►
From YouTube: SimPEG Meeting March 24th
Description
Weekly SimPEG Meeting from March 24th, 2021
A
Yourself
to
the
attending
list,
if
you'd,
like
general
formulas,
we're
going
to
go
through
like
we,
don't
really
have
too
much
of
an
agenda
today.
If
there's
any
items,
kind
of
sort
of
up
there
like
discussion,
items
or
everything,
go
on
and
then
kind
of
go
through
people
just
to
give
a
quick
report
of
what
we've
been
up
to
recently
or
an
introduction,
or
if
you
want
you
don't
need
to
you,
don't
have
anything
to
say:
don't
worry
about
it,
no
pressure!
A
So
with
that
I'm
trying
to
get
into
some,
I
don't
have
any
agenda
items
per
se.
I
borrowed
it
last
week,
but
I
think
we're
going
to
do
a.
I
was
going
to
host
another
friday
afternoon
session
this
week
for
people
who'd
like
to
stop
by
the
friday
afternoon,
happy
hour
coding
party
thing
whatever
you
guys
want
to
do
I'll,
put
up
a
link
friday
around
4
p.m,
pacific
time,
so
welcome
to
the
journal.
B
C
C
Sure
yeah,
I
can
introduce
myself
first,
I'm
geophysicist
by
training.
I
did
my
undergrad
with
dom
so
that
era
and
yeah
I've
been
working
in
industry
for
some
of
the
normal
players
there.
Frontier
precision
last
job
was
at
computational
geosciences,
and
I
did
my
masters
at
memorial
with
colin
farkasen
yeah.
C
A
D
A
B
Hi
I'm
lindsay,
I'm
be
starting
up
at
ubc
as
an
assistant
professor
in
the
summer.
I
did
my
phd
at
ubc
previously
and
then
was
at
uc
berkeley
for
a
postdoc
working
with
folks
on
the
jupiter
team
and
in
the
stats
department,
so
interested
in
pushing
forward
on
geophysical,
inversions
and
data
science.
A
F
F
I'm
a
postdoc
at
thieu
delft
in
the
netherlands
working
here
with
everett
slope
in
in
electromagnetics
and
beside
this
we're
pretty
much
on
our
own,
but
when
it
comes
to
electromagnetics
in
our
department.
So
I'm
very
happy
to
have
a
home
here,
an
em
community
to
bang
my
head
with
questions
and
doubts
and
then
just
generally
some
exchange
of
ideas
and
thoughts.
G
C
C
G
H
A
I
can
go
first
here,
so
I've
got
a
full
request
up
on
discretize
right
now.
Just
add
some
simple
vector
plotting
for
tree
meshes,
built
off
of
what
dom
had
worked
on
with
the
filter.
Flux
just
uses
it
for
just
a
clever
plotting
of
vectors
on
the
cell
centers.
Essentially,
I
also
just
wrote
some
quick
tests
to
make
sure
that
it
at
least
ran
for
every
type
of
input.
A
A
F
A
A
H
F
A
B
Thanks
joe
just
a
quick
note
that
I'm
giving
a
talk
tomorrow
for
a
it's,
a
research
group
or
like
research
collaboration,
mostly
in
with
universities
in
quebec,
it's
students
grad
students
who
are
in
sort
of
sustainability
science,
and
they
were
curious
to
get
a
talk
on
kind
of
open
science.
So
I'll
be
talking
a
bit
about
sort
of
simpeg
and
gsi
as
an
example
and
then
some
examples,
kind
of
from
the
broader
like
python
and
jupiter
ecosystems.
B
B
B
A
B
B
I
Yeah
not
not
not
much
big
development,
but
you
know
ben
being
here.
I
guess
we
should
talk
about
the
kind
of
the
big
picture
that
what
we're
trying
to
do
mira.
So
we
forked
the
current.
You
know
the
current
status
of
absent
bag
onto
the
the
mirage
of
science
get
account
and
the
goal
with
that
is
just
that
we
can.
You
know
we
can
do
dev
on
our
side
without
cluttering
too.
You
know
too
much
branches.
I
But
you
know
our
goal
is
to
constantly
you
know.
You
know,
pull
and
push
stuff
back
and
forth
between
the
two
between
the
two
repos,
because
we're
probably
going
to
be
working
mostly
on
io
stuff
for
for
the
geo,
h5
format,
and
then
you
know
testing.
You
know
more
like
parallelization
efficiencies,
that
kind
of
stuff,
so
so
yeah
we'll
be
working
basically
on
our
own
git
account,
while
keeping
track
of
everything's
happening
here
and
yeah.
D
I
Yeah
I
mean
we
still.
We
still
need
to
scope
out
the
you
know
the
plan,
the
action
plans,
but
I
have
the
feeling
that
if
you're
gonna
do
ios
for
geo
h5,
it
would
probably
be
on
our
side.
You
know
you'll
be
basically
importing
a
geo,
h5
library,
gh5,
jewish
firefighter,
whatever
I'm,
not
sure.
If
simpek
really
needs
it,
we'll
we'll
scope
it
out,
we'll
figure
it
out
and
then
we
can.
D
Yeah,
I
guess
just
just
know
that
there
is
a
place
for
it
and
I
guess
I'm
doing
some
some
development
on
that
trying
to
have
the
I
o
for
all
the
ubc
gif
formatted
stuff
yeah,
that's
great!
There
is
a
bit
of
a
plan,
so
hopefully
we
put
it
in
the
same
place,
oh
for
sure
yeah,
where
we'll
follow
your
lead
on
on
that.
On
that
part,.
A
E
E
And
so
yesterday
I
was
trying
to
kind
of
make
another
version
of
that.
That
would
work
on
tree
meshes.
So
joe
was
helping
me
a
little
bit
with
that.
It's
close,
there's
still
something
still
scrambled
in
the
ordering.
So
it
looks
like
a
bit
of
a
nightmare
right
now,
but
I'm
sure
I'll
get
that
figured
out
here
today,
probably
hopefully
so
I
was
just
wondering
more
than
anything
else.
A
I
think
it's
useful
to
have
something
that
you
know
picks
out
the
boundaries
like
a
general
model,
because
it's
something
we
could
use
to-
let's
enforce
it.
It
could
just
be
like
a
general
utility.
That's
okay,
you
want
to,
you,
know,
define
boundaries
to
find
like
the
faces
to
find
a
you
know
for
boundary
conditions.
Okay,
we've
got
a
model
like
just
kind
of
simple
utility
of
cells
like
active
cells
versus
inactive
cells.
I
want
to
identify
the
boundary
between
them
or
something.
I
I
think
it
might
go
and
model
utils.
I
think
it
should
just
be
an
argument
that
you
can
also
plot
faces
right,
face
values
on
on
the
sections
you
just
expand.
The
current
plotting
functions
to
just
be
able
to
plot
boundaries,
because
I
agree
with
joe
like
we
were
talking
about
it
last
week,
right
regularization,
you
should
be
able
to
take
face
weights
so
we'll
we
need
a
function
that
what
you're
doing
right
now
mike
to
be
able
to
go
from
it.
I
E
I
think
the
approach
I'm
taking
right
now
might
be
a
little
different
than
what
we
would
use
for
that,
probably
because
I'm
literally
just
like
grabbing
the
faces
that
have
kind
of
a
non-zero
gradient
in
that
synthetic
model,
but
that's
it
and
then
yeah.
So
I
guess
it's
just
depends
on
how
we
want
to
plot
that
and
how
we
what's
the
best
way
to
try
to
then
visualize
those
faces.
But
I
think
you
should
just
tease
it
apart.
I
E
E
A
It's
a
nice,
it's
a
neat
little
trick
like
to
do
it.
If
you,
if
you
guys,
haven't,
looked
what
it
essentially
does.
It
goes
through
and
it
puts
it
generates
like
this
fine
2d,
tensor
mesh
and
like
finds
all
of
the
unique
cells
that
intersect
this
2d
tensor
mesh
in
and
like
a
3d
in
the
three
three
bench.
A
So
you
it
checks
all
those
unique
cells.
It
gets
all
of
their
locations
and
their
like
stream
levels,
and
then
it
creates
a
new
tree
mesh
that
inserts
all
of
those
unique
cells
at
that
level.
So
it
creates
like
an
exact
like
tree
mesh
slice
of
black
tree
and
then
interpolates
between
it's
a
neat
little
little
trick.
E
G
So
mike,
when
you
say,
you're
looking
you're
finding
the
boundaries
is
this
of
a
like:
a
rock
model
in
which
you've
got
particular
units,
or
is
this
already
a
you
know,
a
smooth
inversion
and
then
you're
going
to
define
a
boundary
wherever
your
gradient
is
sufficiently
large
or
how
are
you
yeah?
Under
what
circumstances
are
you
finding
boundaries.
E
E
B
Exactly
yeah
sounds
good
wow
I
just
wanted
to
flag
too.
I
put
a
link
in
the
notes
to
a
notebook
dieter-
and
I
talked
about
this,
but
for
the
paperwork
theater.
We
had
a
very
large
sort
of
underlying
mesh
or
very
large
model
on
a
uniform
mesh
and
designed
a
tree
mesh
from
that.
So
I've
also
got
some
code
for
that
sort
of
thing.
So,
if
that's
of
use
feel
free
to
scrape.
D
D
A
I
mean,
I
don't
know
if
it's
significant
but
they're,
just
kind
of
like
it's
a
lot
of
repeated
things
that
I
think
should
be
used
like
should
have
been
used
from
other
areas
or
it's
like
it's
like.
We
create
a
list
of
abm
locations
and
we
each
have
like
the
same
copy
code
like
four
or
five
times
to
make
it
into
a
survey,
and
I
think
we
just
have
like
we
do
have
like
a
you
know,
create
survey
from
a
b
and
then
function.
D
D
A
A
A
A
A
H
H
I
haven't
had
a
chance
to
review
the
syntax
dm1d
stuff,
but
I
would
like
to
do
so
I'll
do
that
in
this
week
and
other
thing
I
went
to
a
conference
yesterday
called
like
a
there's:
a
focus
on
environmental
geophysics
I'll
share
the
link
and
they
probably
upload
their
recordings
later,
and
they
were.
It
was
quite
diverse,
but
there
was
some
interesting
conversation
about
kind
of
thinking
about
like
open
source
software
and
also
integrating
different
disciplines
and
how
the
geophysics
can
play
a
role
there
and
what
are
the
challenges?
F
H
All
right
peter
thanks
for
bringing
that
out,
because
I
need
to
put
things
together.
It's
still
like
a
rough
idea,
but
I'll
share
the
kind
of
outline
later.
But
the
idea
is
yes,
I'm
kind
of
targeted
to
geologists,
but
I
changed
the
title
to
a
version
for
geoscientists
just
to
make
it
a
little
bit
broader,
because
I
I'm
still
not
sure
who
are
the
major
audience
of
that
transform
21.
H
Yes,
I
was
just
focusing
trying
to
focus
on
mental
model
like
okay,
how
we're
setting
up
an
inverse
problem.
What
are
the
major
challenges?
Obviously,
like
we're
doing
a
data-driven
solution,
but
that's
not
enough,
so
we
need
to
put
some
sort
of
irrelevant
knowledge
and
that's
the
connection
point
with
geophysicist
as
well
as
with
the
other
like
the
geoscientists.
H
F
That's
cool,
I
think
the
audience
from
last
year's
experience.
There
will
be
a
lot
of
students.
There
will
be
many
experienced
funds
as
well,
so
yeah,
I
think
it's
cool
I'll,
probably
get
in
touch
with
you
soon
to
have
a
test
recording.
If
I'm
the
one
going
to
host
you.
It
looks
like
well
yeah
looking
forward
to
it.
H
F
F
H
F
F
Yeah,
so
what
they
did
implement
through
chempai
is
just
different
different
objects,
so
you
say
so
like
point
data,
I
think
line
data
like
well
logs
sections
surfaces,
3d
models
and
defined
how
they
are
so
we
would
just
have
to
look
at
it
into
that
and
see
how
we
can
load
it
and
export
to
that
in
in
discretize
and
then
satiando
and
leo.
They
probably
do
the
same
for
harmonica
at
least,
and
some
of
them
by
kimberly
will
do
the
same
for
pi
gimli.
F
And
if
everybody
does
it
on
their
own
package,
then
it
should
be
straightforward
to
just
move
it
along
and
then
maybe
that
would
have
to
go
to
subsurface
to
things
like
omf
or
gh5
pi,
the
definition
how
you
store
it
on
disk,
that's
they're,
not
necessarily
the
same
right.
One
is
how
you
pass
it
along,
and
the
other
is
how
you
store
it
to
disk.
F
A
H
I
So
the
part
that
is
like
you
know
the
visual
of
it
where
people
you
know,
click
buttons
and
stuff
would
be
under
a
paid
on
their
paid
wall,
but
the
code
itself.
That
runs
the
code.
You
know
that
run
simpeg
is
going
to
be
on
an
open,
open
repo.
So
you
know
you
could
always
replicate
the
ui.
If
you
want
to
replicate
the
ui
yeah.
H
There
are
multiple
ways
that
people
are
going,
but
I
thought
like
it
seems
like
people
are
repeating
the
same
procedures,
so
I
was
hoping
like
if
there's
some
sort
of
synergy
that
can
be
made
but
yeah,
I
wasn't
sure
exactly.
But
I
was
just
kind
of
curious:
what's
the
barrier
and
what
are
what
are
the
availability
of
what
you're
developing
yeah?
Okay,
okay
got
it.
I.
I
Mean
there
is
at
some
point
you
know
there
needs
to
be
some
money
generating
part
of
this,
so
the
idea
is
to
draw
people
towards
the
the
paid
version
of
analyst
by
you
know
giving
them
an
easy
access
point
all
right,
just
a
click
button
like
I
want
to
invert.
I
don't
want
to
have
to
think
about
anything
else.
So
right,
that's
where
the
paid
welcome.
H
H
F
I
I
B
H
So
I
think
that
big
umbrella
here
they
want
to
find
more
sustainable
kind
of
ish
project
and
doing
an
open
source
sims
a
good
fed.
So
in
the
department
level
they
have
some
sort
of
initiative
to
do
that
and
it
seems
like
they
have
some
money
to
put
it
in
to
hire
like
a.
It
may
be
a
software
engineer,
not
sure,
but.
I
That'll
be
that
side
we
should
talk
offline,
because
you
know
we
do
this
with
a
lot
of
a
lot
of
other
organizations
that
if
they
want
to
have
a
specific
product
outcome
right,
they
can
be,
they
can
become
a
partner
and
get
you
know,
infinite
licenses
for
for
analysts
and
get
what
they
want
right
so
and
instead
of
going
blind
and
hiring
someone,
we
could
just
merge
the
projects
together.
Basically
right
it
doesn't
happen.
H
Yeah
something
like
yeah,
something
like
that
and
I
I
think
it's
sort
of
both.
I
guess
they
are
interested
in
the
further
developing
that
the
aem
code
a
bit
more
efficient
and
also
having
a
bit
of
good
io
codes
that
or
like
sort
of
graphic
user
interface
that
people
can
readily
run
the
code.
And
I
look
at
the
data
yeah.
So
I
think
rosemary
is
tired
with
that
using
r
whose
harvest
product
so
she's.
She
wants
to
stop.
B
Some
of
these
questions
as
to
like,
even
if
you
know
if
things
go
forward
with
mira,
but
if
they
decide
to
also
hire
a
you
know.
Software
engineer
is
a
part
of
this
project.
Where
would
be
like
strategically
very
useful
for
for
everyone
right
now.
I
don't
think
we're
quite
organized
enough
around
that.
I
think
I
don't
think
we're
too
far
off,
but
you
know
if
they
were
hired
tomorrow
and
we
I
don't
think
we
would
have
a
clear
sort
of
list
of
like
this
would
be
incredibly
impactful
right
now.
H
H
So
in
the
beginning
of
this
grace
project,
it
was
mainly
before
this
grace
is
used
for
mainly
like
monitoring
the
water
storage
in
a
large
scale
and
they
were
reluctant
to
open
up
the
graded
data,
but
the
actually
opening
up
that
breeded
data
without
worrying
too
much
about
misuse,
was
actually
a
big
game
changer
now,
after
10
years,
there
are
a
lot
of
people
using
it.
Although
there
are
some
misuse
case,
but
the
kind
of
intensity
of
using
it
and
kind
of
people
recognizing
how
useful
it
can
be
was
it
was
a
big
difference.
B
I'm
wondering
if
it's
worth
just
regrouping
a
bit
and
then
post
something
I'm
happy
to
sort
of
take
a
look
and
take
stock
again
of
where,
where
we're
at
with
that
conversation
and
then
maybe
come
up
with
a
bit
of
a
proposal
for
next
steps
before
setting
a
meeting.
Would
that
be?
Would
that
be
all
right
with
folks.