►
From YouTube: SimPEG Meeting October 20th
Description
SimPEG weekly meeting from October 20th, 2021
B
A
A
Yeah,
as
always,
add
yourself
to
quick
reports,
I
was
hoping
to
discuss
some
agenda
items,
but
I
don't
think
the
right
people
are
here
today
to
discuss
this.
I
had
a
small
slide
discussion
yesterday
with
doug
and
lindsey
and
teal
about
it,
but
just
putting
it
on
the
radar.
A
This
is
part
of
my
quick
report
anyway,
so
I
have
put
a
bunch
of
I
put
my
changes
onto
the
cross
gradient,
my
pr
my
suggested
changes
for
cross
gradient
on
there.
It's
I
did
it
as
a
pr
onto
the
xiaolong's
branch
like
from
my
own
stuff,
so
we
can
look
at
that
like
so
I
wanted.
I
wanted
him
to
be
able
to
see
the
changes
that
I
was
making,
instead
of
just
pushing
it
directly
into
his
branch.
A
I
just
wanted
and
just
wanted
them
to
see
it
first
and
provide
their
input
as
well
like
make
it
a
little
bit
more
back
and
forth
and
transparent
as
far
as
just
like,
okay.
Well,
this
is
the
way
I
think
it
should
be
done.
So
I'm
doing
it
that
way.
I
just
I
didn't
want
that.
I
wanted
them
to
see
like
well.
This
is
what
I
had
in
mind.
A
B
A
To
the
other
pr
on
jo
long's
branch,
it's
not
as
findable,
so
that's
why
I
put
up
here
or
licked
it
there,
but
in
general
I
would
was
thinking
about
having
a
discussion
on
the
naming
and
we
can
push
to
put
this
off
to
next
week.
It's
okay,
the
naming
scheme
of
what
we
call
when
there
there's
two
models
linked
together
in
a
joint
inversion.
What
that
regularizer
is
called
my
opinion,
like
obviously
joint,
is
too
big
because
joint
encompasses
way
more
things
than
that
right.
A
My
also
personal
thinking
is
that
the
word
coupled
is
also
too
vague,
because
I've
seen
that
used
for
many
other
types
of
inversions,
many
other
meanings
for
that
as
well.
So
my
personal
thought
is
that
I
kind
of
like
the
word
similarity
measure,
because
it's
explicit
in
doing
what
it's
like
and
saying
what
it
is
like.
It's
like.
Oh
I'm,
going
to
measure
how
to
how
similar
these
models
are.
So
I'm
going
to
use
a
similarity
measure
to
do
that.
A
A
D
Joe,
why
do
we
need
that
kind
of
distinction?
Because
a
lot
of.
A
Mostly
because
it's
a
level
abs
of
an
abstraction
that
we
can
use
for
directives,
so
some
of
those
like
a
lot
of
those
directives
can
be
pretty
similar
to
each
other.
It
just
makes
it
easier
to
design
around
okay
we're
expecting
this
regularization
to
have
two
specific
models.
D
A
A
It's
I
don't
know
it's
just
kind
of
I'm
not
sure,
if
still
sure,
if
that's
the
right
word
yet,
but
that's
kind
of
where
I'm
leaning
other
than
that,
so
the
cross
gradient
pr
has
a
little
like.
If
you
see
the
changes
there
that
I'd
suggested
to
it,
it's
a
little
bit
different
internal
implementation.
That,
from
my
little
my
experience
saw
a
little
bit
less
of
the
checkerboarding
issues
that
were
kind
of
apparent
in
some
of
the
other
examples.
A
A
A
B
B
Little
details
that
I
fixed
everything
seems
to
be
working
pretty
well,
except
for
the
frequency
domain
inversion,
where
we
invert
for
the
height
of
of
the
the
source
so
that
one's
a
little
bit
interesting.
I
mean
I
understand
how
the
mapping
is
working
and
I
can
get
it
to
run,
but
some
something's
in
there.
B
I
don't
know
if
maybe
there's
a
different
strategy
for
trying
to
do
regularization
on
it
yeah
some
something
is
not
working
very
well
in
there
and
we
need
to
figure
it
out,
but
mostly
it's
looking
quite
good.
B
The
other
thing
is
trying
to
do
an
initial
build
of
the
website
for
simpeg
using
the
the
same
kind
of
structure.
We
have
for
discrete
ties,
so
I've
got
an
initial
build
of
that
and
the
layout
of
that
that
the
layout
style
of
the
website
is
now
like
it
is
for
discretize
and
just
kind
of
starting
to
build
the
api
and
fill
that
in
slowly.
B
But
I've
got
a
couple,
I
guess
incompatibilities
with
the
graphviz
yeah,
so
I
I
even
tried
to
build
the
discretized
website
from
main
branch
locally
on
my
computer
and
I
ran
into
the
same
error.
So
it's
not
that
there's
something
that
that
I
didn't
do
properly
in
simpeg.
There's
there's
some
kind
of
incompatibility
with
either
the
os
I'm
using
or
the
or
my
my
python
environment.
B
A
D
B
It
might
have
broken
in
the
the
directive
that
chooses
the
initial
beta
based
on
the
eigenvalue,
I'm
not
sure
exactly
where
it
was
was
breaking,
but
it
basically
said
you
need
fields,
it
doesn't
have
fields
and
so
similar
to
some
of
the
other
simulation
classes
that
has
kind
of
a
simple
of
fields,
objects
or
when
deep
red
just
calls
fields.
I
did
the
same
thing.
Yeah.
A
B
A
B
A
A
B
Oh,
that's
that's
a
good
question,
so
I
can
maybe
play
around
with
it
a
little
bit
more
and
kind
of
maybe
document
what
the
error
actually
is.
E
There
might
be
some
places
where,
like
I
think,
we'd
pass
the
fields
around
in
the
data
misfit
because
for
some
examples
like
if
we've
computed
the
fields
for
3d
time-domainium
like
we
don't
want
to
recompute
them,
we
should
allow
none.
But
I
wouldn't
be
surprised
if
there's
places
where
we
assume
that
something's
passed
in
there.
But
if
you
pass
we
should
be,
we
should
be
able
to
pass
around
the
none.
I
think.
A
B
No,
no,
not
not
really.
Those
are
the
two
things
I'm
working
on
these
days.
E
Lindsey
yeah,
a
quick
update,
the
line,
current
pull
request
is
basically
ready
to
go
so
we've
got
it
now
for
time
domain
frequency
domain.
Well,
the
eb
formulation
is
already
there,
but
now
the
hj
formulation
is
there.
So
you're
welcome
to
take
a
look
feel
free
to
add
any
comments,
still
missing
tests.
E
Somebody
has
a
test
or
an
example
that
you
would
like
to
just
turn
into
a
test.
That'd
be
great,
if
not
that's
fine,
I'll,
try
and
get
back
to
that,
but
it'll
probably
be
a
couple
weeks.
So
yes,.
A
E
Yeah,
I
don't
think
it
needs
to
be
too
like
an
analytic
or
anything,
but
I
think
what
would
be
nice
is
to
test
that,
like
the
eb
and
hj
or
giving
something
comparable
for
like
a
wire
would
be
a
nice
test,
but
that
will
take
a
little
bit
to
set
up.
I
could
definitely
set
up
the
very
simple
test
that
checks
that
you
know
ones
are
where
they're
supposed
to
be,
but.
B
A
B
Was
it
I
think
I
already
have
like
a
code
validation
thing
set
up
for
this
already.
I
could
try
and
run
it
through
that
on
the
branch
that
you're
working
on.
E
E
Yeah
and
I
put
some
checks
in
there
just
because
of
things
I've
run
into
in
the
past,
to
make
sure
that
it's
like
the
divergence
is
only
where
it
should
be
so
make
sure
you
haven't,
like
accidentally,
missed
a
face
or
have
a
disconnected
wire
or
anything.
So
there's
there's
a
few
sort
of
sanity
checks
built
in
there
so
feel
free
to
take
a
look.
I'd
be
curious
to
hear
any
thoughts.
F
Yeah
remember
I'll
go
ahead.
I
remember
with
that
original
implementation
that
we
had
of
it
that
I
was
using
for
some
of
the
stuff.
In
my
thesis
it
worked
best
if
we
were
in
an
octree
mesh
if
it
was
kind
of
if
your
source
was
only
running
through
like
your
core
cells
and
if
there
weren't
like
changes
in
the
cell
size.
I
remember
seeing
some
things
with
that,
so
I
don't
know.
I
guess
that's
just
another
thing
to
think
about,
because
I
don't
know
if
anybody
has
looked
at
that
or
addressed
that
yet.
E
That's
a
good
point.
I
don't
check
for
that,
but
at
this
point
I
also
think
that's,
maybe
a
a
user
thing.
We
can
try
and
put
some
guardrails
in
place,
but
yeah
at
least
actually
right
now
it's
checking
the
divergence.
And
so
if
you
did
something
funky
with
an
octree
mesh,
it
might
catch
that,
but
I'm
not
too
sure.
A
Right
now,
with
the
actual
mesh
for
the
other
formulations,
it
just
it,
it
finds
right
the
path,
the
cells
to
the
path
and
makes
sure
they're
all
at
the
same
level
and
like
okay.
Well,
if
it,
if
it
crosses
level
changes,
then
you're,
then
it's
weird
and
it
gives
a
warning
back
at
least
letting
you
know
you
did,
that
you
probably
shouldn't.
G
I
actually
have
a
curve
note
started
where
I
was
playing
around
with
the
line
current
before
so
I
have
like
an
example:
kind
of
set
up
and
just
playing
with
and
and
then
doing
the
code
coverage
for
the
simulation
mt.
So
I'm
writing
tests.
So
I
can
probably
pull
that
in
or
at
least
tackle
or
try
that
out
for
sure.
E
G
Yeah,
I'm
almost
there
seventy-four
percent
just
gotta,
get
to
78
and
then
that
pr
I'll
be
good
to
go
so
I'm
progress.
Finally,
I'm
getting
there.
A
G
Exactly
yeah
already
found
a
few,
so
that's
it
works
out
really
well
and
then
now
I
just
I
don't
makes
for
a
better
understanding
of
the
code.
Actually
it
was
kind
of
nice
to
go
through
see
that
all
the
stuff
you
did
in
the
base
there
might
have
still
some
questions.
I
keep
telling
I'm
gonna
bug
you,
but
I
don't
but
I'll
get
there
one
day.
G
C
You
know
I
made
not
much
much
commit
on
simpek
for
the
past
week,
yeah
like
so
I
moved
all
this
pgi
improvement
to
that
regularization
reflector
here.
So
it's
it's
waiting
for
the
pr
1038
from
them.
C
It
should
be
ready.
This
one
should
be
ready
to
go,
but
I
believe
and
like
he
didn't
update
it
from
it,
didn't
pull
the
latest
main
in
it.
I
think
that
was
the
only
thing
in
that
year
from
done
going
in
so
yeah.
That's
a
that's!
That's
I'm
just
waiting
for
that.
In
terms
of
that
specific
pr
and
yeah.
A
To
edit.
C
D
I
think
you're,
like
particular
for
mayan
a
couple
like
not
that
necessarily
related
to
syntax
a
couple,
interesting
things
that
I
learned
recently
all
the
age.
You
journals.
It's
actually
force
you
to
upload
the
data.
So
that's
the
requirement
actually
now,
which
is
quite
interesting,
so
yeah.
I
was
actually
putting
together
a
data
repositories
through
stanford,
digital
library.
I
think
I
should
I
could
have
used
the
other,
but
yeah
rosemary
prefers
whatever
stands
for
affiliation
stuff,
but
that
seems
quite
interesting.
D
D
Oh
data
is
not
available,
so
I
thought
that's
that's
interesting
and
the
other
thing
we're
sort
of
starting
a
project
with
the
groundwater
modelers
and
one
kind
of
interesting
avenue,
is
how
do
we
really
integrate
the
geophysics
into
the
groundwater
modeling
calibration,
basically
in
the
the
groundwater
inversion
and
the
way
that
they
want
to
do
is
basically
putting
whatever
geophysical
sulfur
into
pests.
It's
a
pest,
I'm
not
sure
you
guys
are
familiar
with
it.
D
It's
just
an
optimization
package,
but
but
it's
quite
general,
so
you
can
put
whatever
program
if
you
follow
the
format,
so
I
thought
it's
actually
a
relatively
simple
task:
that
okay,
we
just
need
a
wrapper
of
simpac
and
then
plugging
in
to
pest,
and
they
can
use
in
whatever
way
they
want
so
yeah.
I
think
a
given
sort
of
the
sort
of
the
size
of
the
community
is
relatively
big
using
that
kind
of
the
types
of
softwares
like
there's
a
couple
others.
D
So
I
thought
this
is
a
good
opportunity
to
actually
think
about
kind
of
plugging
in
syntax
to
other
other
software
which
is
sort
of
where
we're
moving.
I
guess
anyway,
so
that's
that's
where
I'm
not
really
working,
I'm
just
thinking
about
the
potential
and
not
actually
working,
but
probably
one
point
gonna
we're
gonna
work
on
that
anyway.
That's
that's
where
I
why.
A
Okay,
well,
if
there's
nothing
else,
I
think
it's
a
good
meeting
time.
I
don't
think
we
have
anything
big
coming
up
next
week
is
afternoon
meeting
it's
fourth
wednesday,
so
three
o'clock
next
week
see
y'all
there.