►
From YouTube: SimPEG Meeting April 2th Beta
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
F
C
F
C
A
F
C
It's
making
a
very
rough
estimation
of
the
biggest
heigen
value
of
the
data
misfit
and
the
rigor
a
different
function,
and
after
that,
you,
like
here's,
a
beta
of
a
show
is
basically
saying
it's.
It's
basically
said
saying:
if
you
want
it
to
be
10
times,
the
ratio
of
those
two
biggest
eigenvalue
OSA.
F
C
Actually,
like
a
yeah,
it's
like
you,
you
take
your
random,
random,
random
vector
and,
and
you
can,
you
come
and
you
come
and
you
compare
the
two
second
second
derivative
with
compare
multiply.
It
multiply,
method
vector
size,
so
I
can
actually
for
me.
What
I
did,
for
example,
is
that,
like
as
a
victory
random
is
like
yeah?
C
If
you,
if
you
do
it
many
times,
it's
gonna,
give
you
a
better
and
better
estimate,
it's
kind
of
like
it's
like
eight,
you
samplings
found
the
probability
function
of
that
of
that
function,
but
and
basically
like
if
you
take
care
like
what
you
hope
is
that,
like
this
eigen
as
this
eigen
values,
gonna
be
busy
street
or
that
sampling.
But
here
we
just
I
think
that
in
the
queuing
burger.
C
G
C
C
F
Well,
I
mean
with
the
implemented
this
in
some
fashion
or
other,
and
you
know
ever
since
the
magnetics
papers-
yes,
but
still
it's
yet
come
yeah,
it's
always
matter
of
knowing
what
is
actually
being
implement.
So
that
is
how
it's
being
done
here
and
then
for
the
data.
Did
you
use
the
J
trap
estimate
of
J
transpose
J
your
well
you've
got
I've,
got
two
norms
right,
yeah.
A
F
F
A
C
D
E
I'm
working
on
mostly
inversion
course
at
the
moment,
but
I
was
actually
helping.
Mike
Mike
wants
to
like
I,
do
a
grande
source
p.m.
inversion,
just
say
that
basically
same
setup
with
DC,
but
rather
than
using
DC
code.
He
wants
to
use
e/m
code,
so
it
requires
a
bit
of
functionality,
additional
sources
and
receiver
we're
getting
there.
E
So
Mike
worked
on
the
line
integration
code,
so
we
got
a
draft
code
for
our
tree
yeah
I'm,
going
to
work
on
the
receiver
class
so
that,
but
you
can
do
the
actually
same
experiment
having
some
sort
of
finite
length
of
dipole
or
by
pole
for
a
source
and
receiver
and
we're
gonna
implement
that
in
the
frequency
domain
code.
First
and
but
once
that's
done
and
checked
it,
we
can.
We
can
extend
to
time
domain
code,
I'm,
not
sure
other
people
are
interested
or
not.
But
that's
that's
what
I'm
working.
E
E
Cuz,
like
we
extend
the
tensor
code,
so
the
assumption
made
in
the
tensor
like
I
was
was
need
to
be
satisfied,
but
yeah
it'll
be
great
if
you
can
really
generalize
that
for
the
Autry
code,
so
mike
is
kind
of
interacting
with
Joe.
So
hopefully,
Joe
can
help
a
couple
of
like
octree
details,
then
that
that
code
could
be
complete.
F
E
H
We
go
perfect.
E
F
E
Was
like
eh,
yeah,
correct,
but
I
was
thinking
about
a
little
bit
like
a
different
play
for
for
for
SJ.
They
want
an
inductive
source.
So
what
I
was
approaching
it?
It
was
like
planning
to
approach
rather
than
going
that
route.
I
was
actually
thinking
about,
like
writing.
An
analytic
vector
potential
function
for
that,
because
there's
some
benefits
for
that
for
inductive
source.
You
can
make
that
a
divergence
free
magnetic
field,
at
least
for
an
initial
condition,
so
that
that
was
a
route
that
I
was
thinking
but
yeah.
We
can
use
both.
E
But
the
like
that,
if
you
think
about
the
reciprocity,
because
it's
like
once,
you
got
a
source
function
that
suggests
a
transpose.
So
once
you
got
a
source
like
if
you
can
generate
the
integrated
source,
you're
good,
to
go,
it's
a
it's
a
like
it's
like
an
adjoint
function,
so
yeah
you
can
just
transpose
it.
So
once
you've
got
a
source
function,
you're
good
to
go!
Okay!
E
G
Right
now,
I
haven't
gotten
them
to
the
forward
modeling
and
stuff.
Yet
for
the
greedy
amma
tree,
yeah
kind
of
had
some
other
side
project
here,
some
R&D
stuff
to
take
care
of
I.
Guess
one
thing
that
we're
looking
to
maybe
do
say
for
Mt
the
way
that
the
code
forward
models
right
now.
Does
it
just
assume
a
state
of
one
location
for
the
electric
field
measurement,
or
does
it
actually
take
in
a
dipole?
G
G
F
That
that
can
turn
out
to
be
pretty
important
when
you're
doing
kind
of
near
surface
types
of
studies-
and
you
know
if
you
got
lateral,
heterogeneity
and
you've
got
you
know
one
electrode
for
a
potential
on
one
side
of
the
of
the
contact
and
another
electrode
on
the
other
side
can
throw
you
up,
make
things
a
bit
difficult,
so
yeah
you
actually
need
to
be
able
to
get
it
great
along
that
whole
path.
Okay,.
G
F
So
that
is
definitely
on
the
drawing
books.
It
has
been
fur
for
some
time,
but
it's
it's
a
little
bit
second-order
at
this
point,
but
maybe
for
some
of
the
things
that
soggy
has
just
been
doing
with
respect
to
you
know
again
integrating
the
long
wire
path
or
line
both
in
the
time
and
the
frequency
domain.
A
lot
of
those
should
be
completely
transferable.
Oh
cool.
I
G
Yeah,
it's
all
running
now,
again,
I
was
kind
of
the
first
one
was
I
was
using
just
a
just.
A
bad
data
set
I
didn't
have
a
good
discretization
and
as
soon
as
I
put
in
the
data
set,
that
I
have
a
really
good
DM,
yeah
everything
just
started
working
and
then
after
compiling
as
well.
Everything
went
really
well
so
yeah
we're
all
good.
E
For
that,
like
a
mesh
filters
and
refine
XY
z--
s
function,
but
that's
actually
great
and
do
you
have
any
plans
for
extending
to
a
country?
Well,
it's
not
matter
what's
like.
Is
that
an
extend
the
or
not
like
that?
It's.
I
E
I
E
I
E
E
I
E
E
C
E
E
Actually
it
like
at
that
global
domain
is
actually
saying,
and
then
you
kind
of
like
a
defiantly
to
you,
because
you
know
the
older
parents
and
child
cell,
so
you
kind
of
it's
it's
actually
the
same
mashin,
then
you
do
some
sort
of
like
a
very
efficient
interpolation
set.
Is
that
what
you
do
is
that
same.
I
We'll
need
the
I
had
to
do
a
little
work
around,
but
yeah
it's
the
same,
but
it's
not
as
efficient
as
as
it
should
be.
We
need
jool
to
give
us
access
to
the
parents
from
from
the
outside
outside
of
surplus
class,
but
it's
the
same
yes,
so
this
tells
are
all
collocated
up
and
we're
just
looking
for
overlaps
between
between
the
local
and
global
okay.
But
is
there
an
update
today
because
we
were
missing
some
of
the
overriding
cells?
E
E
I
I
C
C
A
I
C
E
I
D
Didn't
do
tons
but
just
went
through
and
started
closing
a
few
issues,
so
some
stuffs
that
had
just
been
open
for
a
really
long
time
that
we
hadn't
touched
or
that
had
changed,
and
so
the
criteria
I
used
just
for
those
handful
were,
it
was
completely
out
of
date,
wasn't
necessarily
fleshed
out
enough.
It
was
just
like
they
were
quite
a
few.
There
were
a
1
statement
of.
D
We
should
use
one
thing
which
at
this
point,
if
it's
3
years
old-
and
we
should
do
it-
we
can
close
it
and
like
where
you
think,
rethink
things.
So,
if
there's
an
issue
that
I
closed,
that
you
were
on,
that,
you
feel
is
still
important
like
please
reopen
it
plum.
My
intention
is
not
to
like
well
take
take
down
items
because
they're
not
important,
but
if
there
hasn't
been
noise
on
it
for
a
while,
then
yeah
I
closed
it
cuz,
I
figure.
The
conversation
has
moved
on
at
this
point.
I
D
They
are
just
a
few
I
mean
if
there's
others
like
it
would
be
great
if
people,
if
you'd
be
willing
to
go
back
and
look
at
issues
that
you
created,
like
I,
just
tried
starting
on
the
last
page,
and
you
know
if
people
could
today
even
just
spend
15
minutes
looking
at
the
issue
list
and
if
there's
something
that
you
posted,
that
is
out
of
date
or
no
longer
relevant
or
is
a
one-sentence.
We
should
do
a
thing.
D
Only
other
things
we
opened
a
couple
issues
just
based
on
looking
soggies
work
prepping
for
next
week,
and
so,
if
there's
issues
like
we,
we
brought
up
the
one
about
the
terminology
around
cooling
rate,
there's
a
couple
other
things
with
directives:
I
mean
the
fact
that
we
don't
print
out
the
last
iteration
I.
Don't
actually
think
we
made
an
issue
on
this
when
you're
running
an
inversion,
we
don't
actually
print
that,
like
the
last
iteration
or
why
the
inversion
stopped.
D
Like
I
mean
if
we've
reached
the
target
misfit,
it
would
be
great
if,
at
the
end
of
the
inversion
it
says
doctor
because
we
reach
the
target.
Misfit
just
gives
the
user
some
really
clear
and
efficient
ones
like
why
we
stopped
right
would
be
helpful.
So
that's
not
actually
an
issue
yet,
but
we
can
document
a
handful
of
those
types
of
things
and
soggy.
D
E
F
The
other
thing
kind
of
a
longer-term
basis
and
I
think
this.
For
me,
this
is
useful
for
any
of
the
invert,
and
that
is
to
break.
Oh
all
of
the
you
know,
the
values
of
the
components
of
the
different
pieces,
of
the
objective
functions
and
all
of
the
three
multipliers.
Just
that
you
can
see
how
big
everything
is
when
we
finished.
F
D
I
D
And
we
really
don't
want
to
be
writing
asking
files
like
that's,
because
everyone
creates
their
own
standards
for
an
ASCII
file,
which
is
really
not
helpful.
If
we
can
do
something
like
with
properties
and
then,
if
you
want
some
beautifully
printed
version
then
like
we
should
be
able
to
translate
between
them.
But
the
thing
that
we
should
the
thing
that
we
save
should
not
be
enough.
D
Some
ways
I
think
like,
so
that
it
would
be
nice
if
we
took
another
look
at
the
iteration
printers
Roland
put
a
lot
of
thought
into
that
originally
and
there's
some
cool
ideas
in
there.
But
it
would
be
great
if,
like
you,
could
just
basically
when
you're
running
an
optimization
or
when
we're
getting
ready
to
run
the
inversion
provide
keywords
of
just
like
what
are
the
things
that
you
want
printed.
So
if
you
want
every
component
of
the
objective
function
or
of
the
regularization
printed
independently
like
we
should
be
able
to
do
that.
D
J
D
I
E
C
C
I
C
F
F
Your
regarding
potential
feels
that
we
have
everything
kind
of
at
hand
for
the
course
I
think
we're
pretty
good
there,
but
you've
been
still
filling
in
some
of
the
figures,
and
we've
got
our
advice:
3d
conversion
for
the
for
the
magnetics
I
ain't
going
to
be
just
nice
to
kind
of
go
through
the
various
elements
of
the
course
and
just
see
where
Dawn's
you
know
where
the
Contra
fields
is
really
coming
in
and
where
we
actually
have
got
all
of
the
results
through
civil
day.
Oh
yeah,.