a to now three money and

it it can um

you don't will have and

uh_huh and silky key for so you know Q to assume karen's and come from japan

and in in just present a and we will present the new the or boring with some four

don't think it detection for two days and that you know that

nick details of several is it didn't it

is that the the ten send up to matters

from thirty and you indicate that you have

you have

is ten so is not only um

a two dimension

um at least two maybe

i two dimensions

so

do would problem uh the the than the friend is about how well i with only that we we we

we bold and with tom was this

the and with them the cater

but this

a big three number of component

it's that a a back on the common

on in them first what in them

that you less and the and uh the a thing with them well

a best only one component in the uh huh uh he that's going last

and what but what i went don't that to to to do to to dry but there are with them

can

can a bit a bit number of component

we we assume like uh we will present the

this a we coding it is that

and you we see that for coding it that

more thank with them can can of

fit the data L

and we will as well

in yeah present here the application

with a easy data

the real were application

so just see here

what is it ten so that so it's is that the moon to way and the

for

a the F is this special case of them

canonical polyadic is to compose and when we call

we so on this one these the uh a five

in the model it like this we have to so and we want to to compile ways to for symbol

if just your time your data have that freedom are

three dimensions

and then we want to find three fact

along

for the one and on the first dimension of the tool of but that's for the tree so on

and the call that's a here

is them

you you have a very uh a special case but very special should die that's only

that kernel that were not uh

uh the and and only a on the diagonal of ten so

so you see is it's worse than a in the mood the linear

oh form

and you know two is a at rate the our motivation we we you is here the first very simple

ah

it's sample

assume that you have a ten so with side of one can dress by one can write my one hundred

and a and just a so is composed by three from the value C yeah

three dimensional so

and one from a to have three

and no one thought the have

five

sparse common and

and we we for the common to be called you we a not my symbol

but simple modification like this

you see uh

the come of a symbol here if it um

i go to you have a common and

the come the second come

by the first come plus the second and and with the scary is zero point five

and if we use the less and with um

the i less camp them can find a a a a way

to not

one to plop by and this and then pro

proof was in two thousand

to tell them

and it cannot fit this like up we call it um

you see here

up the fruit ways and um

a a there and the right the C

in did not improve anymore

even if we have here

we

row

uh no five power than uh

but are and i uh in interest

and ask we

which at

this they that we are not able and a bit on this you see here the monte look at this

and with some ball by so

okay this i was is only fall for for and a map

and has done in to tell than five

he's ten this and an to uh and a no

to at the for three dimensional

symmetric and so

and more of in two thousand six

he also able

it's then the multi this and we some bus

with two

to kind of and was some we have uh

the multi cut the least square and with some i'm want to claremont moaned deeply is

camilla but

and we so

and we chess here these want to plot this i L S

this square and the in

and this can okay is better that uh uh a less it with them but is still

cannot be the

the solution i decide yes

you he use easy G "'em" ten minus five

and the here at go and with them

in a point to it out to two thousand nine

it better but this to

this this to and up that five one than five hundred and in into a recent

and we paul than

one and was a bit of the ten

is but uh and is good up to five hundred into

however

like

the when the this event that of the the clear less and was an that this D very similar to

the

is a on the a less and some that's mean

in me that this and with some

to to a that the the fact that

need to inverse of very last scale matters

if you have a high dimensional so

and you have high come and

you have you have

in the the number of common in this last

and

the

you have to go out with them

is

but on the that

the idea but the the have go like with the with that you best only one component

at it it's race and sell this at the hair the to go with them

have a come that's in the cost but it

it men

a lot of

the both trace and

the clearly a the clear less and was a list

can

but but up uh with free to a it's way and but is to have

a hack "'em" with this in a cost

and the and i was

but what um

well well that's

but uh we propose a and with them that's have some

in the in this reason that's

the at the new and with a suit

but that then the here that's going with them

well

list can with this in cost than that yeah you're like a resume

this is the new

this our motivation

and of of we don't we also have a here the whole one it was on for some the P

M F three and so

and we some but by but the row and that that doesn't

no uh

no but it that T seven and the but see if to get a

okay again to it's is

um but uh to F P and to but now screen

we best in two thousand than a to of than ten in nice gas

and we we we'd

we

yeah

we don't

present all of this and with them

it's could be that the second

uh

or is that a sum

so we see here

well in a in order to derive a new entry time we consider again the

problem the nonnegative

the nonacoustic a trip this programming

like

it's like this we will abide is fun so

uh with three D the

it's square matrix a policy nonnegative symmetric

as i was to different methods and be the best that with item one

and we propose the the recursive approach to solve this problem

oh

first what it uh we have

it at this step one we compute that's

solution of this

but this gradient

so it on the element of this vector is

nonnegative

so this orders distill it at you know we'll be the solution of this but

oh difference and

however

usually it's it could be like this we

that's this they have some

the estimated date

element in this vector and we we have to

two set

one is

i i by

with the in des up normally it like

and they negative

uh as due da

a to

and this it uh this this as a

a well up but an only uh this this these negative and this see it says nonnegative

elements

and

we on

and not a a small problem that's model uh i P be problem with a a what

the

by a a K and K way it

where K the number of active

and one here

and was on this one

in the read seas

a pro

the egg with them like this

you see yeah

a compute the solution and then five

a remove

give only the

you only the non at

am and and then on again

so on

this re would be the will bit this real and to yeah

this that i M T

yes

so you see here

that the eye or less and the clear less of that on the component

and

the here at school i are let the that the two that only one component

and we're new act we will be the new it with them in this

the and we dedicate possible we that they K come

and we assume that

ah

and the

yes i and i and these it

it the fact that and but with

only

"'kay" come and

to be estimated

and we

so the the the the C P i got easy the and F model into two points

one these that the um

one part composed Y

by like come around but it not

in this test

and then not about it does

compose Y

a component

like this

and we define the the to do or

but that's a the the used as a

it you see here the dress you that's so is

we we

is

planned and pressed by

and and T a model

an net of model but with a low

come than is only take a component

is that a

is is only K can is that a a common

so you see here and we

from the cost function fall

for these ten so we uh we we see here

with we want to spend the conference

and we will we convert this cost person to

to this

to this

to just fall

to this song what we do

okay you C

so if it better why the vector i best the

but but the we if you better why the fact that i i and

i and R

uh a when we've press and we can from but this but for sent to this fall

this the and be form and then we can use the egg be

the because see a problem

we at i as and

oh at to just on this problem

and and with them like this

oh okay

oh sorry

the that's that be like this

because

but the but to be here you see the vet to be here

is is uh by on

rest of you ten so

multiplied with

a cocker our products of on the animal

a on the fact that it's set the fact and

and it's K we for the paul high dimensional a so we need to be love

this takes so

and it could demand high and with a an that file

and the one we we we one he that's we

we don't want to to be a to time so

we don't want to be the rest you ten so

and how to how to remove this

we spread that the feedback vector

the fire stress

by

reply

the if you

we the different nice enough

the was you ten

and we we had

uh if

the is present for the fire matters here

C you raw data

and this i

this it the

okay

i agree is bounded location here

but for here

use the case and these uh

do not the high how to mark broader and this

you know the card to dot product is step

all four on the fact that it's of the father and

and this also

on this you how to mark

or or fall on the fact that but is that the from and

so you have a is present fall

it decide for first and for the for a part of a that

five matches

and a

and that we play

in

and i could be a problem for

and we from a with a place that there the with some four

to press

uh i

"'kay" can it is that the on component

and you see here

but it's it's rice sent to look at the that and we select

that's of says

from a component

and then compute like to

yeah my two

oh U

we

yeah

free fine here and then we hope that the fact the

i and a

by

you in the

and can and

and and do this

and to the

will be replicate into a common would be okay that is

that's

the

okay i will explain

and the egg with them these

in this okay if the right within five

but

would be to um

problem for the problem of that for the the and with them alive live it becomes easier

the uh this the the identity matrix with I

uh i think and

the uh going to control of the the this and then they met with with the amount

with this garment maker

and the size of this

this product with that though

i

and by

by well i se and by a

and no i by by K

and then in it K of

this

this i and he's ready last are

this in also very locked the you C

the problem is

uh the problem that is points and

this step maybe

you might high can be and the problem i

we want to

you do this

this that

how to do this

we define this one

is that little back on that

on the role we back only one row by row

and established

new

with

the top with new the new and before i three problem paul

one role

for one rows

in this case you have

a silly a fast

he's only four gram uh and easy four

is the only for gamma we we already removed that acted out border no cortical product

oh wow identity matrix

a sigh i

i se and

active and here

yeah

and the nice nest i next next i we will present how to to

"'kay" command from its of set up

a common

normally we we would that

we than there

normally you see here

the this real we we have that and do you on the common

like this

so

this can be certified

the and

and

the number

the number of them

the number of i you numbers of the case is that you

that

the number of combination

but

i

"'kay" we taken from that

a set of site

so i i

false and if you have take "'em" and and you best only to common

it just K you have

mm forty five

combination

yes just used

finished

a D oh okay K you we have to us another give high up a bit of i know guy

of so that's less than like this we divide that says i'll from one to octave in

into a into to separate into

it is the fact but

this is that it's to be

a a normal less all i'll L a possible

and and and of the K and not a a a to that we

we select

"'kay" highly colour already

correlated component

from a component

from a common or

we can see a

we have many guys select

a but a and

"'kay" K K E

come on from a common a

we to here for

a coding data

and it would be you

the fine fine ten so a to so ten so

with side one hundred fall

for every four on the dimension

and the what on it

if we have here ten component

i don't we also for to the common to be calling yeah

and here you see

uh the

what deep look at effect of them is not

in of fit the data of or what

you the also

problem for the a less

here here you have to a

you have here the the

here here the your with them

and here these the

with K

is weak a four we have this for this can wasn't

and we we K seven we have

this

like

and we K ten we had this line

and the clear less we have

this i

so you using we will put pop here and with them

is

so

but that then the are you less

but are better than the to go ahead with "'em" of got better than the light and them

but

uh sometime time is also but than than the pls and and we some

easy to a bit own application for easy day that i we take can they that from

uh from a of

from the two not from model

and he the the data i like this you you

the data

it you you you recall of from

when you put on you audio how your hands

a a yeah or right hand or left hand then you accommodate that

and more of a can be the uh

in that's in that of fate go

the that they could and

from forty it's of less and we have to task for left and right

for left and right hand

and you have to thirty T for channel

can we have D for yeah sin for channel

so T type time frame rate T one frequency bin from the to estimate the hot

and here we had twenty four

eight mister one

jen before we have we have fourteen subjects

and two task

and we we want to five

for the dominant for this data

you see here

because this late that have really needed

spectral and temporal component

using it

we you

our algorithm weekly

this is not hand

it uh

nist

we here

and we

with K by

week at yeah you go to a three

we see here

we see here the to and in the spectrogram and to common and in them

for a common of sex two gram and for common in this uh and the temporal program

uh in in and

not did D spectral and and is a temporal come and you see

calling it because is very similar

yes and we

we already ready i is that lead the component

this for a is at the get the yeah man "'em" and is a be that we don't

i for the query less about the query less the also what for the data

but you see the for the multi the F

K L can of well for this they now be go

the we we can see though clean near common an here

a no so for the the same

problem for them one look at this

anyway

come and that this

here is not

it's red

well the um

the coding a common and

and the the the data

and i L S the so

and the next that we close the

the easy this at six no from

by on the feature we attracted

by and F

and the and light we see here it we

with the multi that the

K K with the multi that if L west

it's only seventy five also so eighties

seven

at this a i not seventy five here

and to what the i he's of yeah and with our you of the we we have ten I

to but i ninety two or ninety three

sending about greasy

yeah

so

we conclude this uh present a we already propose is then

the new the new and with some of that

uh a between them oh component

but you on the it could be a problem

and we so

it's ten this and want to of this row by row

and we were well here some set so on

so approximate or to select the number of common and to be attracted to be a that

thank you