►
From YouTube: SimPEG Meeting October 6th
Description
Weekly SimPEG meeting from October 6th, 2021
A
A
A
If
you
the
other
and
then
along
with
that
right,
we
have
our
seminar
next
week.
I
believe
it's
on
thursday,
again
thursday,
the
timings,
let's
see
thursday
at
2
p.m.
Next,
week.
A
Then
the
other
thing
talk
about
agenda.
As
far
as
alternating
the
meeting
times
we
set
out
that
link
hope
you
guys.
Hopefully,
everyone
responded
first
and
third
month
mondays.
First
third
wednesdays
will
be
morning
time,
we'll
just
settle
into
that,
because
this
is
the
first
wednesday.
So
this
is
the
morning
time.
So,
let's
just
settle
right
into
it.
A
B
It
looks
like
from
everyone:
who's
responded,
3
p.m,
pacific
works
so
maybe
just
kind
of
nods
or
disagreements
from
folks.
Here
I
know
dieter,
that's
probably
way
too
late
for
for
you,
but
for
most
of
the
folks
in
north
america
does
that
still
does
that
still
work.
C
Hey
joe,
can
you
repeat
when
there's
a
seminar
you
said
on
the
14th
of
october?
Is
that
what
you
said
yeah.
C
Okay
sounds
good.
There's
a
bcgs
talk.
I
guess
most
people
already
seen
it,
but
it's
gonna
be
what
the
hell
look
at
his
name:
empty,
empty
nut,
guy,
yeah
yeah,
so
he's
giving
it
at
one
to
two.
So
that's,
okay,
somehow
good
fyi.
A
See
what
else
we
had
so
lpg
wrapped
up
the
last
week,
everyone
well
for
people
who
had
to
have
some
responsibilities.
There.
A
A
A
B
A
Here,
but
the
next
step
will
be
updating
those.
Oh,
he
had
a
bunch
of
tutorials
that
he
had
written
about
going
through
the
finite
volume.
Just
finite
volume
implementations
in
there
so
we'll
have
to
update
those
with
relevant
information.
Considering
all
the
information
that's
now
in
the
doc
strings
themselves,
it's
pretty
heavy
like
the
doc
strings,
are
much
more
explicit
than
they
used
to
be
so
it'll
be
worth
taking
a
pass
back
at
that
updating
and
maybe
simplifying
them
a
little
bit
more,
which
will
be
good.
A
Also,
the
touch
my
tetrahedral
meshes:
I've
got
all
the
vector
interpolations
working
tests
passing
on
them
works.
Well,
so
the
only
thing
that's
missing
right
now.
I
believe:
there's
a
few
random
averaging
matrices
and
like
a
cell
stencil,
a
cell
gradient
stencil
to
make
sure
it
works
well
with
conversions,
but
other
than
that
it
looks
pretty
decent
to
me,
so
it'll
be
in
there
shortly.
E
Yeah
sure
so
I'm
gonna
start
by
like
thanks,
dom
and
mike
for
the
feedbacks
we
on
the
pgi
branch
and
the
pr
the
more
at
the
at
the
moment.
E
E
What
might
need
a
longer
discussion
either
in
this
meeting
or
maybe
at
another
occasion,
it's
more
a
question
of
of
of
a
timeline
on
how
we
organize
things
because,
like
you
want,
you
are
on
your
side,
modifying
like
a
regularization
but
also
the
sensitivity
weight
directive
to
include
a
volume,
normalization
and
stuff
like
that.
E
So
so
I
I
do
admit
that
at
some
point
I
was
not
sure
on
like
with
how
to
like,
we
should
probably
coordinate
so
that,
like
we
don't
have
like
this
pr
comes
in
and
assumes
this
change
is
gonna
be
making.
This
change
has
not
been
made
yet
or,
like
you,
make
a
change,
and
then
we
need
to
revisit
again
or
like,
or
maybe
that's
the
best
way
to
do
it
like.
C
Yeah,
but
that
was
my
hope,
was
that
you
would
be
just
be
able
to
basically
slash
bgi
and
you
know,
get
rid
of
all
duplication
and
just
keep
one,
because
then,
when
I'm
at
the
time
to
change
regularization,
then
I'm
gonna
have
less
to
look
at
right
right
now.
E
C
Think
we
are
always
going
to
assume
that
volumes.
That's
you
know,
weights
are
don't
have
volumes
and
then
volumes
are
always
going
to
be
multiplied
in
the
regularization
all
right.
So
in
your
in
your
pgi,
you
should
have
a
cell
volumes
being
multiplied
somewhere
on
your
cell
weights
and
then
that's
that's
just
going
to
be
all
up.
All
you
know
implemented
the
same
way.
E
Yeah,
so
that's
what
I
was
thinking
like
if
I
do
that
at
the
moment
on
my
pr,
then
that
means
that
I'm
gonna,
like
the
regular,
is
on
ipr.
Only
like
just
looking
at
my
branch.
It's
gonna
assume
that,
like
the
realization,
gonna
assume
there
is
like
cell
weights,
but
the
sensitivity
weighting
will
not
be
matching
that
assumption
and
then
there
will
be
a
discrepancy
in
the
pr
itself.
E
C
B
C
E
Yeah,
just
thinking
also
that
dji
improvement
branch
is
also
fixing
the
issue
of
the
incompatibility
with
psychic
learn
1.0.
So
it's
and
we
we
had
some
exam.
Some
people
messaging
us
some
slack
about
that.
So
might
be
one
that's
like.
Maybe
I
can
also
just
take
that
six
for
sim
circuit
to
learn
1.0
out
and
just
having
a
small
pr.
Just
on
that.
I
think
that's
probably
easy
to
do
for
a.
C
C
E
Yeah
that
sounds
good
perfect,
just
wanted
to
mention
that
otherwise
I
saw
I
was
looking
at
the
meeting
note
from
last
week
that
I
could
not
attend
because
of
scg,
and
I
saw
that
devon
writes
the
issue
of
like
having
the
simple
regularization
fail
with
the
when
all
your
active
cells
in
the
mesh.
E
Well,
when
all
you're
selling
your
in
your
mesh
are
active,
and
I
actually
just
ran
into
that-
I
was
reviving
some
of
that
of
the
notebook
we
wrote
for
the
cam
project
with
joe,
and
I
mean
in
that
in
that
project,
like
we
had
no
topography,
so
thinking
were
fully
active
and
the
simple
realization
was
failing,
because
of
that
so
might
be
some
and
it
should
be
a
recent.
E
It's
probably
a
reason,
but
my
point
is
more
like
it's
probably
a
recent
bug
because
those
notebooks
were
working
before
so
there
is
something
that
happened
in
between
that
we
haven't
catch
like
I
can
prepare,
maybe
a
small
example
on
that.
That's
something
easy
to
do
and
last
point.
I
wanted
to
do
a
refresh
on
time.
E
Domain
em
recent
recently
for
like
for
myself
and
like
I
haven't,
run
some
of
this
inversion
for
a
while,
and
I
was
just
running
in
a
in
a
couple
of
bottlenecks
with
a
simple
just
because
I
haven't
been
working
with
em
for
a
while
and
yeah.
Like
I
realized,
we
don't
have
the
apparent
resistivity
utils.
I
would.
I
found
one
in
pi
gimli,
so
I
put
that
for
reference,
but
that
might
be
a
good
util
to
to
add.
I
also
found
the
paper
where
those
formulas
are
from,
so
I
can
share
that
as
well.
E
So
like
maybe
something
that
to
put
on
the
radar
to
put
in
and
yeah,
I
was
wondering
about
the
state
of
like
gm
1d
and
inversion
tutorial,
because
yeah
simply
m1d
does
like
the
like.
The
the
separate
repository
doesn't
seem
to
be
maintained
too
much,
so
I
got
it
working,
but
by
you
having
an
environment
with
old
version
of
discretize
and
sim
peg,
and
so
on.
Like
back
when
I
was
doing
the
book
printing
example
with
the
for
my
own
papers
and
yeah
like
and
right
now.
E
The
only
example
we
have
on
the
simpac
website
are
like
in
the
gallery
with
those
different
functions,
so
kind
of
not
a
nice
notebook
to
download,
and
there
is
no
inversion
tutorial
on
it.
So
I'm
just
was
just
wondering
I
didn't
check
the
simple
game
on
the
branch
right
now,
but
maybe
just
put
to
put
that
again
on
the
plate
of
what's
happening
on
that
side
of
things
so
yeah.
So
that's
what
I've
been
up
to.
A
A
B
Unmuted
here
thanks
joe
the
one
point
I'll
raise
well
the
newsletter.
Please
contribute
I'd
like
to
get
that
out
today
or
first
thing
tomorrow
morning,
so
that
there's
at
least
a
week
lead
time
on
the
seminar
and
especially
if
you
have
screenshots
of
things
to
include
it's
always
nice
to
make
it
somewhat
visual.
B
So
if
you've
got
a
screenshot
to
go
with
your
bullet
point,
that
would
be
great,
then
the
other
thing
and
dieter
feel
free
to
chime
in
here
too
tuned
in
with
a
few
folks
here.
So
I
know
joey
were
there
as
well
to
the
subsurface
meeting
on
monday
morning
this
week,
I'm
talking
about
sort
of
compatibility
between
packages
in
the
open
source,
geo
ecosystem,
subsurface
ecosystem.
B
So
one
of
the
outcomes
was
leo,
actually
wrote
a
small
repository
that
actually
just
tests
that
you
know
we
can
install
a
number
of
packages
in
the
same
place,
so
he's
got
a
simple
script:
making
sure
that
simpeg
fatiendo
and
a
few
others
can
all
install
in
the
same
environment.
B
So
that's
a
good
step
forward
is
that
having
the
software
in
one
place
is
a
good
thing,
but
the
other
thing
that
kind
of
came
out
of
that
is
that
we,
it
would
be
great
to
start
collecting
some
examples
of
workflows
and
use
cases
where
people
have
used
multiple
packages
in
the
ecosystem
so
andrea
she
has
they.
She
gave
a
great
presentation
at
the
geologic
modeling
scg
workshop
on
integrating
gempai
and
pai
gimli
super
cool
talk
highly
recommend.
B
Well,
I
guess
they'll
they
weren't
recorded,
but
there
should
be,
hopefully
a
paper
coming
out
at
some
point
with
with
their
work,
and
we
can
take
a
look
at
that.
But
if
you've
got
other
examples
like
that,
make
note
of
them
and
just
chime
in
on
the
I
don't
know
dieter.
B
If
you
have
recommendations
right
now
on
where
to
chime
in,
but
just
over
on
this
one
slack,
maybe
on
the
dev
summit
channel
just
to
start
to
collect
some
of
these
and
we
can
figure
out
what
we're
actually
going
to
do
with
them.
D
Yeah,
basically
chiming
in
wherever
you
feel
like.
I
think
you
had
three
outcome.
One
was
these
examples.
You
say:
if
you
have
cool
examples
that
combine
different
packages,
then
please
submit
them.
I'm
not
sure
where
to
that's
what
we
also
have
to
define
by
the
end
the
channel.
Just
right,
you
have
a
cool
example.
D
The
other
one
was
this
test
smoke
test.
I
think
they
called
them
that
we,
you
can
install
as
many
as
possible
different
packages,
and
I
think
that
the
crucial
part
there
is
to
have
as
few
pinned
dependencies,
but
I
think
simp
peg
is
pretty
good
on
that
front
right.
You
don't
really
have
pinned
dependencies
and
what
was
the
third
one.
D
I
can't
but
yeah
if
you're
interested
in
the
whole
stack
in
in
the
dev
summit.
There
is
all
the
info
and
it
would
be
cool
if
we
get
it
pushing
a
bit.
It
might
also
be
something
if
you
have
this
fifth
wednesday,
occasionally,
where
we
hack
off
that
we
maybe
focus
on
on
some
things.
There.
A
Yeah,
unfortunately,
you
know
yeah
senpai
doesn't
have
too
many
independencies
like
that.
I
mean
you
know:
discretize,
we've
gotten
it
down
pretty
well
sim
craig.
I
think
the
biggest
thing
trying
to
install
now
is
the
dependencies
on
the
older
version
of
scikit-learn
from
pgi,
but
that
should
be
fixed
right
now.
Six
pretty
quickly.
A
Any
other
quick
reports
to
look
like
to
for
anything
else
that
you
would
like
to
chat
about.
G
I
just
had
a
quick
question.
I
don't
know
if
anybody
has
ever
like
written
a
script
or
played
around
with
trying
to
take
like
a
diagonal
section
through
a
model
or
just
one.
That's
not
orthogonal.
I
don't
know
if
anybody's
encountered
that
or
played
around
with
it.
A
The
easiest
way
to
kind
of
do
it
would
be
to
generate
like
a
generate
a
bunch
of
points
on
whatever
plane,
you're
doing
and
then
just
find
whatever
cell
it's
nearest
to
and
get
that
value.
F
D
A
F
Actually
provides
the
section
so
yeah,
I
think
it's
probably.
G
F
G
C
Yeah,
if
it's
on
tree
mesh,
it's
pretty
easy
right
mike
you
just
use
the
in
cell
fine
index
and
cell.
I
can
never
remember
that.
What's
what.
D
C
I
was
going
down
the
dc,
the
dc
rabbit
hole,
and
then
you
saw
joe,
I
ping
you
and
thanks
for
that
ish
for
appointing
that
issue.
So
I
yeah
so
I'll
test
on
my
side
and
then,
if
it
works
for
a
while,
then
I'll
I'll
create
a
pr
for
it.
But
that's
a
heads
up
for
for
john
right,
so
the
the
currently
the
draped
topo
dc
right,
drape
electrolyte
to
toppo
it
moves.
It
moves
the
xy
location
of
your
electrodes
to
go
to
cell
center.
C
So,
instead
of
just
dropping
the
electrode
down,
it
actually
moves
the
xy
to
cell
center
so
that
that
can
be
quite
you
know,
induced
quite
an
error
in
your
simulation.
C
A
H
A
It'll
only
make
a
difference
on
the
cell
center.
It
won't
make
a
difference
on
the
cell
center
dc
and
system
for
treeless,
because
that
interpolation
is
like
nearest
anyway.
So
it's
not
going
to
make
a
difference.
F
C
And
it
does
yeah
I'm
on
model
and
for
the
reason
that's
like
you
just
mentioned,
it
used
to
not
work
on
shimmer.
So
I
won't
know,
though,
and
I
was
yeah-
it
was
getting
me
some
weird,
weird
behaviors
there
so
yeah
did
you
there's
really
no
reason
why
the
heck
did
we
decide
to
move
the
xy?
I
really
don't
understand
like.
A
At
some
point,
but
yeah
anyway
did
you
do
you
get
it?
Did
you
get
the
sketch
of
what
I
was
imagining
like
how
to
do
this
well.
C
A
H
Stuff,
I'm
just
well
well
the
first
round
of
it.
I've
just
got
to
get
the
code
cup
stuff
done
and
then
it's
passing
all
the
other
tests.
I
just
got
to
get
some
more
code
coverage
and
I'll
be
able
to
pull
in
the
first
part,
the
tiled
mt
stuff
that'll
still
be
a
while.
I
still
have
to
work
on
the
past
passing
tests
for
the
tile
contribution
to
that
one.
H
So
yeah
the
speed
will
take
some
time,
but
the
new
implementation
of
the
mt
being
worked
in
with
the
frequency
domain
yeah.
I
should
be
able
to
get
to
that
soon.
I
just
I've
been
busy
with
a
couple
projects.
The
last
little
bit.
D
H
D
D
A
B
I
think
that's
something
we
should
start
sort
of
collectively,
but
let's
we
can
chat
about
sort
of
a
structure
for
that
of
kind
of
maybe
thinking
you
know
like
about
a
couple
different
timelines.
So
you
know
what
is
the
next
release?
So
what
are
we
working
on
in
the
next
couple
of
months?
What
is
the
next
year
and
so
forth,
but
maybe
to
actually
kind
of
kick
that
off.
B
We
could
even
get
a
quick
document
going
for
now
of
like
what
our
folks
priorities
here
for
the
next
month
couple
months
and
just
for
for
people
who
are
sort
of
actively
working
on
different
pieces
of
synthetic.
If
you're
not
sure,
that's
like
totally
fine,
you
don't
don't
feel
obliged
to
sort
of
put
something
down
if,
if
you
don't
have
items
but
if
even
from
everyone
here,
we
had
a
good
sense
of
like
here
my
three
priorities
for
the
next
two
months.
That
would
be
that
would
be
useful.
D
C
Yeah
speaking
of
priorities,
so
tebow
we're
just
talking
about
small
pr
to
fix
psychic,
learn
right
so
right
now,
looking
at
the
pr's
that
devon
just
created
and
they
all
they're
all
failing
on
it
because
of
it.
So
if
we
could
do
this
in
priority,
that
would
be
that'd
be
fantastic
because
we
can't
merge
anything
until
until
this
is
done.
E
D
A
B
A
D
A
F
This
is
something
that
I
haven't
really
thought
about,
and
somehow
I
was
acting
as
to
give
a
talk
about
the
groundwater
component
of
the
sustainable
city.
The
conference
was
actually
the
the
workshop
was
quite
interesting.
A
lot
of
them
were
like
related
seismic,
so
I
think
I
can
clearly
see
this
passive
seismic
monitoring
network.
They
are
really
expanding
the
realm
to
kind
of
near
surface
geophysics,
not
necessarily
petroleum
focus.
F
So
it
was
interesting
to
see,
and
then
people
are
quite
excited
about
using
electromagnetics
and
and
some
of
the
questions
that
we've
got
I've
got
was.
Can
we
actually
used
in
the
urban
setup,
and
I
said
no-
we
can't
really
use
it,
but
that
seems
an
open
question
and
I
don't
know
like
I
was
just
kind
of
I
thought
like
thinking
about
a
sustainable
city
in
how
you
use
the
geophysics.
F
That
seems
like
a
pretty
big
chunk
of
pie
and
people
are
really
interested
in
that
kind
of
avenue.
So
one
avenue
other
avenues
actually
critical
earth
minerals,
because
to
sustain
the
city
development,
you
actually
do
need
to
have
a
lot
of
metals
and
the
rare
earth,
minerals
and
types
of
stuff.
So
that
was
another
topic
in
the
geothermal
development,
in
kind
of
for
generating
small,
like
powers
for
homes,
and
some
of
them
were
using
geophysics
as
well.
F
So
I
don't
know
like
not
kind
of
typical
topic
that
we
deal
with,
but
I
thought
that's
quite
quite
interesting
and.
F
Yeah
urban
mining
is
that
I
don't
know
how,
because
the
I
think
that
I
think
the
major
idea
is
that
geochemistry,
because,
like
the
idea
is,
if
you
like,
if
you
got
old
cell
phone,
then
how
do
you
kind
of
reuse
that
cell
phone
to
kind
of
you
can
actually
extract
the
goals
from
cell
phone?
Or
something
like
that?
F
I
thought
that's
the
general
idea
of
urban,
but
I
think
the
city
has
lots
of
other
problems
like
sinkholes,
like
that
types
of
things,
and
they
are
generally
thinking
about
kind
of
having
a
monitoring
system
in
the
city
and
also
want
to
monitor
the
subsurface
part.
So
I
think
there
is
a
lot
of
role.
I
guess,
but
we
just
don't
know
how
to
do
it,
but
it
seems
an
interesting
avenue.
E
I've
just
done
that
and
also,
I
think
I
remember
a
talk
either
by
dequon
or
one
of
his
students,
and
he
did
something
with
like
a
monitoring,
I
believe,
for
sinkhole
with
a
dc
system
that
would
just
be
like
hooked
to
the
power
grid
and
we'll
just
take
data
regularly.
F
Yeah,
nothing,
nothing
particular
speaking
of
that
that
airborne
like
a
syntax,
1d
stuff.
I
am
to
come
back
and
want
to
get
this
done.
Given
there
are
actually
a
lot
quite
a
bit
of
excitement
in
here.
People
actually
started
starting
to
use
actually
that
that
branch,
joe
some
of
them,
actually
started
to
use
that
without
too
much
instructions,
they're
jumping
in
and
figuring
out
by
themselves,
so
it'll
be
better
if
it's
working
for
them,
I
think
so
I
think
in
general.
F
I
think
it's
mostly
working,
so
it's
actually
pretty
good,
so
probably
may
not
need
to
do
it.
They
may
do
it
by
themselves
anyway.
So
it's
it's
exciting
time.
Actually
they
are
flying
the
airborne
survey
in
california,
so
they
started
actually
just
started
like
a
couple
months
ago
and
they
finished
a
couple
thousand
kilometer
lines
so
yeah
I
and
data
that
data.
I
think
the
they
are
planning
to
open
the
data
for
every
six
like
six
months
after
the
acquisition.
F
F
A
Well
exciting,
hopefully
the
data
will
be
able
to
play
around
with
it
soon.
If
there's
anything
else,.