►
From YouTube: SimPEG Meeting August 25th
Description
Weekly SimPEG meeting from August 25th, 2021
A
C
D
D
A
Excellent,
yeah
john
there's,
probably
a
few
faces
on
here
that
you
don't
you
haven't,
met
before
at
least
I'm
not
sure.
Have
you
been
to
one
of
the
simple
meetings,
yet
I
can't
remember,
and
now
this
is
the
first
one.
Okay,
that's
what
I
thought
you
probably
don't
recognize
ben
mike
or
shalom,
and
because
you've
met
david
and
myself
before
ben
and
mike.
Do
you
mind
just
give
me
a
quick,
yeah
sure
yeah?
I
joined
the
group
fairly
recently
as
well.
E
F
Hi
all
I'm
soggy,
I
used
to
be
in
ubc.
I
got
phd
with
from
from
dunk
and
used
to
work
with
mike
lindsey,
and
I
am
at
stanford
now
as
a
as
a
postdoctoral
fellow
and
mostly
working
on
water
problems,
I
kind
of
switch
it
to
minerals
from
water,
but
I
still
interested
mining.
Actually,
I
just
like.
We
just
presented
some
of
the
mining
work
about
the
en
inversion
in
the
conference
anyway.
So
that's
I'm
great
to
see
you
all.
A
Yeah,
thank
you
as
far
as
an
agenda
for
the
meeting.
I
don't
particularly
have
anything
so
if
I
just
give
some
quick
reports
from
people
and
then
move
on
from
there,
I
I
see
tebow
had
added
something
but
he's
not
on
there.
Yet
I
was
talking
with
him
yesterday
about
if
he
was
trying
to
do
some
pde
simulations
of
the
magnetics
problem,
and
we
realized
that.
Oh,
that
also
is
that
boundary
condition
it's
using
boundary
condition
codes
in
there.
F
Yeah,
I
think
I
don't
know,
I
think
that
pde
version
has
been.
We
haven't
touched
it
for
several
three
of
them
or
something
like
that
so
yeah.
I
would
be
surprised
that
would
work
and
yeah
good
good
luck
with
tebow
and
I'm
happy
to
help,
and
I
I
think
I've
thought
about
that
quite
a
bit
of
and
you
can
actually
do
a
linear
version
of
pde
so
rather
than
so
anyway,
like
I
have
thought
about
that
problem
so
need.
A
Anyway,
I
like
to
move
on
so
I've
messed
up
that
still
just
like
kind
of
I'm
putting
too
much
time
into
their
tetrahedron
acid
meshes
I've
just
kind
of
been
picking
at
it.
It's
still
coming
together
really
quickly,
though,
like
even
what
time
I
do
it
together.
It
just
kind
of
falls
falls
together
very
nicely.
B
A
I
I
think
we've
talked
about
it
before,
but
I
think
at
this
point
having
like
generators,
is
a
little
bit
outside
the
scope
of
discretize.
At
this
point,
like
there's
a
lot
of
great
tetrahedral
mesh
generators
out
there
already,
we
should
at
least
just
be
able
to
interface
with
them
and
use
them
before
we
start
trying
to
like
implement
our
own
stuff
to
create
things.
G
One
of
the
things
that
we
did
with
a
sim
peg:
we
had
that
3d
pseudosection
plot
that
required
plotly
and
it
was
a
utility,
but
it
would
import
that
functionality.
If
you
had
that
package
and
I
think
we
could
do
a
tetrahedral
mesh
utilities
where
it
will
import,
you
have
those
available.
If
you
have
that
the
right
package,
so
we
could
just
kind
of
make,
I
don't
know
you
call
it
a
wrapper
or
whatever
it
is.
We
could
write
that
that
kind
of.
A
F
A
A
Okay,
never
mind,
that's
that's
pretty
pretty
cool
wow,
that's
exciting
yeah!
It
was,
I
don't
know
there
was
a
lot
of
stuff
like
I
was.
I
was
building
up
kind
of
like
the
structure
behind
it.
Like
I
was
thinking
about.
Okay.
What
do
I
need
to
keep
track
of?
Well,
I
need
to
keep
track
of
like
for
every
cell.
I
need
to
keep
track
of
it's
like
the
nodes
on
it.
That's
like
the
general
structure.
Is
you
give
it
like
the
the
points?
A
So
you
give
it
a
list
of
nodes
and
you
can
give
them
their
connectivity,
like
you
define
like
the
nodes
that
are
in
every
cell
and
then
it
goes
through
and
builds
everything
from
that.
So
just
like!
Okay
from
that,
we
can
figure
out
the
faces
that
are
on
each
thing
and
the
nodes
like
which
nodes
are
on
each
face
and
then
which
edges
are
in
each
on
each
cell
and
which
edges
are
each
face.
F
A
F
A
F
Yeah,
what
was
it
interesting?
What
like
what
I
learned
when,
like
lindsay
and
peter
worked
on
that
paper
about
cscm,
the
the
chronology
match
actually
have
a
bit
of
problem
like
that
expansion
rate
of
two
it's
actually
just
too
much
to
handle
like
like
a
the
general
like
a
marine
csm
types
of
problems,
so
which
actually
could
be
a
problem
when
we
actually
want
to.
A
G
Yeah,
I
guess
one
thing
I
I
want
to
bring
up
is
is
pushing
the
discretize
api
across
the
finish
line.
I
think
you've
done
your
final
review.
I've
done
my
final
review.
The
only
thing
really
left
is
to
cache
the
kind
of
expanded
tutorials
section,
because
right
now
that's
we're
not
ready
to
release
it
and
it's
right
now
it's
showing
up
in
in
the
build,
so
I
think
doing
that
part.
We
need
to
do
that
or
maybe
what
part
are
you
referring
to?
G
The
I'm
just
wondering
do
we
make?
It
could
be
something
that
was
in
a
get
ignore
and
then
maybe
not
in
a
gig
get
ignored
on
another
branch
and
why
it's
showing
up?
So
I
guess,
if
you're
not
seeing
it
as
any
problems,
then
it
should
be
fine.
I
was
wondering
if
the
latest
tutorials,
that
we
had
put
that
we
were
developing
for
discretizes
part
of
the
branch
being
merged
or
whether
or
not
that
that's
actually
not
a
problem.
I
don't
think
so.
A
G
A
G
A
G
G
A
G
Yeah,
so
that's
it's
a
useful
little
function,
but
it
definitely
makes
things
a
little
more
interesting
in
terms
of
compatibility
and
kind
of
the
environment,
you're
building
it
in
so
yeah.
It's
something
to
consider
so
that
that
sounds
like
that's
the
only
part.
That's
really
left
to
do
on
that
project.
Yeah,
just
kind
of.
A
G
G
D
B
A
G
A
A
G
G
A
Next
to
each
other
yeah
that
was
kind
of
that's
the
thought
of
this
one
and
then
it
obviously
looks
like
a
d,
but
that's
kind
of
like
at
least
some
of
the
ones
that
we're
working
on
right
now,
yeah.
I
actually
like
that.
Like
I
said
it,
they
look
noticeably
related
to
synthetic
like
they.
They.
All,
I
think
is,
would
keep
that
color
scheme
we'll
find
a
good
good
basis
for
it.
G
I
started
a
google
doc
and
I
started
sketching
out
a
way
that
we
would
be
publishing
our
code
validations,
so
we'll
drop
that
in
the
chat
so
yeah
I
mean
I
I
started
to
work
around.
I
got
a
bit
of
an
example
that
maybe
I
could
show
you
guys
in
like
five
minutes
on
how
I've
sort
of
organized
this
and
I'll
just
start
with
the
the
mags.
If
you
give
me
screen
share
ability,
I
will
go
for
it.
G
I
don't
know
if
it's
going
to
be
a
website
or
if
we
can
deploy
it
other
ways,
I'm
looking
for
any
input
on
that,
but
probably
want
to
section
it
by
geophysical
method
and
then
each
each
particular
kind
of
geophysical,
experiment
or
each
scenario.
Maybe
it's
like.
I
don't
know
the
em
response
over
a
1d,
layered
earth
or
something
like
that.
G
G
G
So
now
you
know
what
you're
modeling
you
know
which
codes
we're
using
and
then
you'd
have
basically
the
lines
of
a
notebook
that
would
download
any
files
from
as
like
a
the
dot
tar.
We
just
import
things
from
from
an
online
storage
place
and
then
maybe
we
would
plot
something
like
what
what
the
model
is
and
the
survey
geometry
so
give
you
a
picture
of
what
you're
modeling
and
then
like
plot
the
simulated
data
using
both
codes
in
some
fashion,
and
in
this
case
it
was
important
to
actually
have
some
kind
of.
G
What
coding
packages
were
compared
and
then
show
visually
that
comparison,
and
if
somebody
wants
to,
I
mean
all
of
these
simulations,
they
should
be
reproducible.
It's
kind
of
the
point
somebody
could
say:
okay.
Well,
how
did
you
actually
get
that
working
so
then
you
could
go
kind
of
down
a
level
and
have
access
to
notebooks
or
instructions
on
how
to
reproduce
the
results.
G
So,
if
you
are
a
ubc
jiff
consortium,
member,
maybe
you've
purchased
a
license.
Maybe
you
work
for
an
academic
institution
and
got
an
academic
license.
You
would
then
actually
have
the
executable
and
you
would
be
able
to
to
run
both
of
these
and
reproduce
it.
So
that's
sort
of
what
I'm,
what
I'm
thinking
with
this
and
the
way
that
I've
written
this.
D
This
looks
really
nice
devin.
Thank
you.
One
thing
I
would
encourage
you
to
look
at
for
deploying
is
jupiter
book.
I
think
that
might
actually
be
a
natural
fit
for
this,
rather
than
turning
all
of
these
into
dot
pi
files,
because
one
of
the
things
that's
important
for
especially
some
of
the
ones
where
you're
reproducing
examples
from
simpeg,
because
I
suspect
that
they're
actually
going
to
take
a
while
to
run
yeah.
D
G
G
D
Yeah,
take
take
a
look
so
actually
the
jupiter
book
docs
is
built
in
jupiter
book.
So
that's
that's
a
nice
way
to
get
a
tour
of
kind
of
what
how
that
works,
and
so,
yes,
you
can
sort
of
have
the
same
nested
tree
approach.
We
can
combine
markdown
files
with
with
notebooks
and
and
things
like
that
and
then
what's
nice
is
is
like
by
default.
G
D
G
G
Well,
I'm
thinking
well
as
as
lindsey
said,
I
think
we're
gonna
put
it
up
on.
I
guess
jupiter
book
and
I
I
don't
know
if
that's
connected
necessarily
to
simpeg
right
now,
but
it
should
be-
and
I
guess
somewhere
on
the
the
sempeg
website,
we
would
have
a
page
about
code
validations
that
would
bring
you
to
the
jupiter
book.
G
So
it
is,
it
kind
of
serves
a
lot
of
purposes.
One
of
it
is
it
it'll,
validate
simple
code
against
analytic
solutions
and
also
against
other
coding
packages,
namely
for
us
would
be
ubc
gif
at
this
point,
and
you
know
how
fast
does
it
run
performance
stuff
and
already
with
some
of
these
validations
validation
codes,
it's
led
to
improvements
in
performance
and
and
other
things
from
from,
like
the
ubc
gif
code
side
and
I
think
also
in
simpeg.
G
F
F
G
G
I
guess
how
jupiter
book
works
and
and
yeah
do
some
learning
for
that
before
I
have
all
the
answers,
but
it
seems
like
it's
going
to
be
hosted
and
it's
going
to
be,
the
files
will
physically
be
located
in,
I
guess,
a
separate
repository
that
will
be,
I
guess,
connected
or
uploaded
to
jupiter
book
and
from
the
simpeg
website.
You
would
get
linked
to
to
that
stuff.
I
feel
like
that's
the
general
idea,
I'm
getting
thumbs
up
from
lindsay,
so
we're
on
the
right
track.
A
D
Yeah
just
a
couple
quick
updates,
so
the
line
source
implementation,
at
least
on
the
time
domain,
is
ready
for
testing
so
devin
you
mentioned
you
might
be
willing
to
take
a
crack
at
that.
If
you'd
like
to
go
for
it,
that'd
be
that'd,
be
wonderful,
I'll,
be
mostly
offline
for
about
a
week.
So
if
you
run
into
any
roadblocks,
feel
free
to
tag
me-
and
we
can
figure
it
out
next
week,
but
if
you're
willing
to
take
a
crack
at
it,
that
would
be
great
and
we
can.
D
D
The
other
one,
that's
a
quick
update,
that's
mostly
relevant
for
dieter
who's,
not
here,
but
we'll
put
it
in
the
recording
form.
Is
the
casing
simulations
repo
that
I
use
for
my
thesis
stuff,
I'm
trying
to
do
a
bit
of
cleanup
there
get
it
onto
the
latest
simpeg
and
also
help
fix
it
up
a
bit.
It
was
one
of
those.
It
was
a
bunch
of
code
that
I
wrote
when
I
was
learning
and
excited
about
object,
oriented
programming.
D
F
I
D
D
I
F
F
F
This
was
my
first
come
first
time
actually
present
there.
I
think
it's
mostly
like
a
very
yeah,
I'm
not
sure,
what's
like
the
real
identity
of,
but
I
think
it's
mostly
kind
of
like
global
global
geophysics.
Very
deep
structures.
Mt
is
their
main
interest,
so
lindsay
was
a
convener
that
was
main
motivation
to
submit,
but
so
it
was
an
interesting
experience.
F
So
we
submitted
an
abstract
and
we
presented
the
work
that
I
collaborate
with
theater
and
yes,
I
shared
that
poster
in
some
of
the
slack
channels.
So
if
anybody
is
interested,
they
can
take
a
look
yeah,
but
I
think
in
general,
like
they
are
getting
more
interested
in
the
near
surface,
so
application
of
mt
for
the
near
surface.
I
thought
that's
some
of
the
interesting
kind
of
discussion
that
that
we
had
in
that
conversation
and
another
thing
I
just
posted
in
that
curve-
note
so
dieter's
code.
F
Actually,
we
used
slightly
different
definition
of
data
so
rather
than
breaking
that
apart
as
a
real
and
imaginary.
So
we're
directly
inverting
for
complex
value
or
output
and
phase,
so
that
makes
a
slightly
different
derivation
of
of
complex
data
to
either
amplitude
or
phase,
and
I
kind
of
I
derived
and
I
got
the
right
solution,
but
I
actually
don't
know
why
that's
the
right
solution.
So
if,
like
I
put
down,
I
I
wrote
down
the
the
math
and
some
of
the
assumptions
that
I
made,
which
I'm
not
sure.
F
So
if
so,
I'm
going
to
take
a
look
and
actually
let
me
know
why
that's
the
case,
I
I
know
that's
the
true
and
that
actually
works
and
what's
interesting
thing
is
the
derivative
of
the
phase.
Regard
in
regards
to
the
complex
value
is
actually
zero.
So
if
you
think
about
that,
the
gradient
of
the
phase
is
zero,
which
was
a
bit
surprising
to
me.
F
So
there's
actually
no
need
to
compute
that
that,
like
the
jt
back
of
this,
like
of
the
of
the
face,
so
most
of
information
is
like
most
of
the
gradient
directions
coming
from
the
amplitude.
It
sounds
weird,
but
I
think
I'm
right,
because
I
tested
the
order
test
and
then
like
everything
worked.
But
I
don't
know
why
so
there's
some
interest.
Maybe
there's
a
potential
that
I
could
be
wrong.
But
if
somebody
could
take
a
look
and
think
about-
and
that
would
be
an
interesting
discussion
that
we
can
have
later.
F
D
E
A
Oh,
I
think
I
remember
when
I
was
looking
at
it,
so
I
just
when
I
was
referring
to
the
normalization
I
just
felt
like
I
don't
have
any
volume
terms
in
it.
You
could
you
can
still
like
divide
it
by
its
max
value.
If
you
want
that's
fine,
oh
it's
fire
yeah.
That
part
is
fine.
Okay,
there's
no
volume
normalization!
A
I
It's
a
bit
ridiculous
that
we
have
that
that's
simple
and
then,
as
far
as
that's
basically
everything
the
same
thing
except
for
like
a
few
flags
right,
so
I
think
we
should
try
to
simplify
it.
A
lot
like
tknav
is
just
a
matter
of
like
a
scaling
factor.
It
could
be
all
just
flags.
I
think
if
you
want
to
use
you
know
the
the
usual
and
anyway
I'll
I'll
sketch
up
like
a
new
structure,
and
then
we
can
talk
about
it
next
time.
E
F
The
synthetic
looks
pretty
good,
so
yeah
don't
be
too
optimistic,
but
yeah
we're
like
we're
just
ready,
because
we
like
yeah.
The
major
challenge
was
the
the
notations
and
the
the
complex
value
and
they
provide
like
apparent
resistivity
and
phase
so
yeah.
We
had
to
kind
of
come
up
with
better
form
of
data
like
an
amplitude,
so
we're
just
ready
to
run
the
amplitude
csmg
data,
so
yeah
I'll
keep
you
posted.