►
From YouTube: SimPEG Meeting June 9th
Description
Weekly SimPEG Meeting from June 9th, 2021
A
Oh
hello,
everyone!
It's
nice
to
see
you
guys
all
again
a
good
contingent
here
this
week,
so
I'm
excited
to
get
everybody
back,
get
into
it
again.
If
you'd
like
to
talk
about
anything
as
far
as
agenda
lindsay,
I'm
not
sure
if
you're
ready
to
talk
about
governance,
yet
it
might
be
a
good
time
as
far
as
agenda
items
for
things
today,
if
not,
we've
also
got
some.
A
A
A
C
Yeah,
I
probably
wanted
to
pick
up
what
we
had
the
the
the
discussion
this
week
in
in
general.
I
think
about
the
inversion
that
is
used
and
which
is
used,
and
I
talked
to
zoggy
about
it
and
said:
maybe
it
would
be
cool
if
I
don't
know
lindsey
dom
so
yeah.
I
don't
know
who
are
the
main
contributors
to
the
inversion
once
could
give
like
a
a
session
on?
A
C
D
Well,
that
would
be
great
dieter
I'd
be
happy
to
to
to
work
on
that
and
taki.
If
you're
interested,
we
could
work
on
something
together
or
anybody
else,
and
maybe
we
could
actually
first
get
a
few
questions
from
folks
and
that
can
maybe
focus
what
we
give
an
overview
on.
I
could
so
in
some
ways
actually
imagine
this
going
coming
in
two
parts
is
one
is
kind
of
the
inversion
components
of
simpeg
as
they
are,
and
I
think
that
that
will
rapidly
lead
into
a
discussion
of.
C
D
A
F
F
F
F
So
this
this
was
on
my
to-do
list,
but
that
yeah,
I
think
I
got.
I
got
something
still
analytic,
but
it's
piecewise
analytics.
So
it's
okay,
but
I
I'm
pretty
sure
like
implementing
the
ubc
way,
would
would
actually
be
quite
better
and
by
doing
so
you
can
actually
handle
the
on-time
response
which
is
coming
in
these
days,
especially
for
skytemp
system.
Maybe
couple
other
systems,
so
it
would
be
quite
nice.
I'd
like
to
just.
A
Get
in
and
just
get
it
like
that,
so
it's
it's
general
and
it
can
handle
that
it's
not
as
just
kind
of
getting
in
terms
of
that
way.
It
makes
a
little
bit
easier
to
program
in
general
because
you
don't
get
if
you
can
be
more
general
with
the
convolutional
area,
you
don't
have
to
have
nearly
as
many
special
cases.
F
F
That
you
could
think
of
like
even
for
a
pde
code.
That's
some!
That's!
Actually!
What's
what
I
locally
worked
on,
basically
updating
that
the
first
order,
a
backward
euler
to
a
second
order
and
actually
doing
that
kind
of
that
numerical
convolution
after
you
get
the
step
response
actually
simplifies
on
the
user
end.
F
So
now,
at
that
point,
if
you
can
come
up
with
a
good
time,
steps
that
can
work
for
all
the
cases,
then
what
the
user
need
to
put
it
in
is
just
just
the
waveform
and
and
where
the
time
gates
are,
that
sort
of
solves
kind
of
this
complicated
problem.
How
is
that
the
time
steps
accurately
to
get
the
good
response,
so
this
is
a
potential,
a
pathway
that
we
can
think
of.
That
complicates
the
inverse
problem,
but
at
least
for
the
fourth
sense,
it's
actually.
A
Yeah
and
the
way
that
I'm
looking
at
doing
it
right
now,
it
doesn't
really
complicate
it
any
much
more.
I'm
just
trying
to
go,
I
go
through,
and
then
I
mean
it:
it's
not
that
expensive
to
generate
a
spline
or
something
and
then
generate
the
and
generate
the
linear
representation
of
that
spline
or
the
linear,
matrix
representation
of
that
operation.
A
A
A
G
Yeah
well,
I
was
working
on
the
mtpr,
but
then
I
kind
of
put
that
down
got
distracted
with
the
tiling
yeah.
I
started
there
did
a
bunch
of
tiling
tests
on
the
cluster
for
small
problems,
just
normal
synthetic
models,
but
then,
as
I've
moved
into
more
of
our
recent
projects,
where
we
have
like
over
5
million
cells
into
a
model,
the
ram
is
just
adding
up
so
quickly.
G
You
make
like
a
few
tiles
and
sure
the
the
sensitivity
matrix
is
being
stored
on
the
main
machine,
but
it's
just
calculating
the
fields
and
like
still
having
like
a
2
million
mesh
on
every
tile
it
just
it
adds
up
pretty
quick,
so
yeah.
It's
really
important.
I
think
right
now
to
have
the
octree
mesh
and
doing
the
sub
or
the
nested
arteries.
G
The
other
thing
that's
working
against
me
right
now
is
I'm
they
want
their
everybody's
doing.
Inversions
and
diaz
right
now
wants
tensor
meshes,
so
I'm
doing
a
global
tensor
and
that's
killing
me
as
well
so
yeah
just
trying
to
deal
with
the
ram.
Oh
yeah,
I
did
some
subtiling
there
dom
again.
I
was
trying
it
on
one
of
our
recent
data
sets
and
it's
just
too
much
ram,
so
I'm
gonna
have
to
make
up
a
like
a
little
synthetic
model
to
test
further.
G
Otherwise,
it's
just
as
you
keep
getting
those
tiles,
it's
just
more
and
more
ram,
because
it's
it's
got
to
make
its
own
model
for
each
one
of
them,
and
that
seems
to
be
the
killer
right
now.
Sorry,
what
the?
What
method
are
you?
Are
you
looking
into
the
tile
or
sorry
that
would
be
the
styling
by
frequency
and
by
space,
so
empty
yeah
using
the
empty.
H
G
H
E
E
A
E
A
Yeah-
and
it's
like
like,
if
you
don't,
I
mean
I
can
look
over
to
while
you're
in
the
middle
of
things,
if
you
want
to
like
do
that
as
well,
so
you
have
to
wait
until
the
final
draft.
Have
it
reviewed
or
anything,
can
you
repeat
that,
if
you
want,
I
can
like
look
over
some
of
the
things
already
that
you've
added
just
kind
of
like
oh
catch,
a
few
things
in
there
for
other
stuff
going
on
that
could
be.
E
Done
soon,
I
think
there's
a
couple
questions
that
I
have
about
the
level
of
detail.
I
want
to
describe
some
of
the
the
base
mesh
classes
because
there's
no,
those
are
so
general
that
I
don't
want
to
make
the
description
of
what
that
functionality
is
very
specific
to
the
the
coordinate
system
and
things
like
that.
So
yeah
that
would
be
pretty
useful.
A
D
Yeah
sure,
thanks
for
bringing
that
forward,
I
think
the
biggest
ask
right
now
is,
if
folks
can
take
a
look
at
the
pull
request.
That's
on
there
right
now.
I
see
both
dieter
and
soggy.
Have
it
would
be
great
to
just
see
I'll
just
drop
it
in
the
chat
there
and
I'll
put
in
the
notes
here
too,
just
to
see
some
sort
of
indication
I
mean
if
you've
got
any
suggested
changes.
That's
that's
great
feel
free
to
make
some
suggestions
if
you're
happy
with
it.
D
It
would
be
great
to
see
just
indications
from
folks
that
that
sort
of
you're
you're
on
board.
So
if
you
don't
mind
just
leaving
a
comment
or
given
a
thumbs
up
or
or
something
like
that,
just
to
indicate
a
bit
of
you
know,
community
consensus,
I
think,
is-
is
important
to
show
that
people
have
people
in
the
community
have
looked
at
this
and
it's
not
just
sort
of
three
of
us
who
are
like
yep.
This
is
the
simpeg
governance.
D
So
if
you
don't
mind
sort
of
giving
a
bit
of
a
an
indication
of
just
if,
if
this
works
for
you
and
if
it
doesn't
feel
free
to
raise,
raise
questions
and
and
thoughts
and
ideas
there,
and
so
we
can
sort
of
put
a
check
mark
on
that
and
move
on
towards
the
the
next
things
which
I
think
what
we
should
start
looking
at
is.
I
think
I've
got
a
couple.
D
I
haven't
written
those
down
too
much
yet
but
sort
of
some
some
steps
forward.
I
think
you
know
we
want
to
think
a
bit
about.
We've
talked
about
sort
of
website
updates
things
like
that
that
we
might
want
to
be
including
to
highlight
visibility
of
folks
in
the
community
and
contributions
from
multiple
institutions,
but
also
showing
you
know,
collecting
papers
and
collecting
places
where
simpe
is
used
in
a
sense
to
track
some
of
our
impact.
D
A
A
D
That
sounds
good.
We
can
plan
to
hit
the
merge
at
the
meeting
if,
if
that
works
with
folks-
and
if
you
do
have
questions
concerns
thoughts,
anybody
wants
to
chat
about
this
more
than
happy
to
yeah,
yeah
and
then
feel
free.
Also
to
if
there
are
things
that
you
think
should
be
prioritized.
I
know
we
sort
of
just
put
this
on
the
spot,
but
to
open
up
issues
in
in
that
repository.
D
So
those
for
me
are
kind
of
the
two
main
things
that
come
to
mind.
But
if
there's
other
things
that
you
think
should
be
prioritized
or
that,
even
if
you
don't
think
it
needs
to
be
prioritized.
But
it's
something
we
should
be
thinking
about,
feel
free
to
just
open
up
some
issues
there.
And
we
can
use
that
repository
as
a
bit
of
a
place
to
to
track
ideas
and
and
aspects
of
the
community.
D
But
you
know
what
right
now,
if
you've
got
ideas
in
mind
for
for
road
map
items,
let's
track
them
as
issues
and
then
we
can
sort
of
figure
out
from
their
like.
What's
what's
the
best
way
to
kind
of
consolidate
all
of
this
stuff,
but
yeah,
let's
get
them
down,
and
so
if
an
issue
works,
that's
great
or
even
if
you
want
to
draft
it
in
a
google
doc
and
then
just
link
that
in
an
issue-
that's
that's
fine
too,
but
yeah.
A
Are
there
any
other
items
outstanding
items?
People
would
like
to
chat
a
little
bit
more
about.
Do
I
want
to
did
or
do
you
want
to
make
it?
I
want
to
set
a
time
for
that
q
a
I
want
to
try
to
get
it
hey.
So
what
sort
of
time
period
are
you
thinking
about
it,
or
do
you
want
to
do
that?
Can
we
set
that
up
afterwards
at
some
point.
C
All
right,
no,
I
thought
I
I
make
this
this
course.
This
year
we
can
post
there
maybe
a
week
or
two,
and
then
I
mean
it's
gonna,
basically
link
the
soggy.
Whoever
is
working
on
it.
I
we'll
have
to
save
and
a
half
time,
but
maybe
if
there
is
a
short
meeting
like
today,
we
could
do
it
afterwards,
half
an
hour
meeting
and
then
do
this
session
or.
A
F
Has
nothing
I
met
up
with
nick
and
other
people
from
barrick
of
the
cscmt
theater?
This
was
kind
of
ignited
of
a
hackathon
between
myself
and
peter,
so
we
kind
of
implemented
the
jvac
and
jtvac
in
in
the
syntax
slide,
using
the
em
pipe
mod
as
a
as
an
engine.
So
I
was
looking
for
on
a
good,
a
large-scale.
E
F
F
So
I
got
one
data
from
that
about
what
mike
mcmillan
did
in
san
antonio,
not
antonio
deposit
and
that'll,
be
a.
I
haven't
inverted,
but
I'm
just
collecting
the
data,
but
what
we
are
going
to
get
from
nick
and
beric
seems
quite
interesting
and
also
it
was
good
to
chat.
What
were
the
challenges
that
they
have.
So
the
typical
tool
that
they
are
using
currently
is
a
2d
mt
code,
and
so
you
can
still
fit
the
data
well,
but
their
confidence
is
very
low,
so
they
get
the
result.
F
It
only
provides
apparent,
resistivity
and
phase,
and
sometimes
the
phase
information
is
wrong.
So
I
think
all
this
technical
challenges
kind
of
made
them
hard
actually
to
run
the
3d
version,
even
though
they
have
the
good
code.
I
think
that's
where
they
are
and
what's
actually
interesting.
A
lot
of
the
historical
data
set
doesn't
know
where
the
source
locations
are,
but
they
do
have
lots
of
historical
data
that
can
potentially
be
used
and
reinverted
using
the
latest
code.
F
So
I
asked
them
if
we
can
actually
kind
of
revisit
this
problem
and
make
it
possible
to
reinvert
all
data,
as
well
as
the
high
quality
latest
data.
Then
that
seems
like
it's
quite
impactful,
because
what
they
really
want
to
do
by
using
the
csmt
they
want
to
go
a
little
bit
deeper
than
dcip
so
like
the
maximum
of
dcip,
is
500
to
700
meter
or
maximum
kilometer.
F
But
by
doing
csmt
they
come
go
a
little
bit
deeper
like
a
couple
of
kilometer
or
maybe
even
a
kilometer,
is
actually
pretty
good.
So
it
seems
like
a
timely
topic,
so
anybody
interested
and
yeah,
let
me
know
I'm
happy
to
it's
mine
nobody's
paying
me
so
I'm
just
doing
my
own
and
my
own
interest
in
volunteering
to
do
so.
But
I
think
there
are
some
potentials,
I
think,
and
if
you
actually
do
it
right
at
this
point,
I
think
that
could
be
a
really
useful
tool
for
the
society.
F
H
F
C
G
G
I
Maybe
we
see-
maybe
we
think
so
just
like
some
just
wanted
to
mention
that
I
was
part
of
the
discussion
like
the
access
to
efficient
object.
Code
was
part
of
the
discussion
because,
like
his
thing
was
that
he
had
like
some
sort
of
like
the
regional
course
empty
station
and
in
some
area
wherever
you
can
put
it,
because
that's
kind
of
like
a
very
bushy
area.
I
So
he
was
mentioning
that
field
is
kind
of
complicated
but
like
in
some
area,
I
was
trying
to
have
like
a
like
a
a
finer
stay
final
station.
So
he
was
kind
of
looking
at
this
like
regional
and
then
looking
at
the
local,
local
one,
starting
and
constraining
with
the
regional,
but
he
was
limited
in
in
what
he
could
do
with
with
the
measure.
I
Three
was
kind
of
part
of
it
yeah,
maybe
that's
very
minor
to
I
just
I
haven't
checked,
but
I
just
realized
that
somebody,
like
maybe
joe
you.
You
saw
that
that
somebody
on
pi
mat
solvers,
mentioned
that
by
my
server
was
not
working
for
him
for
with
mkl
2021,
and
I
just
checked
for
myself
and
I'm
actually
working
with
mkl
2020.
So
I'm
not
sure
if
it's
something
or
if
that
was
just
this
person
issue,
but
I
just.
A
Did
you
did
you
see
that
yeah?
I
thought
I
haven't
looked
into
anything
more
with
it.
I
think
it's
it
might
have
something
to
do
with
a
form
like
a
change
in
the
way
that
they're
pushing
like
printing
out
the
mko
version,
but
I
haven't
looked
into
into
it
too
much
and
I
just
kind
of
googled
it
didn't
you
kind
of
like
the
changes
between
them
really
quick
and
test.