01and the
0:16:03that
0:16:04can find a document
0:16:06and can these two area
0:16:09uh
0:16:10that's
0:16:11is is as the same
0:16:12oh this method of does need to write
0:16:14right
0:16:14all
0:16:15well
0:16:17and also we need to
0:16:18i i it's no that
0:16:21i this using it
0:16:22send
0:16:23a
0:16:24for example but
0:16:25what to know that we cannot use lots or not
0:16:28yeah
0:16:29start
0:16:30is
0:16:30you have a lot of neighbours and all
0:16:32that's that's i mean
0:16:34uh i one
0:16:35i can make a wrong decision that
0:16:37a wrong not as well but
0:16:38is that although
0:16:40i are also so
0:16:42a
0:16:43is a lot of
0:16:44note
0:16:46a of the net
0:16:50and
0:16:50seems to you sort this paper
0:16:52for
0:16:54that's to mention
0:16:55it
0:16:55a a similar form
0:16:57a a of fine for all point be possible as something
0:17:00it's
0:17:01one can
0:17:02is
0:17:03or
0:17:04additionally as the little so
0:17:06report
0:17:08two button
0:17:09role in a low dimensional space and then for like in
0:17:12i
0:17:17so are also i that also mention and
0:17:20that's not is that
0:17:22and not like a shot
0:17:24first one
0:17:25this is
0:17:26is this
0:17:27it is that a
0:17:28my
0:17:29really
0:17:29okay
0:17:30you
0:17:31i
0:17:31is it a little
0:17:33at this
0:17:34and it really
0:17:35and also what
0:17:38uh
0:17:38gain to the standard net
0:17:40and say that the product
0:17:42we apply applied
0:17:43the the this of for ldpc codes
0:17:46this case
0:17:47yeah
0:17:47i i the next
0:17:49also be
0:17:51not in for is beyond that
0:17:53oh
0:17:54yeah and so we did
0:17:55the the
0:17:57oh
0:17:59you know case
0:17:59three
0:18:00oh but four a B B or
0:18:02and and be like in case
0:18:04is there is
0:18:07so that you know thing
0:18:08i
0:18:22uh here
0:18:23i
0:18:24do could be this
0:18:26and uh it but you know that it i
0:18:28i
0:18:29yeah but
0:18:30additionally expense
0:18:44a to the resin scaling so
0:18:48this would
0:18:49but here we have a
0:18:51this is completely different for you get this output
0:18:54a probability density function is probably i guess
0:18:57only you
0:18:58uh
0:18:59estimate of the with this
0:19:01so
0:19:02it could be optimal Q the gaussian noise
0:19:05but here we can be completely independent can i
0:19:09and you noise
0:19:11you
0:19:13this is a lot more expensive
0:19:15um
0:19:16the