►
From YouTube: SimPEG meeting Aug 12, 2020
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Perfect
well
joe,
do
you
want
to
start
us
off.
B
Sure
I've
been
kind
of
young
about
getting
coming
up
catching
up
on
more
writing.
This
past
week
played
around
with
the
with
the
geosci
labs,
getting
those
jupiter
like
a
jupiter
hub
going
for
them
using
the
repo
to
docker,
which
worked
out
really
well
like
I've.
Never
I've
never
used
it
before,
and
I've
never
used
docker
before
it
was
very
easy
to
get
going
on
that
to
get
an
enviro,
a
docker
environment
set.
C
B
B
B
We
probably
should
have
taken
a
little
bit
more
meetings,
but
we
have
a
few
links
to
some
things
about.
We've
got
a
document
discussing
the
big
picture.
Now,
that's
in
the
documentation,
just
kind
of
fleshing
that
out
a
little
bit
wider
lindsay.
Do
you
you
have
that
document
right.
B
We're
all
for
doing
that
thing
that
we
can
see
if
we
do
a
short
30
minute
one
after
a
meeting
next
week.
Maybe
people
are
open
for
that
sure.
I
know
we
said
we
kind
of
alternate
them
back
like
every
other
week
or
so
we
talked
about
doing
that,
so
people
are
free
for
30
minutes.
Afterwards.
Next
meeting
we
have
a
nice
another.
A
Cool
thanks
joe,
I
see
john
next
on
my
screen.
E
Yeah
still
playing
with
the
mumps
there,
I've
pretty
much
got
it
all
the
way
done
and
implemented
in
a
sense,
I've
just
made
like
just
a
cli
or
just
a.
What
do
you
call
it
like
a
functional
method
as
part
of
the
simulation
right
now
in
the
dc
to
try
it
out
yeah,
and
I
was
just
using
the
test
model,
the
one
that
you
guys
used
in
your
tests
and
everything
was
working,
and
then
I
threw
some
a
data
set
at
it
and
it's
just
sitting
and
spinning
in
the
solve.
E
Now
I
don't
know
what
I
did
different
like
it's
all
set
up
the
same:
I've
even
calculated
the
fields
for
a
bunch
of
other
projects
and
it
seemed
to
work,
and
then
this
data
is
just
kind
of
throwing
me
into
the
weeds
again
so
just
kind
of
picking
away
at
that,
and
it's
pretty
much
ready.
I
just
need
to
confirm
that
it's
going
to
work
every
time,
kind
of
thing
and
then
yeah
maybe
sit
down
and
start
talking
about
how
to
maybe
add
it
to
the
solver
utils.
E
I
think
that's
practically
wrapping
everything
yeah.
I
just
have
to
sit
down
and
go
into
it.
I've
been
looking
through
it
a
little
bit
to
see
how
I
can
put
it
in
there
so
that
we
don't
really
have
to
change
much
because
yeah,
the
having
it
as
just
added
to
the
simulation
class
isn't
really
efficient.
It's
because
you'll
have
to
add
it
to
every
other
simulation
class
or
just
put
it
up
on
the
high
one.
So
yeah
I'll
try
to
get
it
integrated
into
the
solver
utils.
Next
and.
B
You're
johnny's
still
just
using
the
mumps
or
pie,
monks.
B
B
E
B
So
that
one
still
only
has
the
the
floating
point
or
the
double
floating
point
operations,
there's
no
complex
numbers
in
there.
Yet.
E
Yeah,
so
I
built
both
of
them.
That's
why
I
said
I
got
like
a
little
bit
of
a
hybrid
of
pi
months,
because
I've
also
incorporated
the
the
complex
as
well
right
now,
I'm
just
testing
the
dc
yeah
and
that's
where
I
kind
of
came
into
the
problem.
I
didn't
test
this
far
yet
on
the
z
mumps.
I
guess
they
call
it
for
the
complex
stuff
for
frequency
domain.
E
I'm
just
trying
to
hook
them
in
to
make
sure
that
they
can
work
yeah.
Just
at
that
point
now,
yeah
though
it
handles
the
multiple
right-hand
sides
really
well,
that
was
actually
another
thing,
so
they
can
ex
you
can
exploit.
If
the
right
hand
side
is
sparse,
is
that
going
to
be
the
case
at
all,
or
is
it
always
going
to
be
a
dense
matrix.
E
E
Okay,
so
we
should
be
able
to
get
some
speed
ups
again
right
now.
I
just
have
it
as
a
dense
matrix,
but
if
we
can
yeah
do
it
as
a
sparse,
they
say
that
it'll
yeah
it
helps
again
but
yeah.
When
we
get
that
when
I
get
there
anyways
I'll,
let
you
know
and
yeah
hopefully
lindsay
I'll,
have
something
for
you
to
test
here
soon
I
did
lose
the
big
computer
here
for
a
bit.
E
E
B
It's
fortunately
like
one
like
the
tool,
mod
solvers
like
it's.
I
don't
think
they
take
advantage
of
mpi
processes,
but
I
think
it's
a
little
better
openmp
and,
like
thread
parallelism.
E
E
E
B
Because
it
can
do
essentially
all
of
our
applications,
at
least
I
understand
they're
all
symmetric
matrices.
I
I'm
not
quite
sure
where
it's
and
once
the
metric
makes
shifts
anywhere.
E
Right:
okay,
yeah.
Let
me
take
a
look
at
that
today,
I'll
not
put
that
off
again.
A
Than
that
yeah.
F
Finally,
have
a
date
for
the
defense:
it
was
a
long
run
to
get
everybody
everybody's,
but
that's
gonna,
be
on
september,
the
september,
the
17th.
It's
like
it's
like
a
thursday
morning.
F
And
yeah
like
so
because
it
got
a
bit
delayed
compared
to
what
we
want.
F
I
was
thinking
maybe,
instead
of
doing
the
the
practice
like
almost
a
month
early
because
I
was
I
was
asking
for,
if
we
can
do
it
next
week,
maybe
the
week
aft,
maybe
the
26th
or
even
the
second
of
september
might
be
better
just
to
have
it
closer
to
the
actual
defense
sure,
just
other
than
that
I've
been
going
back
to
to
try
to
merge
my
branch
with
simulation.
F
F
So
whenever,
when,
when
I've
investigated
that
more,
maybe
I
will
ping
you
soggy
and
send
you
like
a
script
to
see,
try
to
pinpoint
where
the
difference
like
yeah,
it's
a
bit
of
a
it's
a
bit
long
to
to
do
like
my
branch,
like
I
was
keeping
it
up
to
date
with
simplex
so
far,
which
was
good,
but
there
was
a
lot
of
change
in
my
branch
and
there
was
also
a
lot
of
branch
in
simulation
so
and
because
the
merging
is
not
one-on-one
because
of
the
of
the
name,
change
of
the
file
etc
and
the
conventions
there
is
a
lot
of
conflicts,
and
sometimes
I
just
have
to
do
things
by
hand.
F
I
have,
I
have
all
of
my
example
running
running
again
with
simulation,
so
there
is
it's
a
bit
of
a
lot
of
free
write
but
like
that
will
be
also
a
good
test
to
see
if
I
can
get
the,
maybe
not
the
exact
same
thing,
but
something
that's
close
enough
to
say
that
it's
working
and
that
it's
just
like
fiddles
parameters
in
the
background
that
are
changed
slightly
changing
things
but
yeah,
I
at
least
have
good
archive
of
all
the
examples.
So
so
that's
so
that's
that
will
be
so
the
bench.
C
F
C
No,
I
don't
have
anything
new.
One
thing
that
I
would
like
to
ask
is:
we
need
to
submit
our
list
of
presenters
and
themes
for
scg
by
friday.
I
think
so.
If
anyone
else
wants
to
like
I
don't
know,
if
john
you
wanted
to
do
a
part
of
it,
because
right
now,
I
I'm
giving
two
of
this
in
the
three
sections
would
probably
be
nice.
If
someone
else
would,
you
know,
would
present
something
else.
E
How,
how
long
is
it
supposed
to
be
sorry
or
like.
C
E
E
C
If
you,
if
you'd
like
yeah,
it's
just,
I
guess
an
opportunity
for
for
you
guys
to
show
what
you
have
right.
Yeah.
E
That
would
be
probably
the
thing
I
could
spend
most
time
on
kind
of
thing
and
easier
to
present
yeah
sure
goodbye,
awesome,
all
right,
yeah.
Let
me
get
back
to
you
I'll,
just
talk
to
glenn
there
and
just
yeah
I'll
get
back
to
you
this
afternoon.
C
Sounds
good
yeah,
I'm
pretty
sure
glenn
was
already
on
the
because
he's
part
of
the
yeah.
E
Yeah
the
mining
thing
right
yeah.
He
was
telling
me
that
everybody
was
things
were
starting
to
roll
again,
because
everything
was
going
online,
so
yeah,
I'm
sure,
it'll
be
all
good.
C
C
Well
running
on
that
excel
spreadsheet,
but
I
can
I
can
convert
it
to
a
google
sheet
yeah
I'll
I'll
forward.
It
now.
G
Yeah,
it
seems
actually
pretty
hard
to
make
a
good
match
with
the
analytics
there
are.
There
are
quite
a
few
things
going
on,
so
it
seems
like
a
good
opportunity
to
I
don't
know
like
it
makes
kind
of
the
the
hardship
is.
You
can
generate
the
audrey
mesh,
but
you
need
to
do
quite
a
few
things
to
make
sure
that
you
got
good
accuracy.
G
G
It's
it's
very
hard
to
get
good
accuracy
at
the
far
offset,
and
so
you
probably
define
where
you
need
to
discretize
more
to
make
sure
you
get
good
accuracy
at
the
far
offset,
and
usually
we
have
a
worse
accuracy
at
the
near
offset,
but
the
mirror
offset
accuracy
is
pretty
good,
even
if
you
use
pretty
poor
mesh,
which
I
was
a
bit
surprised
anyway,
so
I
think
a
little
bit
of
fun
with
that
other
than
that
yeah.
G
G
H
Yeah,
I
guess
I
can
sort
of
continue
on
that.
One
of
the
things
I've
been
doing,
soggy
is
going
through
all
of
the
tests
and
I've
been
updating
the
tests
to
be
running
with
a
simulation,
so
I
was
able
to
get
started
on
it,
but
I
haven't
been
able
to
continue
on
for
the
past
a
little
bit
because
I've
been
working
on
other
things,
but
that's
sort
of
where
I'm
putting
most
of
my
time.
H
H
No,
not
really
I've
just
been
finishing
off
some
other
stuff
with
dc
octree,
but
I
guess
I'm
seeing
some
things
that
can
ultimately
be
added
to
simpeg
when
we're
done.
This
we've
been
playing
around
with
this
sort
of
sensitivity,
waiting
for
the
dc
octree
problem
and
I've
been
playing
around
with
like
a
really
basic
interpolation
right
now,
it
seems
like
we
don't
have
kind
of
a
satisfactory
way
of
just
going
from
any
mesh
to
any
other
mesh
and
joe.
C
George,
just
posted
the
it's
already
on,
though
yeah
it's
a
match
to
mesh.
It
works
with
all
measures,
we're.
H
It's
nothing!
It's
nothing
too,
fancy
one
thing
that
I
noticed
and
it
came
up
when
I
was
doing
the
inverting
a
set
of
2d
dcip
lines,
and
so
then
I
end
up
getting
a
bunch
of
slices
2d
models
and
then
we're
going
to
go
and
interpolate
that
on
onto
a
3d
mesh
of
some
kind,
and
so
we
really
only
have
kind
of
an
understanding
of
generally
what
the
model
does
immediately
within
the
the
data
points.
H
And
if
I
do
something
like
a
nearest
neighbor,
if
there's
some
conductor
or
some
anomalous
body,
that's
sitting
near
the
edge
of
it,
it's
going
to
create
this
giant
conductor
or
or
structure.
That's
going
to
impact
my
data
or,
if
I
just
truncate
it
and
say,
okay,
I'm
just
going
to
set
a
background
value
in
the
padding.
H
Then
I
noticed
that
you
get
a
really
big
penalty,
sometimes
on
the
regularization,
and
so
this
kind
of
choice
on
how
you
want
to
put
something
in
the
padding
is
really
impacting.
My
final
inversion
result,
so
I'm
kind
of
playing
around
with
some
stuff
to
try
and
maybe
balance
it
or
sort
of
decay
it
to
a
background
value
so
that
I'm
yeah
not
running
into
those
problems.
H
But
yeah,
going
going
from
2d
inversion
results
to
3d
and
wanting
to
to
create
a
reference
model
is
not,
I
guess,
as
trivial
as
I
thought.
B
Have
you
do
you
know
about
the
natural
neighbor?
No,
so
it's
kind
of
like
a
it's
almost
like
a
blended
nearest
neighbor
thing,
it's
another!
It's
another
method
of
kind
of
interpolation.
That
kind
of
you
can
imagine
like
it
looks
at
like
so
many
nearest
neighbors,
and
then
it
blends
them
together
as
it's
extrapolating
like
it's
a
little
bit
more
complicated
than
that.
But
it's
kind
of
like
a
the
idea
of
what's
happening
is
that
it's
planning
the
idea
of
different
neighbors.
H
C
B
C
H
H
H
E
H
H
Yeah
some
I'm
pro,
I
don't
know
I'm
playing
around
with
something
right
now,
where
I
mean
I
just
kind
of
said
what
the
two
problems
were
right.
One
is,
you
might
accidentally
nearest
neighbor
everything
and
just
put
some
giant
structure,
because
there's
a
single
cell
there
hanging
out
in
the
corner,
that's
got
a
big
value.
C
C
Just
your
it's
just
it's
your
background.
H
Yeah
yeah,
that's
sort
of
what
I'm
gonna
do
that's
what
I'm
playing
with
right.
Now,
it's
it's
gonna
start
with
the
nearest
neighbor
and
then,
as
you
go
further
away,
it's
going
to
decay
with
some
level
of
smoothness
to
a
background
value
yeah,
because
if
you
don't
yeah,
if
you
don't
do
that,
then
yeah,
your
your
regularization,
can
become
a
problem.
H
H
Yeah
we
successfully
implemented
it
in
the
in
in
the
manual
these
for
that
problem.
When
there's
anomalous
bodies
that
aren't
hanging
out
at
the
surface,
it
seems
to
work
pretty
well.
H
I
haven't
really
gotten
anything
satisfactory
for
mount
isa
as
of
yet
so
I'm
finding
that
with
mount
isa,
putting
some
measure
of
near
surface
like
horizontal
interface
weights,
to
try
and
get
it
to
be
a
little
bit
smoother
near
the
surface.
That's
giving
me
results
that
I'm
much
happier
with
and
it's
converging
to
target
misfit
in
a
way.
H
That's
a
lot
more
satisfactory,
even
putting
very
minimal
scent
like
using
a
very
high
truncation
value,
so
that
I'm
only
putting
the
weights
immediately
immediately
next
to
the
receivers
I'm
still
pushing
stuff
out
and
now
it's
sort
of
clustering
at
the
surface,
but
in
between
the
survey
lines
as
opposed
to
clustering
along
the
survey
lines.
So
I
haven't
found
a
good
way
to
just
get
that
that
balance.
H
So
I
think
it
it
works.
Like
the
the
I
would
say
when
I
look
at
the
sensitivities
and
I
I
look
at
the
the
shape
of
my
my
sensitivity
weights
model,
it
passes
the
eye
test,
but
I
haven't
been
able
to
go
and
find
kind
of
a
consistent
workflow.
That
would
make
me
say:
oh
it's
putting
everything
in
the
right
place,
most
of
the
time
or
all
the
time.
So.
H
Yeah
yeah
it
you
would
see
basically
separated
into
two
executables.
H
One
executable's
got
a
set
of
iterative
methods
to
go
and
approximate
the
sensitivities
and
it
will
output
the
true
sensitivities
and
that's
the
time
consuming
part
and
then
there's
a
second
executable
which
will
take
those
and
we'll
turn
it
into
a
a
weights
model.
So
it
will
pick
a
truncation
factor.
It
will
normalize
it
and
that
is
really
really
quickly.
It
will
also
there's
a
you:
can
apply
a
smoothing
to
it
and
so.
H
C
So
when
you
say
he
passes
the
ites,
I
mean
you're
not
really
seeing
the
real
sensitivities
you're.
Just
looking
at
the
approximate
essentials
against
your
weight,
so
you're
saying.
H
Yeah,
I'm
kind
of
looking
at
the
the
sensitivities
that
I
got
out
and
the
sensitivity
and
the
dc
weights
and
kind
of
seeing
is
it.
Is
it
decaying
at
the
same
rate
vertically
and
horizontally?
Is
it
I
might
end
up?
Do
I
end
up
with
a
kind
of
smoother
approximation
or
representation
of
my
approximate
sensitivities,
but
yeah
right
right
now,
I'm
still
anyone
who
wants
to
play
with
it
is
welcome
to
to
play
with
it,
because
it's
not
as
robust
right
off
the
bat
as
we
were
hoping.
D
Of
a
question,
so
I
wasn't
really
sure
where
we
are,
it
has
to
do
with
the
license
for
the
data
for
gwb,
so
I'd
like
to
get
back
to
them
and
yeah.
Let
palavi
know
that
you
know
they
can
have
the
data
sets,
but
I
wasn't
quite
sure
exactly
where
we
were
with
how
to
describe
the
license
or
just
where
to
where
to
allow
her
access
to
everything.
D
A
So
we
should
do
this
through
the
website.
Maybe
you
me
and
devin
can
just
hang
on
briefly
after
the
call,
and
we
can
sort
that
out.
Would
that
be
all
right.
Yep!
That's.
D
Fine
yeah
and
then
I
guess
the
other
big
thing
is.
We
finally
have
this
american
gringo
coming
out
of
quarantine
here.
So
yeah
we're
gonna,
have
a
pizza
party,
so
everybody's
invited
and
yeah.
D
So
we'll
just
have
the
the
limit
that
we're
allowed
for
our
phase
2.5
here
in
british
columbia,
for
a
number
of
people
at
a
pizza
party
and
yeah,
we
could
probably
keep
everybody
socially
distanced
and
I'm
hoping
people
would
be
yeah
going
to
kind
of
bring
something
a
little
bit
novel
for
whatever
they
want
on
a
pizza
on
their
pizza
crust,
marion's,
making
some
dough
so
we'll
have
and
we've
got
a
you.
G
D
Pizza
we've
got
a
pizza,
stone
and
a
paddle,
and
we
can
put
whatever
we
want
on.
We
got
a
good
oven,
so
it
cooks
it
in
about
five
minutes,
so
everybody
can
bring
whatever
you
might
like
to
have
we'll
have
some
basic
stuff
and
yeah
meet
at
our
place
at
at
six
o'clock.
So
soggy
lindsay
might
be
a
little
bit
of
a
hoof
for
you,
but
you
can
see
what
you
can
do.
D
A
D
A
Excellent,
so
I
have
a
couple
updates.
I
guess
one
thing
I've
been
doing
is
there's
the
ubc
jupiter
days
is
happening
right
now,
today's
like
very
basic
sort
of
intro
stuff
and
then
the
next
two
days
are
going
to
be
a
bit
more
in-depth
I'll,
be
presenting
on
friday
on
widgets
and
things
like
that.
So
I'll
give
a
demo
on
widgets
and
dashboards,
but
there's
some
interesting
talks.
I'm
excited
to
see
some
of
the
stuff
where
folks
are
talking
about
pedagogy
and
things
like
that.
So
I'll
report
back.
A
If
there's
anything
that
stands
out
the
other
thing
I've
gotten
so
after
transform,
I
got
an
email
from
some
folks
at
the
geologic
survey
of
brazil
and
they're
really
interested
in
having
well.
They
first
suggested
a
course,
and
I
said
well
right
now-
I
mean
that's,
that's
a
lot
to
put
together
and
also
you
know,
we're
not
traveling,
and
so
that
makes
it
pretty
tricky.
If
you
want
to
do
something,
that's
a
bit
longer,
but
also
the
time
thing
is
what
I'm
most
concerned
about.
A
So
what
I
suggested
is
that
why
don't
they
actually
have
folks
like
watch
the
transform
tutorial
and
then
we'll
just
show
up
and
do
a
q
a
for
an
hour
and
have
it
be
a
fairly
informal
but
sort
of
first
just
approach
to
sort
of
get
people
in
like
engaged
in
the
community,
maybe
meet
some
friendly
faces
and
just
have
a
chance
to
to
chat,
and
hopefully
they
could
submit
some
questions
beforehand.
A
But
I
figured
that
that
seemed
like
a
pretty
not
too
much
overhead
on
us,
but
a
good
chance
to
meet
with
some
people
who
are
keen
to
get
up
and
running
with
simpeg,
so
they
suggested
in
like
february
march,
like
they're
thinking
next
year.
So
I
don't
know
if
folks
have
constraints
or
things
like
that.
If
not
we
can
just
kind
of
like.
A
I
think
we
can
basically
just
suggest
a
date
and
then
you
know
do
a
one,
maybe
two
hour
q,
a
session
and
maybe
a
couple
of
demos
or
things
like
that
of
stuff.
That's
going
on
so.
A
Are
they
at
they
are?
A
Okay,
I'll
get
a
sense
of
the
the
audience.
I
think.
Well,
I'm
pretty
sure
that
marco
who
reached
out
to
me,
I
believe,
he's
a
geophysicist.
So
I
think
I
think
some
folks
probably
are
okay
yeah,
so
was
it.
A
B
A
Yeah,
but
we
could
also
reach
out
to
alini
and
see
if
there's
sort
of
other
connections,
because
it
would
be
nice
actually
to
get
and
because
I
think,
she's
involved
with
the
geophysical
society
there.
So.
B
A
D
A
Yeah,
so
maybe
what
I'll
do
is
I'll
draft
just
like
a
quick
sort
of
email
response
of
what
we
kind
of
have
in
mind,
so
I
think,
like
they
sound
like
they're
good
with
the
sort
of
watch
the
tutorial
and
then
q
a
session,
but
we
can
also
put
in
their
request
they
reach
out
to
other
groups
and
universities.
We
have
a
few
contacts
from
the
disc
and
things
like
that,
so
that
would
be
good
other
than
that.
A
H
Is
sort
of
the
time
that
we
should
get
busy
and
finalize
it?
Is
that
like
months
away,
or
is
it
weeks
away.
H
Oh
yeah,
so
we
submitted
the
agu
abstracts
for
our
our
final
gwb
project
and
I
guess
I'm
wondering
is
that
months
away
or
weeks
away
so.
A
Yeah,
well,
I
mean,
I
think,
we'll
be
able
to
reuse
a
lot
of
this
stuff
from
scg,
but
actually
that's
a
really
good
point.
Scg
like
we
need
to
have
recordings
by
september
1st,
which
is
actually
soon.
H
A
And
actually,
on
that
note,
with
seg,
there
was
an
email
sent
out
to
authors
that
I
wanted
to
flag
so
that
everyone
knows
you
have
to
take
action
by
the
17th
and
either
give
them.
So
there's
a
couple
different
options
for
the
licensing
like
agreement
for
your
recording
is
you
can
give
them
rights
to
use
it
for
a
month
or
you
use
it
indefinitely
or
not
at
all?
But
that
also
means
that
you
talk.
D
So
soggy
I'm
wondering
how
that
would
work
for
the
sdg
because
you
actually
submitted
it.
You
were,
you
were
the
person
that
was
officially
acknowledged
or
you
got
the
initial
email
about
acceptance.
G
D
A
Sort
of
thinking
that
I
just
want
the
month,
I
don't
know
how
I
feel
about
sdg
or,
like
I
don't
know
what
their
plans
are
for
keeping
these
recordings.
I
don't
really
know
so.
G
A
Cool
okay!
So
then,
on
that
note
for
the
seg
for
the
gwb
presentations,
if
folks,
who
are
lead
on
each
one,
want
to
start
sketching
out
what
your
presentation
looks
like
and
devin,
maybe
if
you
wouldn't
mind
just
pinging
ren,
just
to
make
sure
that
this
is
on
on
her
radar
too
yeah.
That
would
be
great.
A
D
E
D
All
right,
fair
enough,
okay,
so
dev
and
lens
will
stay
on
for
a
bit
after.