►
From YouTube: SimPEG Meeting October 27th
Description
SimPEG weekly meeting from October 27th, 2021
A
Other
thoughts
over
the
weekend
about
those
naming
conventions
for
the
joint
inversion
packages
or
if
any,
also
kind
of
curious.
If
anyone
ever
has
gotten
a
chance
to
play
around
with
the
cross
gradient
branch,
I
had
suggested
on
michelle
wrongs.
I
still
have
not
heard
back
from
them
about
any
things,
they've
any
other
stuff
they
played
with
yet
so
it's
still
there.
A
If
not,
I
can
move
on
to
some
quick
reports,
as
always
just
add
yourself
to
the
list
or
a
little
bit
as
far
as
today,
something
we
can
just
do
it
around
for
everyone
just
to
get
a
quick
update,
not
too
much
me
senpaig
related.
I
pushed
a
quick
update
to
the
discretized
package,
saying
that
you
can
refer
to
levels
and
some
and
the
refined
functions
as
negatives
and
they
will
index
from
the
back.
A
So
if
you
like
want
to
refine
at
the
maximum
level,
you
can
just
do
it
as
negative
one.
Instead
of
having
to
figure
out
what
the
max
level
was
or
pass
it
just
a
little
bit
cleaner
implementation,
and
as
I
was
as
we're
doing
that
we
still,
I
would
still
like
to
separate
up
that
helper
function
into
the
three
different
parts
of
it.
A
A
So
it
makes
a
little
bit
cleaner
as
far
as
what
the
parameters
do,
because
they
do
slightly
different
things
depending
on
what
option
is
being
passed
to
it,
just
a
little
bit
unclear
and
then
there's
also
the
issue
that
I
feel
like.
I
should
have
caught
it
much
sooner,
but
it's
the
way
that
the
levels
are
defined
in
that
refined
tree
xyz
function
is
backwards.
The
way
if
they
that
they
are
defined
on
the
mesh
itself.
A
A
A
And
that
I
think,
there's
there's
been
a
few
issues
with
azure
pipelines,
not
launching
tests
of
some
of
the
pr's
I'm
still
working
on.
It
seems
really
odd.
It
doesn't.
I
mean
obviously,
we've
changed
nothing
in
our
scripts
and
now
it's
just
all
of
a
sudden
they're
just
automatically
canceled.
A
C
And
not
much
in
terms
of
synthetic
development.
Well,
if
you
got
dom's
pr
in
going
might
be
worth
looking
at
the
one
like.
I
know
it
didn't
touch
pgi
and
I
would
have
to
go
back.
It's
been
a
few
weeks
now,
but
I'm
just
wondering
if
you're
gonna
break
it
like
if
it's
changed
or
this
chain,
if
he
has
changed
sensitivity,
weighting
and
the
normal
regularization,
but
not
pgi,
it
might
breaks
the
pgi
example.
If
that's
just
maybe
something
to
look
at
just
and.
C
A
Is
gonna
be
helpful,
whether
you're
using
a
simple
with
a
tree
mesh
versus
when
you've
got
the
you
know,
since
updating
the
sensitivity
weightings
it'll
be
it'll
lead
to
less
confusion
about
what
you're
supposed
to
be.
C
Yeah,
so
the
pia
I
put
that
saying
after
that,
after
dom
was
basically
just
merge.
Merge
like
the
simple,
pgi
and
pgi,
so
that
now
it's
just
one
and
that
it
has
automatically
the
cell
weights
in
it.
But
before
like.
But
all
the
examples
use
the
simple
pgi
without
the
cell
waiting
so
like
you're
losing
that
cell
actually
yeah.
C
A
B
A
C
Okay,
so
then
it's
we
can
make
an
easy
fix
on
spl
to
just
calling
pgi
instead
of
simple
pgi
in
the
example,
and
that
will
fix
that
will
be
a
simple
effects.
Okay,
if
you
don't
want
to
bring
the
more
massive
pr,
that's
where,
like
I've,
actually
collapsed,
a
simple,
pgi
and
pgi,
but
that
should
probably
happen
at
the
same
as
we
we
collapsed,
simple
and
thick
enough,
but
anyway
like
easy,
let
me
know
and
that
we
can
make
an
easy
fix
on
that
for,
for,
if
you
just
want
to
have
a
release.
A
Okay,
oh
another
thing
that
I
started
doing
locally
as
I'm
working
on
branches
that
will
remove
the
deprecations
for
synthetic
and
discretized,
because
that
should
be
coming
soon
within.
A
I
think
so
I
think
the
process
is
the
next
one.
They'll
be
changed
to
future
warnings,
so
they'll
be
louder.
B
B
E
Yeah
not
not
too
bad
just
kind
of
working
on
this
paper
draft
still
trying
to
remake
some
figures.
I
had
a
situation
kind
of
where
my
main
gravity
anomaly,
that
kind
of
sits
over
this
little
volcanic
center
and
this
other
active
area
were
kind
of
offset
at
like
a
45
degree
angle
from
I
were
kind
of
pointing
straight
northeast.
If
I
wanted
a
cross
section
through
there,
so
I
came
up
with
a
way
to
make
diagonal
cross
sections.
E
I
don't
know
if
it's
the
best
way
to
do
it,
but
it
works
yeah,
so
that
was
kind
of
what
I
was
playing
around
with
a
little
bit
crispy,
getting
a
nice
way
to
draw
that
diagonal
cross
section
and
put
it
on
to
an
actual
2d
mesh
so
that
I
can
still
use
like
the
gradient
based
function
that
I
use
to
kind
of
draw
the
outlines
of
my
synthetic
models
in
the
topographic
surface.
E
C
C
Do
we
want
to
have
like
the
the
possibility
to
do
diagonal
cross
section
with
with
like
the
plot
slice
tool
within
discretize,
or
is
that
just
not
something
we
we
should?
We
want
to
have.
C
That's
it.
It
just
feels
it's
not
the
first
time
we
talk
about
it
or
that
someone
asked
for
it
or
need
it
didn't
need
it.
So,
just
asking
if
like
if
that,
if
mike
can
contribute
something
or
if
there
was
like
a
deeper
reason,
that's
we
should
not
have
it
because
you
well
mike.
You
probably
wrote
that
in
python,
while
plot
slice
maybe
might
have
some
roots
on
the
c
code.
E
Yeah,
I
don't
know
the
way
I
did
it
was
using
pi
vista.
I
wrote
it
out
to
like
a
vtk
object,
sliced
it
sliced
it
using
pie,
vista
and
then
got
the
points
off
of
the
slice
found.
The
cells
in
my
model
that
were
kind
of
closest
or
kind
of
that
those
points
fell
within
and
then
just
kind
of
did
a
nearest
neighbor
interpolation
to
get
it
on
to
a
a
2d
mesh
that
I
then
kind
of
translated
from
x
y
z,
points
to
z,
points
and
profile
distance.
So.
B
Something
like
that
might
be
a
really
useful
example
for
the
documentation.
Is
that,
even
if
it's
not
something
that
we
necessarily
want
to
include
having
something
that
we
can
point
people
to
and
say
like
here's,
one
way
to
do
it,
and
even
if
you
want
to
put
at
the
top
that
like
hey
this
is
my
my
approach.
If
you've
got
improvements
be
great
to
find
out,
it
could
be
kind
of
a
cool
way
to
crowdsource
some
of
the
some
of
this
problem.
E
E
E
A
E
That's
a
good
point.
I
actually
meant
to
ask
I
thought
about
that
a
week
or
so
ago
and
meant
to
ask
you
about
it,
but
obviously
didn't
get
around
to
that.
So
all.
B
A
A
B
B
B
B
What
are
folks
thoughts
on
having
one
in
december
like
we
could
try
to.
I
also
appreciate
its
agu
end
of
term
christmas.
It's
a
month
that
disappears
quickly,
so
we
could
also
just
postpone
and
and
do
something
in
january.
I
don't
know
anybody
have
opinions.
C
B
Okay,
well,
let's
tentatively
plan
to
punt
it
because
we
don't
have
anybody
lined
up,
but
that
is
then
maybe
a
call
for
folks
here
to
tap
people
that
you
want
to
hear
from
and
just
say
we're
looking
for
summon
our
speakers
in
the
new
year.
So
that
gives
people
plenty
of
time.
But
if
you
know
people
who
would
like
to
come
and
connect
with
the
group
and
even
if
they're,
not
specifically
working
with
simpeg,
if
they're
doing
inversions,
that's
great
and
having
them
come,
give
a
talk
would
be
would
be
awesome.
B
C
I
mean
I
have
a
then
I
have
a
question
on
that
is
like
he's
like
I
I
I
don't
know
if,
if
everyone,
but
me
knows
those
people,
but
who
are
those
people
working
on
that
oh
simpek
over
project,
I
don't
even
know
who
they
are
or
like.
I
know
it's
for
empty
processing,
but
it's
that's
all
my
all.
I
know
about
that.
B
Sure
that
was
the
meeting
joe
and
I
were
just
in
it's
carl
kapler
and
jared
peacock
are
two
people
who
are
the
driving
forces
and
joe
myself
and
doug
are
kind
of
trying
to
help
out
in
the
in
the
background
yeah.
We
can
definitely
float
that
the
project
is
sort
of
wrapping
up.
So
I
think
something
well
this.
B
This
round
of
the
project
is
wrapping
up
and
we're
putting
in
hopefully
a
new
proposal
with
iris
soon-ish,
but
something
in
the
new
year
yeah
we
can,
we
can
ask
and
see
if
they
would
be
game.
C
B
B
They
wanted
to
plug
into
a
community
that
already
exists
and
see
if
there's
folks,
who
might
be
interested
in
sort
of
crossing
in
over
there
and
dabbling
in
with
the
processing
software,
so
actually
having
having
a
seminar
that
invites
invites
folks
here
so
like
the
first
year
of
the
project,
was
basically
kind
of
getting
a
base
level
in
place
and
so
transferring
gary
egbert's
emtf
code
from
that.
Well,
he
he's
got
an
original
fortran
version.
B
He's
got
a
matlab
implementation
and
the
goal
was
to
basically
get
up
to
the
same
level
of
functionality
as
the
matlab
implementation
and
so
actually
with
carl's
work.
He's
done
some
really
great
work
and
gotten
it
I'd
say
in
some
cases
surpassing
the
the
matlab
code
that
gary
provided,
but
there's
still
it's
still
kind
of
in
the
early
stages
of
a
project
right
like
we.
We
were
a
few
years
ago
in
simpeg,
where
you
still
need
to
get.
B
D
F
Just
mostly
more
doc
strings
I've
gone
through
the
utilities
and
for
the
most
part
I
think,
there's
probably
one
or
two
that
I
missed
and
then
I
started
documenting
all
the
mappings
mapping
classes.
There
is
a
lot
of
undocumented
math
in
there,
so
I'm
almost
through
those,
but
it
was
quite
tedious,
but
I
I
am
able
to
build
the
website
with
the
latest,
the
latest
style,
the
pi
data
theme
and
getting
all
that
stuff
due
to
kind
of
build
with
no
errors.
So
that's
really
good.
F
I
think
my
approach
right
now
is
that
I'm
going
to
start
with
documenting
the
base
classes
for
simpeg,
I'm
going
to
stay
away
from
regularization
and
directives
for
now.
So
I
think
there's
a
lot
of
active
work
being
done
on
those
and
I'm
I'm
putting
off
putting
any
doc
strings
or
any
doing
the
api
for
simulation
classes.
Yet
I
guess
what
I'm.
What
I
want
to
do
is
start
with
the
things
that
I
don't
believe
are
actively
being
worked
on
or
changed
and
then
move
to
the
other
stuff.
F
I
also
did
some
work.
I
can't
remember
if
I
talked
about
it
last
week,
but
I
I
did
do
some
work
on
the
em-1b
stuff.
I
just
can't
remember
if
it
was
one
week
or
two
weeks
ago,
so
I
I
did
find
a
couple
of
bugs
and
I
did
update
the
the
tutorial
examples
to
to
be
consistent
with
how
the
coding
is
managed,
so
that
stuff's
kind
of
been
done
so
maybe
soggy.
F
If
you
want
to
take
another
look
at
it,
that
would
be
great
one
of
one
of
the
ones
that
I'm
finding
is
a
little
strange
is
inverting
a
frequency
domain
sounding
for
the
conductivity
structure
and
the
flight
height.
That
does
not
seem
to
be
working
very
well.
It
was
working
before,
but
that
doesn't
seem
to
be
particularly
stable
for
some
reason
and
I
haven't
dug
into
it
yet.
G
I
think
it's
it's
not
ready
for
inverting
the
height.
I
think
we
haven't
clarified
the
definition
of
what
is
the
height
means
when
you've
got
multiple
sources
and
receivers.
So.
G
It's
it's
not
supposed
to
work,
so
it's
yeah.
I
was
just
like
kind
of
putting
that
as
a
leftover
that
we
can
tackle
later
so.
G
Works,
I
think
we're
good
to
go.
A
A
B
F
A
G
Yeah,
I
don't,
I
don't
think
at
least
I
think
that
that
that
mechanics
has
not
tested
not
sure
like
whether
code
runs
or
not,
but
yeah
we
haven't
actually
addressed.
F
B
F
It
should
be
yeah
and
I
found
a
couple,
a
couple
bugs
which
got
fixed,
so
that
was
that
was
good.
I
and
I
I
did.
I
updated-
I
updated
the
branch
so
that
everything
seems
to
be
up
to
date
with
the
current
simulation
or
at
least
last
week,
unless
things
have
been
merged
in
since
so
it's
in
a
much
better
place
to
get
brought
in
than
it
was
okay.
G
We'll
take
a
look
and
yeah
made
a
clean
it
up
and
make
a
pull
request
to
the
main
sounds
good.
F
Yeah,
that's
that's
it
for
me.
G
G
Code
for
a
single
single
time
series,
it
seems
working
relatively
well
yeah,
just
I'm
kind
of
up
to
expanding
that
to
an
entire
like
a
large
area.
So
that's
where
I
need
the
couple
of
like
a
solving
rc
equations
to
constrain
the
problem
with
the
physics
but
anyway.
So
that's
that's!
I'm
making
some
updates,
but
not
very
directly
to
the
impact.