►
From YouTube: SimPEG Meeting May 14 part 2
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).
A
I
shouldn't
have
yeah,
go
ahead
and
try
again
and
we'll
see
what
happens.
I
can
see
your
screen
now
so.
B
They
were
talking
about
the
tree
mesh,
so
all
the
internals
like
say
all
the
calculation,
heavy
and
memory
heavy
parts
are
numbered,
cheated
functions
right
so,
and
so
there
are
many
loops.
It's
not
numpy
it
it's.
There
are
many
loops.
That's
why
I'm
asking
so
like
in
solving
the,
but
we
say,
aim
at
X,
so
the
left
hand,
side
yeah.
A
B
Basically,
just
looping
over
number
of
elements
in
set
directions:
number
of
elements
in
y
directions,
number
of
element
in
X
directions
right
and
solve
it
for
each
node,
the
problem
for
all
neighboring
cells.
So
if
this
is
can
be
done
easily
on
a
tree
mesh
to
then
it
shouldn't
be
too
difficult
to
implement.
If
that
needs
some
serious
reconsideration,
then
it'll
be
different.
I
guess
the.
A
Loop
would
have
to
end
up
being
different.
We
can't
loop
easily
over
the
range
of
each
oh.
We
would
basically
loop
over
cells
like
instead
of
having
three
nested
loops.
They
would
actually
just
be
a
single,
a
single
loop
right
over
all
of
the
cells
in
the
in
the
mesh,
because
the
way
that
we
walk
them
is
a
bit
different.
Like
you
sort
of
you
walk
along
each
each
dimension,
I
think
that
docks
on
actually
how
the
tree
meshes.
Stewart
Joe's
done
a
nice
job
explaining
some
of
those
things
yeah,
but
it
would.
A
A
B
That's
one
place
the
the
a
times
X
and
the
other
is
in
the
in
the
smooth
thing,
which
is
them
the
turret
of
solver.
If
you
want
that,
that
is
run
at
each
level
and
it
again
it
runs
over
all
the
dimensions
over
set
over
Y
over
X
and
then
inside
you
solve
the
system
again
and
so
yeah.
This
could
probably
done
be
done
with
parts
of
same
pack.
I
don't
know
because
you
have
all
the
curls
and
stuff
included,
so
yeah
be
an
easier
way.
A
B
A
A
B
B
A
C
A
B
A
Ubc
folks,
you're
John
are
there
any
other
questions
or
thoughts?
You
guys
gonna
bring
up.
A
B
I
compared
the
party
so
and
my
mother,
so
it's
not
just
a
difference
over
it's
also.
The
physics
is
implemented
separately
in
both
and
the
error
are
basically
1
to
the
power
of
minus
5
or
less
the
relative
error
or
the
relative
difference
between
the
two.
So
this
is
really
yeah.
It
kind
of
gives
a
lot
of
confidence
that
both
do
the
right
thing.
That's
good!
That's.
A
Excellent
one
of
the
things
that
I
think
would
be
kind
of
cool
that
we
think
about
when
we
write
what
a
makes
sense,
what
would
be
to
write
like
a
leading
edge
article
or
something
like
that
that
just
tells
a
bit
of
the
story
of
some
of
the
overlaps
of
like
iam,
hi,
Maude
and
now
iam
its
iam
3dg.
Sorry,
EMG.
A
B
One
of
the
things
are
who
my
to-do
list
for
the
next
two
three
months
or
something
it's
really
child
to
you
guys
both
put
two
caste
you've
checked
out
that
ones
Castilla
Costigan.
It
was
just
published
in
to
physics
when
my
50
papers
have
to
I
think
a
month
ago.
Okay,
so
I
think
it's
associated
or
at
least
loosely
associated
with
pie
chimney.
B
It's
a
finite
element,
see
SEM
code
yep
using
Phoenix
under
the
hood,
there's
calcium
and
then
there's
pet
Chang
from
Barcelona
mmm-hmm,
which
is
also
finite
element
code,
also
using
Phoenix,
but
they
use
it
differently.
I
think
one
of
them
uses
time
for
PBS
I,
don't
remember
they
kind
of
use
different
hooks
into
Phoenix,
but
both
his
Phoenix
I
I
thought
it
could
be
nice
now.
B
Are
kind
of
different
3d
CSUN
cooks
out
there
we
could
have
like
no
the
competitive,
more
comparable
article
about
what
is
the
strength
of
one
and
what
is
the
strength
of
the
other
and
what
cans
impact
or
what
can
the
guys
in
Barcelona,
so
I'd
like
to
reach
out
to
them
and
see
with
you
guys
to
have
kind
of
this
article
of
dying
thing,
great
that'd
be
great,
because
I
think
yeah
a
lot
changed
in
the
last
five
years.
I
just
know
it's
from
the
CSE
M
word,
but
a
lot
of
change
there.
B
Mean
it
would
say,
man
I,
went
to
transform
and
I'm,
not
ask
us
to
write
down
what
what
is
staring
in
your
daily
work.
Frequent
business
I
never
had
done
one
of
the
things
it's
kind
of
missing.
We
create
our
mothers
and
in
lack
of
realistic
geological
model,
we
just
put
a
cube
into
the
hemisphere
and
we
model
a
cube
right.
It.
C
B
Us
there
it's
kind
of
missing
a
geological
modeling
thing,
and
then
there
were
the
champ
ice
there
that
I
think
I
had
to
go
Clark
but
I
never
looked
into
it,
and
so
then,
just
also
having
this
possibility
to
have
a
really
geological
model
and
then
model
that
forward.
The
Cueto's,
oh
yeah,
I
think
there's
a
lot
of
things
and
we
don't
know
of
many
things
that
actually
exists.
So
yeah.
A
B
A
A
B
A
A
B
A
C
A
B
A
B
B
Would
turn
me
away
if
you're
on
an
inversion
with
simple?
Could
you
define
an
external
model
like
EMP
3d,
but
I
run
an
inversion?
Is
there
an
interface
I
think
we
already
discussing
that
in
Python?
Now
it's
there
kind
of
an
interface
hook,
so
you
could
define
the
forward
model
but
use
the
simple
conversion
things
so.
A
There's
the
the
challenge,
though,
is
derivatives,
so
we
do
need
write
derivatives,
yeah
for
EMP
mode,
it's
okay
to
use
the
perturbation
technique
like
we
can
estimate
with
perturbation
I'm,
not
entirely
sure
how
feasible
that
is
on
the
multigrid,
although
it's
something
that
we
can
try.
Oh
yeah,
so
I
think
the
simulation
refactor
I'm,
hoping
to
have
made
a
bit
more
progress.
That's
sort
of
related
to
that
data
data
issue.
A
If
it
was
in
simple,
that's
all
related,
but
if
we
can
kind
of
get
that
up
and
running
and
we
can
clean
up
first,
maybe
the
EMP
mod
example.
Yes,
how
to
plug
it
in
and
then
we
can
play
with
it
and
see,
see
what
it
does
and
see.
Yeah
mm-hmm
yeah
see
what
we
can
get
out
and
it
might
be
I,
don't
know
if
you've
looked
into
like
any
of
the
automatic
differentiation
stuff
in
Python.
I
know
like
PI
torch
uses
that
and
things
like
that.
C
A
A
Not
rather
it's
not
necessarily
difficult.
It's
just
like
it's
unnatural.
If
that
makes
sense
like
the
this
pairing
that
we
do
between
the
problem
and
the
survey
like
is
just
clunky
to
work
with
when
you're
plugging
in
another
code.
Where
is
it?
We
unite
that
under
just
a
simulation,
then
it's
very
easy
to
write.
A
single
wrapper
on
assume
you
like
on
any
external
simulation
to
basically
give
give
sensible
inputs
to
simpang
so
like
there
is
I.
A
E
E
A
Helmet
is
the
textbook
chapter:
oh
okay,
and
if
you
don't
have
it
I
have
a
PDF
copy
that
I
can
slack
to
you
right
after
this
yeah
sure
yeah
I.
Don't
we'll
have
to
have
a
look
at
like
what
the
solutions
are
in
there,
but
I
think
there's
there's
some
things
that
we
can
certainly
play
around
with
so
okay,
cool
yeah,
I.