►
From YouTube: SimPEG meeting July 22
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
Perfect,
so
who
would
like
to
kick
us
off
this
week?.
B
I
got
short
because
I
was
on
vacation
there
yeah,
I'm
so
yes,
kind
of
still
fighting
with
the
solver
there
there's
lots
of
packages
that
you
can
download
that
are
pre-built
and
whatnot,
but
they
don't.
They
all
seem
to
not
have
everything
that
I
I
want.
C
B
Ideally,
when
you
read
the
documentation
they're
like
well,
if
you
want
full
availability,
you've
got
to
use
both.
You
need
like
a
certain
scala
pack
package
yeah,
so
I'm
just
trying
to
optimize
it
as
I'm
building
it,
because
it's
actually
not
hard
to
build.
You
just
have
to
put
all
the
pieces
together
and
then
compile
so.
B
The
best
one
for,
like
the
purposes
of
our
computers,
like
150
cores
how
we
can
actually
use
all
that
so
yeah,
I'm
stuck
testing
mostly
that
right
now
and
then
I'm
actually
playing
around
with
x-ray
again
with
the
fields
for
the
mt
stuff.
I
just
got
it
so
that
I'm
just
I'm
trying
I'm
gonna,
try
out
storing
the
fields
to
using
das
and
then
just
access
them
when
needed.
I
don't
know
if
we're
really
going
to
get
any
speed
up,
but.
B
And
because
I
like
xray,
I'd
like
to
try
and
get
it
a
little
bit
more
used,
but
yeah
just
got
to
find
where
it
will
perfectly
fit.
A
That's
really
cool
I'd,
be
super
keen
to
take
a
look
at
what
you're
up
to
there
yeah
is
it?
Are
you
working
on
like
a
simulation
mt
branch.
B
B
Yeah,
I'm
kind
of
emulating
kind
of
what
you've
done
with
the
field
class,
as
is
just
because
you
can
have
the
pegs
and
stuff
with
x-ray,
but
yeah.
That's
why
I
was
wondering,
if
that's
really
now
that
I'm
getting
into
it
it's
like.
Is
it
purposeful,
because
we
already
have
the
fields
project
or
the
fields
class?
And
it
really
does
all
that
already
so.
C
D
So
it
with
the
fields
like
it,
I
think
it
does
deserve
a
re-look
or
look
over
just
because
they
it
the
access
for
them
is
generally
pretty
slow
right
now,
if
you're
not
accessing
all
of
them
at
once
or
grabbing
like
everything
at
once
from
a
specific
source.
Otherwise,
it's
like
really
it's
actually
pretty
slow.
C
Right
even
mt,
I
think,
if
you're
handling
like
regular
birch
mesh,
so
if
you
could
do
a
lazy
loading
stuff
like
with
x-ray
support,
I
guess
that
that'll
be
quite
nice.
Yeah.
D
A
Other
than
that
yeah.
A
Thanks,
john
dear.
E
Yes,
yeah,
I'm
I
mean
the
same
trying
to
get
stuff
that
we
tried
in
the
last
months
to
master
and
documenting
and
testing
is
a
slow
process,
but
it's
mostly
rewarding
because
you
find
all
the
little
bugs
because
you
did
it,
you
implemented
it
something
quickly
and
then
you
want
to
do
it
nicely,
and
you
think
why
did
I
do
that
so
yeah,
it's
a
nice!
It's
a
nice
job!
E
E
Also
it's
it's
quite
new
for
me
that
it's
not
yet
inversion,
but
it's
going
to
the
immersion
side
and
like
the
first
year
or
two
of
this
3d
project,
that's
mostly
involved
with
the
solver
so
really
on
on
a
much
more,
let's
say,
fundamental
basis:
how
to
generate
the
field.
So
now
the
inversion
is
kind
of
it's
also
a
solver
issue,
but
it's
a
completely
different
story.
A
E
Two
things
so
the
solver
that
it
first
multigrid
solver
itself
is
just
yeah.
It
has
a
verbosity
parameter
and
it
prints
it
and
that's
now
hidden
away.
But
then
the
stuff
I
do
for
the
command
line
is
just
with
the
login,
the
inbuilt
logging
package,
and
then
I
print
it
to
a
file
and-
and
I
print
it
to
screen
and
they
can
choose.
A
E
And
also
the
the
x
array,
because
you
mentioned
it
at
the
moment-
it
just
makes
me
easy
because
you,
if,
once
you
have
a
data
set,
then
it's
easy
to
just
slap
another
date
array
on
it.
So
if
you
have
one
data
set,
you
just
slab
as
many
on
as
it
and
then
each
dataset
knows
then
which
source
which
receiver,
which
frequencies
is.
So
you
don't
have
to
take
care
of
that.
E
So
in
this
I
didn't
find
yet
other
quite
usage
other
than
a
very
easy
way
to
store
and
access
data,
but
yeah
for
that.
It's
really
handy
also.
If,
if
I,
if
I
suddenly
want
to
have
like
the
fields
that
the
response
of
every
step,
then
you
just
add
to
this
data
set
a
new
data
array
for
each
step,
which
is
kind
of
easy.
C
What
I
wasn't
sure
like
it's
if
you've
got
very
uniform
data,
set
it's
kind
of
making
sense
like
using
like
something
like
an
x-ray,
but
sometimes
data.
It's
not
uniform
like
it.
It's
it's
hard
to
put
it
into
just
a
single
table
yeah.
So
it
has
a
little
bit
more
complicated
structure
in
general
for
the
problem
that
we're
dealing
with
so
yeah.
I
wasn't
sure
I
was.
I
thought,
like
a
field
object
is
a
little
bit
more
sort
of
aligned,
yeah.
E
C
Yeah,
but
like
even
that,
stick
even
that
for
that
case,
like
like
you,
don't
have
necessarily
in
the
reality
like
your
receiver,
will
have
some
missing
point
in
certain
frequency.
So,
even
though
you
got
relative
uniform
set
up,
but
things
can
go
fairly,
complicated.
E
If
it's
just
a
few
but
then
like
for
the
ip
or
dc
you
do
where
you
have
some
sources
with
only
one
receiver
or
vice
versa,
then
you
get
a
lot
of
nons.
Then
it
might
not
be
that
cool
anymore.
If
it's
just
a
view
of
data,
then
it
shouldn't
matter
too
much
yeah,
so
yeah.
I
do
think
in
simpek
it's
much
more
difficult
because
you
are
so
so
flexible
with
your
survey,
layouts
and
methodology
and
all
that
stuff.
So
it
would
require
a
lot
of
thinking.
I
think.
C
Nothing
particular
I
spit
busy.
I
think
I'm
gonna
be
busy
until
end
of
july,
so
yeah
I'm
just
working
on
geostatistics,
although
it's
actually
kind
of
fun,
I'm
mostly
using
the
program
called
sgm.
I'm
not
sure
you
guys
have
heard
about
that.
It's
actually
open
source
code.
C
An
interesting
and
yeah,
so
what
I'm
mostly
working
on
is
like
kind
of
combining
that
geophysics
into
geostatistics.
So
that's
what
kind
of
seems
like
people
do
so
they
use
the
output
of
geophysics
and
then
fit
into
their
geostatistical
algorithm
to
get
much
higher
resolution
picture
rather
than
a
smooth
image
that
we
provide.
A
C
That
geological
patterns
to
generate
much
higher
resolution
images,
it
seems
kind
of
fun,
but
nothing
really
related
to
simpac.
But
one
good
news
is
like
a
probably
interesting
point
is
if
you
can
have
a
simple
geostatistical
algorithm
that
we
can
generate
those
fine
resolution
image
and
use
like
a
synthetic
like
a
kind
of
or
geophysical
simulation
to
generate
their
the
geophysical
response.
A
A
C
E
C
C
I
think
he's
he's
he's
like
he
has
much
more
updated
version,
but
yeah
the
older
version
is
open
source.
D
Yeah
sure
so
I've
been,
I
got
a
bunch
of
those
pull
requests
on
into
simpeg.
All
those
little
bug
fixes
that
people
brought
up
got
a
few
new
contributors
on
there,
which
is
kind
of
nice.
D
You
know
every
small
bit
little
helps
so
it's
always
good.
I
figured
out
why
the
dot
tests
were
failing,
because,
essentially
what
happened
was
the
files
that
we
had
used
that
we
uploaded
to
our
bucket
google
bucket
whatever
were
protected,
and
the
only
and
only
branches
that
the
sim
peg,
that
was
under
the
simpeg
repository
could
access
the
necessary
credentials
to
download
the
files,
so
they
just
weren't
getting
downloaded
for
pull
requests
that
were
originating
from
outside
the
simpeg
organization
repo.
D
So
I
just
switched
around
to
make
it
publicly
so
now
those
should
they
should
not
fail
when
other
people
do
pr's
on
us,
which
is
nice,
I'm
not
sure
if
you
guys
have
noticed,
but
I've
been
trying
to
at
least
follow
that
practice
myself
as
far
as
originating
pull
requests,
so
I
fork
it.
I
fork
sent
back
to
my
repository
and
then
branch
it
off
from
there
and
then
do
a
pull
request
onto
the
simpegmaster
from
my
own
version
on
here
from
my
own
branch
on
my
own
repo.
D
D
D
D
So
if
you
for
you
can
fork
it,
I
try
to
keep
my
own
master
branch
on
my
personal
up
to
date
with
the
syntax
master.
So
the
process
is,
but
you
don't
necessarily
have
to
do
that.
You
just
have
to
fork
it
to
your
own
repository
and
then
you
can
literally
branch
off
from
the
syntax
master
and
then
push
it
to
your
own
copy
of
the
repository.
D
Yeah
seems
to
work
out
pretty
well
what
else?
Oh
I've
been
working
on,
getting
everything
ready
for
the
eoas
350
course
as
part
of
that
I've
gotten
all
the
geosci
labs
notebooks
repository
the
gsi
lab
notebooks
updated
for
the
new
simulation
branch,
as
well
as
a
few
little
operations
to
make
them
run
a
little
smoother
like
caching,
some
of
the
things
that
don't
need
to
be
rerun
if
all
you're
doing
is
like.
Oh,
I
want
to
look
at
the
electric
field.
D
Now
I
want
to
look
at
the
you
know
the
currents
or
the
potential
like
there's
no
need
to
rerun
the
entire
simulation,
so
it
just
makes
a
little
bit
more
responsive
like
that
yeah
and
then
I've
also
updated
all
the
gpg
labs
to
to
work
off
the
simulation
branch
as
well.
D
D
A
D
So
like,
if
you
look
at,
I
think
scipy
does
it
some
if,
with
one
of
their
with
one
of
the
documentation
things
there's
like
a
or
like
a
hook
there,
it's
kind
of
it's
neat
made
it
just
links
to
other
repository.
D
D
No
michael
bostock
is.
C
D
A
D
Want
to
get
it
up
on
their
quest,
have
a
jupiter
hub
hosted
so
for
the
students
to
just
log
into
that
to
have
so
the
environment's
all
ready
to
go
for
them
got
it
yeah.
D
A
I
went
through
the
zero
to
jupiter
hub
guides,
and
so
I
have
launched
a
jupiter
hub
on
kubrick
or
on
google
yeah,
but
I
haven't
really
ever
like
maintained.
One
are:
are
they
not
going
to
use
the
sysg
one?
Are
they
launching.
D
D
Yeah,
so
they
are
using
like
they're,
they're
planning
and
they're
planning
on
using
g.
I
believe
I'm
just
curious
if
there'll
be
a
way
to
use
the
kubernetes
to
get
up
on
system.
If
that
would
be
easier.
A
If
you
have
questions
on
that,
ian
allison
is
a
good
person
to
to
connect
with
he's
the
one
who
like
really
actually
runs
at
ubc.
D
That
sounds
good.
I
just
and
then,
during
the
semester,
I'm
going
to
be
helping
out
make
sure
things
keep
running
during
the
course
if
they
have
any
problems,
they'll
be
on
top
of
that
just
a
little
bit
of
extra
stuff
to
do,
hopefully,
a
little
experience,
helping
them
and
probably
have
a
little
bit
more
exploration
geophysics
than
the
other
tas
that
they
have
helping
so
I'll,
be
hopefully
a
good
thing.
I
hope
they
can
take
advantage
of
that.
D
Oh
other
than
that,
I
I
don't
think
I
have
anything
else.
Oh
the
volume
average
stuff
dom
tested
that
out.
Looking.
D
Think
it's
worth
it
to
put
in
a
special
hook
for
when
that,
like,
when
a
tensor
and
an
octree
share
the
same
bass,
mesh
or
two
type,
the
two
octres
share
the
same
bass
mesh
because
it
there
is
a
much
faster
way
to
do
it
in
those
cases.
But
what
I
did
does
work
in
general.
So
that's
good
to
test.
It's
good
to
know.
F
Yeah-
and
I
can
you
know,
I
can
just
keep
relying
on
on
what's
in
the
map
mapping
for
now,
because
it's
like
10
times
a
time,
you
know
sorry
10x
is
kind
of
it's
not
negligible
for
large
large
meshes
and
then
once
you
have
it,
we
can
just
swap
over,
but
basically
they're
doing
the
same
operation.
So
that's
that's
a
good
thing.
F
F
The
new
implementation-
yes,
because
it's
more
general
right
so
I'm
sure
I'm
sure
joe
you're
doing
some
extra
special
cases
right
to
be
able
to.
I
guess
a
very
general
approach,
rather
than
what's
in
the
tile
map
is
always
assuming
that
the
the
cells
are
nested,
so
you
can
shortcut
basically
a
lot
of
operations.
I'm
sure
that
well
yeah.
D
And
those
same
like
it
should
work
as
long
like
we
can
do
a
lot
of
shortcuts.
If,
like
I
said,
if
the
octrees
and
tensors
share
the
same
base,
then
there's
a
lot
of
shortcuts,
you
don't
have
to
do
it
takes
the
most
time
is
just
like
searching,
but
that
doesn't
that
won't
take
nearly
as
much
time
if
we're
just
looking
at
cell
centers
and
like
finding
either
thing,
because
what
it
does
right
now
is.
F
Whereas
the
the
tile
map-
because
I
know
it's
nested-
I
can
use
joe's
like
in
cell
in
cello
tree
call,
which
you
know
always
returns,
only
the
cells
that
are
inside
so
then
I
don't
have
to
search
right.
I
know
exactly
where
cells
are
inside
and
just
do
a
calculation
like
vectorial,
basically.
D
E
F
C
And
that's
that's
great,
like
or
joe
have
you
like,
tested
with
the
actual
simulation,
so
it's
it's
great
that
you
got
this
averaging
algorithm
then
have
you
tested
with
like
some
simulation?
Let's,
you
generate
some
coarse
mesh,
but
you
gotta
find
mesh
that
you
have
fine
scale
structure.
Have
you
like
tested
some
of
those
like
kind
of
things
like
because
it'll
be
different
depending
upon
what
simulation
you
would
use,
you
probably
need
to
better
use,
either
conductivity
or
resistivity
or
like
yeah
it.
C
D
Yeah,
when
I
was
I've
used
it
before
I'm
like
mt
and
like
ip
and
dc
as
before
the
volume
averaging
operation,
I
was
having
better
look
volume
averaging
log
conductivities
instead
of
conductivities
generally,
but
it
seemed
to
work
out
pretty
well.
D
F
Yeah
and
I've
only
tested
with
the
integral
integral
equation.
So
as
long
as
the
mass
is
preserved
and
then
you
know
far
field
enough,
then
you
don't
see
you
don't
see
the
difference
right,
but
I'm
sure
with
pde
it's
much
more
critical.
How
you?
How
you
do
it.
A
Thanks
joe
dom,
do
you
have
any
updates.
F
No
well,
I
did
I
did
that
pr
review
for
joe
and
then
I'm
still
elbow
deep
in
testing
the
m1d
code
for
more
systems.
So
you
know
I
looked
at
the
helitam
geotown.
I
did
more
v10,
so
yeah,
it's
just
more.
When
the
inversions,
I
actually
got
managed
to
get
the
harus
connectivity
model
like
their
stitch
when
the
inversion.
So
I
could
I
benchmarked
against
against
our
code
and
the
models
were
pretty
similar.
So
that's
that's
really
nice.
F
That
means
we're
at
least
you
know
for
like
straight
up
like
l2
or
really
close
to
what
the
what
the
danish
people
are
doing.
So
that's
that's
great
good
check
mark
for
us.
F
That's
and
then
nothing,
nothing
really
related
after
this.
A
That's
fair,
I
guess
the
the
other
thing,
though
we
are
on
the
hook
now
for
sdg,
so
it
looks
like
that'll
we'll
have
to
figure
something.
Do
you
know
when
we
need
to
have
those
recorded
by.
F
I
will
check
my
emails
and
then
forward
it
to
you.
I
can
remember
right
now.
D
F
The
the
low
moment
was
only
like
two
extra
early
time
so
like
it
didn't
it
wouldn't
change
much.
I'm
guessing
and
the
result
looks
very
similar,
but
yeah
we
need,
we
should
talk,
should
can
we
actually,
because
I
saw
that
you
hard
coded
that
right
to
the
yeah
like
a
dual
moment,
yeah
system?
Could
we
just
treat
it
as
a
joint
as
a
joint
misfit
problem.
C
F
Because
I
I
kind
of
diverge
from
the
script
that
we
were
working
on
and
right
now,
I'm
kind
of-
and
you
know
the
way
they
input
files
they're
the
one
dealing
with
the
inputs
and
everything
and
it's
two
different
waveforms,
I'm
just
wondering
if
we
could
just
not
do
like
like
a
joint
problem
and
then
that
we
would
be
able
to
do
like
both
the
time
and
frequency
system
at
the
same
time.
You
know
for
that
kind
of
that
kind
of
idea.
C
Yeah,
but
I
think
I
really
like
it,
but
like
the
computation
wise,
like
it
only
actually
like
for
computing,
both
high
and
low
moment.
It
only
requires
a
single
single
computation
of
like
frequency
response,
and
then
you
just
need
to
do
a
convolution
twice,
because,
okay,
like
it's
two
different
waveforms,
so
yeah.
C
Handle
two
different
system:
sure
you
could.
You
could
set
up
as
a
joint
joint
misfit
functions
right.
F
C
G
Not
really
I
mean
I,
I
went
and
tried
to
fix
the
import
issues
that
dieter
was
having
so
that
he
could
contribute
to
the
em1d
stuff,
but
I
haven't
really
touched
too
much
simpag
recently,
it's
been
all
gwb
and
dc
resistivity
stuff
for
jif
tools.
E
G
Well,
I
mean,
I
guess.
The
first
thing
is
just
to
see
if
you
can
import
like
just
run
the
script
right,
so
it
looked
like
the
init
dot.
Pi
files
were
just
not
not
good,
so
I
kind
of
clean
that
up
a
little
bit
and
yeah
yeah.
Okay,
so
I
mean
that's
at
least
the
first
step.
So
if
you
can,
if
you
can
run
any
of
those
forward
modeling
scripts,
then
that
kind
of
says:
okay,
you
forgot,
you
fixed
the
problem.
G
I
tried
running
it
in
spider
and
I
tried
running
it
just
from
a
command
line
and
I
didn't
get
any
issues,
but
I
didn't
get
any
issues
before
so.
E
G
C
Cool,
hey,
theater,
related
to
that
in
em
pie
mode.
Can
I
set
a
arbitrary
shapes
source
like
a
like,
not
just
a
rectangular
source,
but
that
if
you
got,
I
don't
know
like
some
sort
of
polygon
type
like
source?
Can
I
set
that.
D
E
C
So
if
you
got
a
like
arbitrary
shape
source
and
you
compute
bunch
of
dipole,
but
still
somehow
you
need
to
integrate
them
over.
Yes,
you
have
that
function.
E
I
I
have
them,
for
you
can
have
finite
length,
dipoles
right
and
I
have
there
the
gaussian
quadrature
of
low
a
few
points.
Who
does
that
integration,
but
yeah?
If
you
want
to
do
it
for
an
arbitrary
shaped
source,
you
would
have
to
do
it
yourself,
but
you
could
use
the
tool
simple
and
I
think
there's
just
no.
A
So
I
guess
the
updates
I
have
I've
been
playing
around
well
again,
sort
of
with
the
the
modeling
paper
of
theater
and
came
up
with
a
super
hacky
higher
higher
order,
interpolation,
and
so
what
I
ended
up
doing
is
basically
just
grab
like
a
chunk
of
cells
around
the
receiver
and
use
the
sci-fi
radial
basis
function
to
interpolate,
and
it
actually
makes
I
mean
if
you
really
do
want
to
drive
down
errors,
it
does
make.
It
does
make
a
big
difference.
A
The
other
thing
that's
kind
of
been
interesting
to
get
a
sense
of
is
when
we
really
are
trying
to
drive
down
errors
is
how
sensitive
we
are
to
padding,
so
I'm
using
the
autry
meshes
and
then
the
planting
cells
to
try
and
get
the
boundary
conditions
satisfied
without
you
know
introducing
too
many
more
cells,
but
we're
you're
pretty
sensitive
to
that.
So,
if
you
really
do
want
to
drive
down
the
error
it
takes,
it
takes
some
thought
to
put
that
in
place
so
kind
of
interesting
kind.
A
It's
all
the
natural
boundary
conditions,
so
yeah
yeah,
so
that's
actually
something
that
would
be
fun
to
play
around
with
and
see
see
what
we
can
do.
A
Yeah,
actually
so
maybe
that's
that's
one
thing
I
can
try
to
is
switch
formulations
and
see
see
what
kind
of
difference
that
means
the
other
thing
so
agu
abstracts
are
due
next
week.
A
So
I'm
going
to
put,
I
would
like
to
put
one
in
on
mt
and
sort
of
in
connection
with
the
jupiter
meets
the
earth
project
so
sort
of
focusing
in
on
what
are
kind
of
interactive
visualizations,
and
things
like
that
that
we
can
be
doing
well
and
inversion
to
running.
But
we'd
like
to
you
know,
come
up
with
some
good
empty
examples.
So
anyone
who
wants
to
participate
in
that
abstract,
like
feel
free
to
add
your
name
and
add,
adjust
the
abstract
to
make
suggestions
or
anything
like
that.
A
I'm
happy
to
have
a
big
list
of
co-authors
so
feel
free
to
take
a
look.
If
there's
anyone
else
who
wants
to
submit
an
abstract
on
anything,
simpeg
related
feel
free
to
just
drop
drop,
an
abstract
or
a
sketch
of
an
abstract
in
this
google
doc
and
ping
people,
and
we
can
all
kind
of
jump
in
and
help
out.
It
would
be
nice
to
have
a
few
different
examples.
A
I
know
we
talked
about
a
bunch
for
the
gwp
project,
so
if
anyone
else
has
ideas
go
for
it
yeah,
then
I
guess
the
last
thing
is
following
up
on
dieter's
comment
on
slack-
and
we
talked
about
this
briefly
last
week
with
sort
of
bug,
bug
fixing,
sprints
and
or
time
to
sort
of
talk
about
issues
or
pull
requests.
That,
like
just
need
a
brief
conversation,
do
you
want
to
share
what
you
suggested
with
with
everyone.
E
Yeah
well,
I
just
thought,
because
I
sort
of
accepted
last
time
to
find
a
time
slot
and
I
thought
well,
it's
a
bit
difficult.
I
tried
to.
I
actually
wanted
to
make
a
doodle,
and
then
I
got
so
confused
because
of
this
the
time
difference
that
everything
that
would
fit
me
is
somewhere
early
in
the
morning.
It
doesn't
make
sense
that
I
do
it.
E
I
just
thought:
well
there.
I
think
there
are
two
possibilities.
One
is
very
short
but
intense,
where
it's
more
discussion
and
distribution
and
the
other
one
and
probably
more
regularly
and
the
other
one
would
be
more
in
less
frequent
but
longer
and
actual
hacking,
or
I
think
you
said
we
could
even
do
both
combined.
E
I
don't
know
so
I
guess
we
would
have
to
find
that
a
time
that
fits
like
say
one
hour
or
two
hour
for
the
longer
one
or,
as
I
said,
for
the
shorter
one
just
going
through,
I
don't
know,
starting
with
the
oldest
one
going
through.
We
could
do
it
half
an
hour
after
the
meetings
or
I
think,
as
as
soon
as
it's
a
routine.
We
would
actually
do
it
and
if
it's
not,
we
don't
yeah.
E
D
E
A
A
D
B
F
A
Well,
it
keeps
ourselves
busy
right,
yeah,
okay,
that
sounds
good.
When
do
we
want
to
schedule
sort
of
the
first
well,
we
can
maybe
does
next
week
work
for
everyone
for
a
first
slightly
longer
simpeg
meeting
and
we'll
just
start
tackling
well,
I
guess
maybe
joe,
that
won't
work,
because
you
might
be
on
the
road.
F
D
A
Okay,
that
sounds
good,
so
in
two
weeks
time.
So
maybe
what
we
can
try
for
next
week
is
even
if
folks
just
want
to
take
a
quick
look
at
the
issues
just
so
that
you
sort
of
see
what
we're
what
we're
up
against
and
if
there's
any
that
you
started
you
know
three
years
ago
and
just
want
to
close,
because
they
look
like
we've
addressed
them
or
things
like
that.
Go
for
it,
cool!
Okay!
D
Let's
start
with
the
two
week,
one
first
and
see
what
problems
we
identify
at
that
one
and
then
we'll
know:
okay
well,
maybe
next
weekend
or
something
like
do
them
every
other
week
and
this
week
it's
on
that,
it's
a
30
minute
after
the
next
week
or
two
weeks
later.
It's
you
know
the
friday
session
or
something.
A
Excellent,
okay,
any
other
things
that
folks
would
like
to
bring
out
before
we
wrap
things
up
for
the
day.
B
E
Although
we
we
had,
we
had
a
meeting
when
was
that
two
weeks
ago,
three
weeks
ago,
but
then
I
realized
the
first
email
I
sent
a
year
ago
and
I
think
we're
having
four
different
projects
involved.
It's
not
too
bad
a
year's
time,
so
we
said
end
of
august.
We
would
like
to
submit
it
so,
but
it's
it's!
The
repo
is
online.
Now
it's
not
oh
yeah.
B
Yeah
I
asked
because
yeah,
if
I
can
get
this
mumps
version
built
like
the
optimal
one,
I'd,
maybe
like
to
sneak
that
in
there.
It
sounds
like
it's
an
interesting
problem
to
see,
and
you
guys
have
so
much
other
comparison
to
do
or
other
comparisons
to
make
with.
So,
if
I
can
slip
that
in
there
before
you
guys
are
done,
you
know
it'd
be
kind
of
cool.
G
And
then
you
had
that
there
was
that
paper,
you're
thinking
of
doing
lindsey
on
the
mt
stuff
and
reproducing
things,
and
I
guess
I
see
the
biggest
hurdle
being
getting
getting
the
raw
data
into
the
simpeg
framework.
What's
what
I
guess,
what
file
format
do
you
want
to
start
with.
A
Is
this
so
this
is
the
model
you
were
thinking
of.
We
were
trying
to
go
on
slack.
E
G
Well,
I
mean
the
whole
the
whole
project
and
all
the
steps
are
actually
documented,
so
you
could
start
off
with
raw
edi
files
that
have
been
formatted
or
they
there
is
the
data
files
that
are,
according
to
the
e3dmt
version,
two
code,
okay
and
so
that's
actually
kind
of
nice.
It
gives
you
a
separate,
a
separate
file
where
you've
defined
all
of
the
receivers
and
then
there's
a
file
that
actually
just
gives
you
an
index
like
it
indexes
all
of
the
receivers
and
provides
you
the
associated
data.
With
that.
A
Okay,
send
the
link
just
to
that
page.
That
would
be
great
because
I
think
ideally,
we
should
actually
be
able
to
read
both
of
those
things
into
simpeg,
so
both
getting
a
raw
edi
file,
which
I
think
you
did
a
bunch
of
work
on
a
while
back.
G
Yeah,
I
would,
I
would
almost
suggest
just
using
the
the
jiff
ones,
because
the
raw
edi
files
has
the
whole
suite
of
frequencies
right
and
we're
we're
just
going
to
pick
a
small
amount
of
those
and
invert
them
and
recreate
the
results.
So
maybe
it's
actually
best
to
start
with
the
version
2
ubc,
but
it's
it's
all
in
there.
So
I
can
just.
I
can
just
send
that
link.
A
That
would
be
great
yeah
and
so
and
for
now
that's
just
in
an
egu
abstract,
but
hopefully
it
is
actually
something
that
we
could
turn
into
into
a
paper.
So.
F
G
There
was
there
was
a
demand
to
create
a
sensitivity,
weighting
function
for
the
dcip
autry
code
and
to
implement
a
line
where
you
could
put
in
separate
cell
weights
that
only
act
on
the
smallness
term
and
we're
figuring
that
out,
because
you
can't
you
can't
explicitly
compute
the
sensitivities.
You
have
to
do
some
hutchinson's
method
and
so
we're
fiddling
around
with
a
way
of
kind
of
creating
a
sensitivity,
weights
model
and
we're
making
some
progress.
G
So
you
did
your
mount
isa
results,
I'm
going
through
and
applying
all
the
new
stuff
and
testing
all
the
the
tools
that
we've
developed
and
then,
in
the
end,
we're
going
to
have
a
standard
workflow
for
dcip
inversion
and
when
we
get
to
the
end
of
that.
G
G
So
we'd
have
this
yeah
we'd
have
a
comprehensive
workflow
we'd
have
new
tools
where
we
have
sensitivity
waiting
for
the
the
3d
like
the
dcip
octree
code,
and
then
that
could
be
reproduced
with
simpeg
and
it
seems
like
there's,
there's
a
big
desire
to
have
simpeg
and
ubc
gift
codes
matching
and
compare
their
performance
and
compare
the
things
that
we're
getting
out
and
if
we
can
do
that,
it's
it's
that's
pretty.
That
would
be
pretty
great,
so
yeah.
G
Once
once
roman,
once
we
figured
out
the
that
synthetic
sensitivity,
weights
situation,
then
it's
getting
thrown
in
the
gui
and
then
there'll
be
a
new
release
and
then
hopefully
I
won't
have
to
develop
that
again.