►
From YouTube: SimPEG Meeting September 22nd
Description
Weekly SimPEG meeting from September 22nd, 2021
A
B
Mike
did
a
few
up
and
down
the
stairs
with
a
lot
of
steel
rods
and
the
hammers
in
his
life
did
his
fair
share.
B
C
A
B
A
It
was
nice
day
yesterday,
actually
down
there
sunny
in
the
afternoon,
especially
wednesday
or
monday
afternoon,
was
a
little
bit
more
cloudier,
but
it
was
still
pleasant,
weather
or
pleasantly
warm.
A
B
A
D
In
october,
no
that's
a
good
question.
I
was
wondering
that
I
don't
know
if
we,
if
somebody
volunteered
or
if
we
ask
someone,
I'm
not
sure
yet
either
but
I'll
look
into
it.
When
I
start
getting
that
together,
there
is,
I
mean
if
I'm,
if
I
can
volunteer
someone
that
is
not
even
in
the
meeting
and
that
I
don't
even
know
like
I've,
seen
a
lot
of
things
happening
on
the
over
like
the
repo
on
the
simple
again,
I'm
just
not
sure
and
don't
know
what's
happening
there.
D
B
Fair
enough
next
week,
though,
it's
scg
right,
so
it
might
be
a
little
a
little
busy.
A
Okay,
so
one
of
the
really
quick
reports
from
me
I've
been
going
through.
I
got
a
lot
more
time
to
work
on
some
stuff.
A
Back
through
started
going
through
the
cross-creating
stuff
again
last
night,
so
it's
almost
it's
getting
it's
almost
there.
I
just
want
to
write
a
few
more
tests
for
it
to
make
sure
it's
passing,
hitting
everything,
and
then
I've
got
my
interpolation,
mostly
working
for
the
tetrahedral
meshes,
and
that's
that's
pretty
much.
The
last
part
that
I
think
we
need
to
get
in
if
you're.
A
If
you're
curious
about
it,
it
makes
use
of
a
directed
search
to
find
a
simplet,
a
simplex
that
it's
a
part
of
so
part
of
interpolating
at
a
tetrahedral
mesh
is
finding
like
what
triangle
is
located
in
right
and
it's
not
always
as
simple
as
just
looking
for
the
nearest
cell
center
cyaniding
that
one
because
it
could
end
up
like
outside
of
the
triangle
and
still
be
close
to
a
different
cell
center.
So
what
you
do
what
I?
A
What
I
did
was
since
we
need
cell
centers
most
of
the
time
anyway,
we
just
compute
a
tree,
okay,
kd
tree
from
that
to
get
close
to
one,
and
then
you
use
a
directed
search.
So
on
the
tetrahedral
messages.
There's
you
work
often
with
very
centric
coordinates
and
what
you
can
do
is
so
you
get
a
triangle
and
then
you
compute,
its
very
centric,
coordinates
if
it's
a
negative
value.
It
means
it's
outside
the
triangle
in
a
certain
direction.
So
you
you
compute
the
coordinates.
A
A
A
I'll
put
that
up
today
on
discretize
for
the
logo
thing
I
was,
I
got
sidetracked
I
was
like.
Oh,
I
was
looking
around
if
we
had
a
a
pole
bot
on
the
sim
pic
slack
channel,
but
I
don't
think
we
do
so
I'll.
Probably
just
do
it
in
terms
of
react
to
this
message,
depending
on
which
one
you
want
just
simple
and
horrible.
A
Sorry,
I'm
at
tevo,
I
see
you
got
some
stuff
here.
D
Yeah
so
last
week
I,
like
I
pushed
more
on
that
branch
that
I've
been
working
on
for
a
while
for
some
improvement
of
the
pgi
implementation
and
it's
in
the
state.
Now
that
I
think,
is
ready
to
to
be
reviewed
and
like
and
like.
I
did
a
lot
of
things
that
I
wanted
to
put
that
in
that
pr
like.
There
is
a
couple
of
numerical
improvements,
especially
when,
like
physical
property
distribution
are
overlapping,
so
we
do
a
better
job
with
the
gradient
animation.
From
those
cases
I've.
D
There
is
a
lot
of
like
quantity
of
life
improvement
and
like
having
better
default
parameters
for
a
lot
of
things,
so
that
we
don't
have
to
specify
all
of
the
parameters
at
all
the
time,
and
there
is
like
a
couple
of
new
functions
that
I
added
to
like
in
like
now.
The
quasi
geology
model
is
really
accessible
just
through
a
single
function,
call
with
no
argument,
and
rather
than
calling
like
a
series
of
functions
to
create
it
when
we
want
to
access
it.
D
So
it's
a
so
it's
much
easier
on
these
on
the
user
way,
and
one
big
thing
too,
that
has
gone
being
nice,
for
the
users
too.
Is
that
the
gm
for
the
gmm?
I
actually
actually
now
created
the
embedded
function
in
it
that
you
can
just
call
it
and
it's
gonna
plot
the
gmm
for
you.
D
Instead
of
having
like
a
in
the
tutorial,
you
all
had
that
script
of
like
50
lines
for
plotting
it
so
like
now,
it's
just
like
a
single
line
of
like
gmm
dot,
plot
pdf
and
that's
and
that's
it
and
then,
if
you
have
like
several,
if
you
have
a
multi-dimensional
gmm,
you
just
have
a
flag
to
tell
that
you
want
to
plot
it
in
2d
instead
of
1d,
and
you
can
just
specify
which
component
you
want
to
plot.
So
it's
very
very
like
much
easier
to
much
easier
to
use.
So
that's
so
yeah!
D
So
that's
kind
of
so,
and
that
was
really
the
big
youtube
that
I
wanted
to
add
in
that
in
that
pr
too
so
yeah,
I'm
happy
with
it
now,
like
better
numerical,
better
numeric
numericals,
like
some
bug,
fix
better
defaults,
easier
to
use
and
new
functionalities
for
plotting
and
accessing
various
values.
D
Yeah,
so
a
lot
like
I
can
yeah,
I'm
a
vip
if
you
can
review.
I
know
like
mike.
Also
you,
like
you,
like
your
feedbacks
on
using
pgi,
are
very
instrumental
for
some
of
those
changing
so
like.
If
you
want
to
have
a
look
too
and
see
if
things
works
for
you
with
that
new
branch.
That
would
be
great
also.
B
And
that
fixes
the
the
update
for
scikit-learn
right.
D
Yes,
so
actually
it's
kind
of
nice
in
a
way
that
it
that's
a
very
good
point,
though
it's
like
I,
like
my
previous
version
of
pgi,
was
not
compatible
with
scikit-learn
1.0.
D
So
now
this
this
sphere
also
addressed
that,
like
that's,
it's
now
compatible
with
scikit-learn
1.0,
but
it's
still
compatible
with
previous
version
of
scikit-learn.
So
if
you're
using
scikit-learn
like
0.24,
it's
still
working
as
well.
So
it's
it's
a
it's
compatible
with
a
new
one,
because
it's
also
back
compatible.
So
that's
actually
quite
neat.
A
F
Yeah,
I
also
don't
have
any
updates,
but
yeah.
I
had
a
sort
of
kind
of
vacation
slash
conference,
so
we
went
to
la.
There
was
a
big
like
a
water
conference
in
l.a.
So
last
week
we
went
there
with
angela
and
daniel,
so
they
had
a
lot
of
fun
to
enjoy
la.
I
had
fun
to
meet
up
like
like
it
was
the
it's
more
like
a
most
of
the
participants
were
coming
from
either
the
governmental
organization
or
commercial
sectors.
So
it
was
interesting
to
see
kind
of
what.
F
What
is
the
need
of
geophysics
for
these
water
managers
and
what
are
the
major
applications
that
they
were
thinking
and
hoping
to
yeah
do
so
yeah.
It
was
interesting.
I
had
a
good
conversation
with
somebody
developing
an
em
system
called
bagten,
so
yeah
we're
planning
to
have
a
survey
using
their
system
and
comparing
with
the
aarhus
totem
system
yeah.
F
So
it's
interesting
kind
of
seeing
interesting
opportunities
and
I
think
the
like
generally,
the
use
of
simpac
could
be
quite
effective
for
like
their
purposes
because
they
want
to
use
a
lot
of
geophysical
data
and
interior
integrate
that
into
their
hydrological
modeling
process.
So
yeah
it
was,
it
was
interesting.
I
think
it's
an
open
question
like
how
are
you
going
to
actually
integrate
geophysical
data
into
kind
of
in
a
different
discipline
to
actually
make
a
decision
or
to
actually
manage
how
well
more
resources
so
yeah?
A
C
That's
probably
true,
yeah
yeah,
I
mean
nothing
too
no
development
type
stuff
going
on
for
my
end,
just
kind
of
mostly
been
working
on
putting
together
a
draft
of
a
paper
for
some
of
that
gravity,
data
from
the
clear
lake
area,
so
some
cool
results
from
that,
but
yeah,
I
think
that's
coming
together.
So
that's
exciting.
F
Oh
speaking
of
that
mike
so
dwr
just
actually
started
flying
over
entire
california,
so
I
think
I'll
send
you
a
website.
You
can
actually
check.
They
have
a
flight
plan,
so
you
can
check
whether
that
that
flight
plan
covers
your
study
area
or
not
because
it's
quite
big
areas,
so
it's
likely
cover.
B
B
I
have
I
did
work
on
on
refactoring
the
radarization
last
weekend
and
I
created
four
requests.
I
managed
to
go
down
all
the
way
to
merging
together
bass,
sample
antique
enough
refactoring
it
out
running
the
spars
and
now
I'm
down
to
the
pgi.
I
guess
I
need
to
update.
B
I
should
reduce
the
amount
of
lines
of
codes
by
at
least
50
percent,
so
it
will
be
a
lot
nicer,
a
lot
more
english
and
then
yeah.
Hopefully
next
meeting
will
be
ready
for
review.
B
The
question
is:
do
we
want
to
merge
them?
The
issue
1037,
where
we're
dealing
with
the
the
cell
volumes
before
or
we
just
abandon
it
and
and
deal
with
it
with
that
new
reg,
because
the
new
reg
is
gonna
handle
handle
it
all.
B
It
would
replace
basically
the
antique
enough-
I'm
not
I'm,
not
I'm
not
too
attached
to
the
what
the
name
of
the
class
is,
but
basically
right
now,
it's
l2,
we'll
we'll
do
like
the
the
usual
like
on
normalized
gradients
and
then
sparse
is
a
is
a
is
a
child
class
of
it
same
thing
as
pgi
right,
so
we
can
start
attaching
to
it,
but
they're
just
sharing
a
whole
lot
more
with
each
other,
and
it's
much
more
obvious
where
the
length
scales
are
happening
if
they're
normalized
or
not,
it's
basically
just
a
switch.
B
B
I
think
you
will
reduce
the
duplication
of
code
a
lot
because
there
was
a
lot
of
like
yeah
same
things
being
reused,
and
I
also
took
out
properties
out
of
regularization,
so
we're
just
going
to
do
like
get
our
sorry,
setters
and
regular.
Prop
python
property.
B
E
B
Bunch
of
of
duplication
warnings
right
because
some
of
the
names
will
will
be
changing
a
bit
so
skip
them
around.
D
A
D
D
File,
but
inside
there
is
like
two,
there
is
two
pgi
regularizations.
There
is
simple
pgi,
which
actually,
like
is
a
mirror
of
the
simple
regularization,
and
there
is
pgi,
which
is
a
mirror
of
the
tick
on
enough.
A
D
B
B
That's
great
that'll,
be,
I
think
that
should
be
pretty
easy,
and
I
also
I
did
the
so
we
have
cell
weights.
I
didn't
change
the
name
for
that,
because
I
I
think
it's
quite
well
established,
but
I
had
the
free
weights
and
the
free
weights
are
yeah
for
the
gradient
for
the
derivative
terms
for
the
gradient
terms.
B
If
the
free
weights
have
the
have
the
length
of
cell
salsa,
you
know
the
number
of
cells,
then
it's
projected
on
the
faces
or
you
could
strictly
directly
give
it
face
weights
right,
so
that
will
get
open
up
the
the
idea
that
we
can
have
weights
that
are
fixed
and
weights
that
are
open
free
for
you
know
our
like
directives
to
to
update
without
having
to
to
worry
about
interfering
with
like
sensitivity,
waiting
or
or
whatnot.
So
it's
kind
of
like
a
second
level
of
waiting
on
top.
A
I
always
thought
it
could
be
interesting
like
I
know
I've
talked
to
you
before
about
it,
making
use
of
some
of
those
inner
product
matrices
from
from
discretized
to
do
some
of
those
things,
because
that
would
also
allow
us
to
handle
like.
A
B
Okay,
yeah,
that's
great,
I
would
say:
let's
just
do
it
sample
for
now
and
then
we
can.
A
B
F
F
Like
a
dumb
you've
done
this
right,
like
it,
was
think
of
that
kind
of
assigning
that
local
or
like
a
rotation.
I
think
I
thought
you've
done
this
before.
B
Yeah
and
it
turned
out
to
be
a
little
bit
more
complicated
than
just
combining
xyz,
because
you
kind
of
need
to
touch
the
other
neighbors.
That's
what
the
work
I
was
working
on
right,
but
it
doesn't
matter
yeah
we
can.
We
have
a
space
to
be
able
to
explore
it
and
yeah.
That's
good,
that's
great
right,
great
great!
B
E
B
No,
I
just
branched
off
of
main
yeah,
okay
yeah,
because
the
thing
is
that
1037,
you
know
it's
it's.
It's
updating
the
current
regularization
framework,
but.
A
A
E
D
The
working
on
the
pr
last
friday
at
some
point
code
coverage-
it
was
not
like
the
code
coverage
test
was
just
not
starting.
It
was
not
like
you
see
it
in
the
list
that
it's
not
there
and
it's
it's.
It's
done
that
for
like
a
few
comments
now,
I'm
like
and
I'm
just
wondering,
what's
happening.
If
you
know,
if
you
notice.
D
D
Just
just
just
a
warning
that,
like
at
the
last
time
he
was
checking
like
the
coverage
was
like
good
but
yeah,
it's
just
it
just
doesn't
get
started.
When
I
push
a
new
comment
on
that
one
for
some
reason.
A
I
will
check
I
I'm
curious
if
it's
just
that
it
doesn't
hasn't
changed.
Maybe
I
don't
know
but
I'll
check
again.