►
From YouTube: SimPEG meeting Sept 30
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).
B
Yeah
I'm
getting
used
to
this
whole
thing
and
little
three-hour
stints
of
sleep
here
and
there
I'm
still
alive
so
yeah.
It's
all
good
she's
actually
really
well
behaved,
not
too
much
crying
and
cute
as
hell.
So
excellent.
C
B
Baby
but
yeah
other
than
that
yeah.
I
don't
have
really
much
too
much
new
right.
I'm
trying
to
pick
away
at
that
tiling
stuff
there
dom.
I
see
you
finally
picked
up
some
stuff
there
and
yeah
I'm
just
getting
into
it.
So
maybe
I'll
dug
you
here
as
I
slowly
get
it
back
into
things.
B
Oh
yeah,
yes,
well,
once
we
got
back
from
the
hospital,
the
whole
culvert
thing,
it's
really
weird
like
it's
just
like
the
baby
was
born.
It
was
just
like
us,
like
no
parents,
nobody
could
come
and
it
was
just
like
what
the
what
would
I
do
so
yeah
and
then
being
we
had
to
stay
there
for
a
couple
days
and
it
was
just
oh,
it
was
tough
just
being
alone
without
much
help
so,
but
we're
all
good
excellent.
A
Excellent
cool:
well,
I
see
dom
in
the
middle
of
my
screen.
Would
you
like
to
go
next.
C
That
skip
chantal
is
behind
me
and
I'll.
Talk
later
sounds
good.
D
Sure
we've
been
just
been,
I
put
out
that
pull
request
after
the
meeting
last
time,
people
have
been
looking
through
it.
I've
been
trying
to
implement
some
a
few
little
things
in
there.
There
is
the
I
think.
It's
almost
it's
basically
ready
to
go.
I
think
we
kind
of
come
to
consensus
on
most
things,
except
I
does
still
can't
think
of
a
good
word
or
a
good
name
for
that
mesh
tensor
function.
D
I've
been
trying
to
write
my
head
around
it,
and
I
just
I
don't
so
if
anyone
what
the
function
does
is
it's
the
function
that
takes
that
kind
of
fancy
input
for
the
cell
width
vectors
like
that,
and
then
it
changes
like
that
expansion
factor
out
to
the
like.
It
takes
a
tuple
of
numbers
that
describe
like
the
padding
cells
and
it
expands
that
out
to
a
like
the
actual
wits
like
it
does
that
operation
for
us.
E
D
H
A
G
D
D
It
in
the
name
that
it
has
something
to
do
with
cell
wits,
because
that's
what
we're
getting
out
of
it
yeah
I
I
can't
figure
out
what
to
call
what
we're
putting
in,
because
it's
like
it's,
the
let's
double
check
here.
His
primary
use
is
that
it
takes
that
tuple
of
you
know,
number
number
number,
an
expansive
release,
expansion
factor.
A
Cell
widths
from
tuples,
so
it's
util-
you
can
just
make
it
vague.
A
H
D
We
don't,
if
the
user
does
that
on
our
end,
we
don't
want
to
fix
it
for
them,
because
if
we
do
that,
then
that
is
what
matplotlib
is
trying
to
get
trying
to
avoid.
They
don't
want
people
to
do
that
right.
They
want
people
to
fix
it
on
the
way
that
they're
calling
the
functions,
but
then
because,
if
we,
if
we
try
to
take
those
inputs-
and
you
know
fix
it
ourselves-
then
we're
committing
the
same
error
that
matplotlib
is
trying
to
avoid,
but
is
it
is.
E
It
is
it
logical
for
a
user
that
uses
a
discretized
plotting
function
that
he
can
provide
a
c
limb
if
he
doesn't
use
a
norm,
but
he
can't.
If
you
use
a
known,
then
I
would
get
rid
of
the
c
limb
and
just
use
the
matte
block
with
functionality,
v
min
and
v
max
as
matpative
uses,
because
like
this
with
ceiling,
they
kind
of
provide
some
functionality,
but
only
for
some
cases
and
that
doesn't
make
sense.
Vlan
ceiling.
D
It
did
and
in
the
way
that
our
plot
image
cloud
grid,
plot
slice
commands.
D
And
put
it
into
v
min
v
max
of
a
p
color
mesh.
We
did
that
every
time
right.
So
that
way,
no
matter
what,
if
the
user
supplied
a
norm
argument
to
the
plot
slice
and
plot
image
commands,
we
would
recognize
that
our
code
would
recognize
that
there
was
that
celine
was
none
and
it
would
create
beam
and
v
max
from
the
function
or
from
the
values
and
then
put
that
into
the
color
bar
or
put
it
into
the
pass
that
to
the
function.
D
The
so
that,
therefore,
if
the
user
supplied
a
norm
to
p
color
options,
we
would
take
the
norm
and
still
pass
v
min
v
mac
or
still
create
a
v
min
v
max
and
use
that
which
is
what
was
causing
deprecation
warnings.
So
we
we
don't
want
to
do
that
internally
and
what
matplotlib
deprecation
is
doing.
Is
it's
discouraging
users
from
providing
a
vmin,
vmax
or
equivalently
climb
in
our
functions
that
we've
been
using
and
a
norm?
At
the
same
time,
they.
H
D
A
Norm
should
we
just
get
rid
of
like
the
default
behavior
that
we
always
grab
that
and.
D
F
Like
I,
I
just
think
there's
why
my
lips
sort
of
depleting
that
function
like
a
website,
it's
kind
of.
D
I
don't
think
it's
defined
and
it's
unclear
which-
and
they
can't
really
document
like.
Oh,
if
the
norm
has
v
mini
max.
It's
going
to
use
those
over
the
v
and
b
max
provides
the
function
or
the
other
way
around
it's
hard
to
get.
That
kind
of
behavior
is
undefined
and
that's
what
they're
trying
to
avoid.
H
D
F
F
D
I
don't
have
to
do,
but
if
you
do
want
limits
on
on
a
log
scale,
it
has
to
be,
it
should
be
provided.
When
you
create
the
norm,
you
can
create
the
norm
with
with
without
them,
in
which
case
it'll,
just
you
know,
automatically
choose
them,
but
if
you
do
want
to
give
it
limits,
you
need
to
provide
it
on
the
norm.
Well,.
E
E
D
E
D
A
D
There's
a
lot
of
changes
to
look
through.
I
have
I
tried
to
get
document.
I
think
I
put
in
all
of
the
renaming.
A
D
Specifically
into
the
comments
of
there,
like
I
said
before,
I
left
the
tests
as
untouched
as
I
possibly
could
just
so
that
way,
we
ensured
that
the
deprecations
were
being
handled
correctly
and
then
and
that
it
wouldn't
really
break
old
code
right
away
or
should
be
just
fine,
unless
the
only
thing
that
might
change
is
that
oh
yeah.
D
The
only
thing
that
might
change
is
that
if
people
were
importing
utility
functions
from
the
modules
that
they
were
defined
in
as
as
opposed
to
importing
them
directly
from
discretized.utils
that
will
kind
of
break,
but
if
they
were
importing
it
directly
from
discretized.utils
it'll
still
work,
just
fine,
which
is
what
everything
was
imported
as
in
in
simpeg.
It
was
imported
that
way.
So
it's
not
it
shouldn't
break
anything
in
something,
and
the
other
thing
is
like
I
said
the.
D
The
those
like
vector,
n
thing
properties,
the
shape
properties
as
they're
not
going
to
be,
are
going
to
be
tuples
instead
of
numpy
arrays.
So
you
won't
be
able
to
just
like.
The
point
is
that
okay,
those
are
actually
like
quantities
that,
if
somebody
got
back,
they
shouldn't
be
able
to
change
it
like.
If
somebody
asked
for
the
shape
of
the
mesh
before
you,
you
could
reassign
a
new
number
to
that
array,
which
kind
of
define
is
like
undefined
behavior,
so
we're
going
to
keep
them
as
tuples.
So
that
way,
people
that
way,
they're
immutable.
D
We
can't
change
them,
but
a
side
effect
of
that
is
that.
D
H
A
D
D
The
other
purpose
is
making
it
like.
If
you
look
at
the
way
the
source
code
is
organized
now,
I
think
it's
a
little
bit
more
clear,
like
okay,
we're
putting
the
emphasis
on
the
meshes
here.
Those
are
the
base
things
in
the
package
right,
so
those
are
the
things
that
are
easily
see
open
up
me:
okay,
if
you're
looking
at
the
source
code.
H
A
Excellent,
I
guess
what
else
do
you
want
or
need
from
folks
here
for
that
pull
request.
D
If
I
mean
I
guess,
if
people
wanted
to
take
it
in
and
start
using
it
and
make
sure
it
didn't
break
any
of
their
stuff
too
crazily,
that
would
be
good,
so
people
could
just
you
know,
install
it
like
use
it
locally
for
a
little
bit
and
see
if
it
it's
unmanageable.
E
Would
could
you
do
like
aloe
vera
beta
for
a
week
or
two
that
might
probably.
E
Yes,
yes,
we
did,
I
think,
in
in
this
joint
paper
video
we
had
the
help
of
bain
to
get
the
teacher
tetrahedra
meshes
out
to
matplotlib,
so
we
have
now
quite
exciting
results
that
compare
the
tensor
mesh,
octree
mesh
and
tetrahedra
mesh
cscm
results
and
the
corresponding
fields,
which
is
kind
of
nice.
I
don't
think
I've
ever
seen
a
comparison
like
this.
E
It's
not
that
easy
to
actually
plot
the
tetrahedra
mesh
in
matplotlib,
so
bain
promised
he
will
play
around
a
bit
further
to
improve
that
image.
It's
full
of
lines
which
is
not
really
nice
to
see
yet
yet
and
other
than
that.
I
had
this
chat
with
joe
and
soak.
It
helped
me
to
get
my
head
around
simpek
internals
and
I
created
a
branch.
If
anyone
is
interested
where
you
can
use
emg
3d
as
a
backend
for
simper
computations,
given
you
do
em
not
dc
but
em
at
a
certain
frequency,
so
yeah.
E
F
E
F
Right
so
have
you
like
tried,
like
let's
say,
you're
solving
e,
so
in
our
field
object
it
supports
that
to
compute,
b
or
h,
j.
Have
you
sort
of
tried
that
kind
of
functionality.
F
F
In
theory,
but
you're
using
a
slightly
different
sort
of
discretization
and
formulation,
so.
E
A
F
Still,
what
I
meant
is
slightly
different
in
a
sense,
so
that
field
object
gives
you
an
a
numerical
operator
that
can
evaluate,
let's
say,
b
from
e
or
j
fraud
and
that
kind
of
assumes
some
sort
of
formulation
that
we're
specifically
using.
So
I
I
wasn't
exactly
sure
how
I
mean
in
theory
that
might
work
well,
but
I
wasn't
sure
that
you
are
getting
the
expected
result.
I.
E
E
E
E
E
F
F
C
Okay,
I
guess
yeah.
E
D
What
what
were
you
using
to
do
the
tetrahedral
simulation.
E
In
the
comparison,
there's
custom
and
petcham
and
they
use
well
custom
uses
phoenix
and
patchem
uses
pet
sc.
I
think.
H
D
All
right
there's
something
that
I
brought
up
earlier,
but
whereas
thinking
of
adding
like
tetrahedral
mesh
functionality
to
discretize,
so
we
could
kind
of,
we
could
use
that
as
well
in
there.
So
it's
just
something
that
I
was
curious.
What
you
were
using
to
do
this
simulation?
Actually,
obviously,
it's
not
supported
in
simpek
right
now,.
D
E
D
Did
you
so
exciting,
but
I
mean
like
setting
up
the
actual
pde
that
you're
solving,
so
you
don't
know
how
that
was
done.
F
B
F
I
E
I
yeah
I'm
not
sure
like
like
custom
everything
that
they
use
with
phoenix
is
on
conda
but
pet
gem
with
peta
c.
They
it's
still
a
bit
more
involved
to
install
it,
because
I
think
in
their
install
instructions
you
have
to
compile
some
stuff
yourself
but
yeah.
A
Cool
thanks
dieter
any
other
is
that.
G
Yeah
sure
so
I'm
just
kind
of
finalizing
the
1d
stuff
yeah
I
added
dispersive
magnetic
susceptibility,
so
we
can
simulate
the
vrm
response
got
all
the
bugs
that
I
was
trying
to
fix
fixed
and
all
the
tests
passed
and
I've
added
new
tests
for
all
the
the
things
I've
added.
So
I'm
getting
close
to
a
point
where
I
think
I
may
want
to
bring
it
into
simpeg,
because
I'm
there's
some
stuff,
I
think
basic
utilities
and
analytic
functions
that
probably
already
exist
in
simpeg.
G
So
I
don't
exactly
want
to
perfect
it
here
and
then
bring
it
in,
because
I'm
going
to
have
to
kind
of
remove
any
duplicate
functions
and
some
other
stuff
anyway.
So
I
think
once
I
get
through
kind
of
finalizing
the
doc
strings
and
making
sure
that
the
formatting
and
variable
names
and
properties
are
all
done
in
the
same
fashion
as
simpag
I'd
like
to
bring
it
in
the
only
kind
of
remaining
item
for
this
is
the
the
multi-processing
part
that
seems
to
be
broken.
G
So
I'm
trying
to
to
fix
to
fix
that.
So
it
would
be
nice,
especially
for
the
time
domain,
because
it'll
be
faster
to
to
forward
model
each
sounding
and
compute
sensitivities
using
multi-processing.
So
I
just
need
to
kind
of
figure
that
out.
A
A
Yeah,
so
that
I
don't
think
that
needs
to
be
a
prerequisite
for
for
getting
it
into
into
synthetic,
you
should
be
able
to
import
it
directly
and
that'll
actually
make
it
easier
down.
The
road,
too,
does
that.
G
Make
sense
sort
of
I
mean
I
mostly
meant
that
there's
like
some
analytics
solutions
that
are
duplicates
of
analytics
solutions
there.
So,
but
it's
it's
like
tiny
details,
it's
not
something
that
fundamentally
changes
how
the
code
runs.
A
A
Yeah,
that
would
be
great,
I
think,
is
remove
as
much
sort
of
duplication
and
yeah,
because
then
then
it's
already
much
more
in
line
with
simpeg
before
before
thinking
about
merging
it
in
because
it's
going
to
take
some
thought,
I
don't
know
joe
what
your
take
is
because
simpek
em1d
has
fortran
code
in
it.
It's
it's
more
complicated
to
distribute
joe.
I
don't
know
what
you're
sort
of
taking.
H
G
D
Just
usually,
I
mean
it's
easier
to
distribute
c
code
than
it
is
for
train
codes.
D
There's
no,
I
mean
the
good
fortran
or
most
professional
compilers
and
windows
are
either
hard
to
get
or
they
kind
of
don't
work.
Natively
with
the
windows
kernel
or
something
like
just
weird
they're,
just
kind
of
weird
in
windows
I
mean
there's
definitely
ways
around
it.
Sci-Fi
has
I
mean
I
know,
sci-fi
uses,
fortran
they've
been
they've
done
a
lot
of
work
to
make
it
easier
to
use
it
than
what
it
does.
G
G
Fortran
that
fortran
90
file
is
going
to
need
to
be
updated
because
we've
added
a
frequency
dependent
susceptibility.
So
if
you're
doing
the
reflection
coefficient
for
the
te
mode,
it's
not
set
up
to
do
that
right
now.
So
that
needs
to
be.
D
H
H
G
Yeah,
I
guess
I
I
like
bringing
it
into
a
branch
of
simpeg
earlier
rather
than
later,
because
if
you
kind
of
try
to
perfect
it
too
much
when
it's
living
in
its
own
little
space,
you
might
have
to
undo
stuff
it's
it's
kind
of
at
a
point
where
it
could
be
brought
in
anyway.
I
mean
people
have
preferences,
but
it
feels
like
six
of
one
half
dozen,
the
other.
A
A
Maybe
we
just
come
up
with
a
little
bit
of
a
road
map
for
what
things
we
want
in
place
before
we
bring
it
into
simpeg,
because
I
think
this
this
question
about
code,
are
we
distributing
fortran
in
simpeg
that
that
is
actually,
I
think,
that's
an
important
question
to
to
address,
and
so
just
coming
up
with
a
with
a
bit
of
a
game
plan,
maybe
once
once
the
the
pull
request
is
in
and
evaluating
from
from
there
does
that
sound?
Does
that
sound
all
right.
G
Yeah,
I
think
that's
a
good
idea.
Definitely
before
we
bring
it
in,
it
should
be
reviewed
and
then
yeah
there's
still
a
few
details
that
I
need
to
finish
off
before
I'd
say
it's.
It's
ready
so
I'll
I'll,
complete
that
stuff.
I
want
to
fix
the
the
multi-processing
stuff
and
then
once
it's
done,
yeah
make
a
pull
request,
get
it
reviewed
and
when
everyone's
happy
we
can
take
the
next
step.
A
That
sounds
good
thanks
thanks
devin
joe,
I
don't
know
if
you
wanted
to
touch
a
little
bit
on
sort
of
documentation,
because
this,
I
think,
is
related
to
some
of
the
things
that
devin
might
be
working
on
next
or
we
can
table
that
and
chat
about
that
later.
What
would
you
or
maybe
overview.
D
I
guess
the
overview
is
that
we
would
like
to
move
to
a
it's
like
a
sort
of
numpy
style
documentation
of
functions.
It
just
makes
it
look
kind
of
it
makes
it
easier
to
read
the
text
strings
of
the
functions
and
easier
to
write
them.
D
Okay,
as
part
of
that
I'd
like
to
have
more
small
examples
of
each
of
like
the
functions
and
methods
being
used,
so
things
like.
Oh
how
what
does
this
you
know,
little
utility
function
actually
do.
What
is
this
function
actually
do,
because
it's
a
lot
of
the
utility
functions
are
just
like
not
documented
at
all
there.
D
So
I
think,
that'd
be
a
good
place
to
start
as
far
as
that,
especially
it's
probably
easier
to
start
on
discretize,
because
it's
easier
to
move
then
discretize
already
set
up
to
handle
them
by
documentation,
style
right
now,
so,
probably
easier
to
start
there
there's
also
less
utility
functions
and
just
kind
of
have
an
idea
of
okay.
This
is
what
we
want
it
to
look
like
when
we're
going
forward.
Part
of
that
eventually
it'll
we
start
with
that.
D
Documentation
is
not
nice
properties
package
and
we're
not
really
using
it
to
do
too
much
in
discretize,
because
all
we're
asking
it
to
do
is
basically
verify
the
inputs
to
a
couple
things
and
then
we're
using
it
to
serialize
some
of
the
base
meshes
or
some
of
the
mesh
or
serialized
meshes,
and
I
don't
think
it
would
be
too
hard
to
just
kind
of
have
that
like
implement
something
simple
that
does
that
ourselves
within
it.
D
D
So
that's
kind
of
looking
and
then
I'm
more
concerned
about
getting
like
the
api.
These
are
mostly
changes
on
the
api
documentation
side
of
it.
So
things
like
okay.
If
you
look
at
how
numpy
does
it,
they
don't
really
write
too
many
of
the
dedicated
api
documentation
files
in
like
the
documentation,
folder,
it's
right
at
throw
into
the
source
code
itself,
it's
written
in
module,
string,
yeah
it's
in
like
even
in
the
document
of
like
the
modules
themselves.
D
So
if
you
look
at
a
module
and
you
go
to
the
module
like
the
init.pi
module
or
function
inside,
these
modules,
you'll
see
the
entire
documentation
right
now
there,
which
is
kind
of
a
nice
way
like.
Oh,
I
import
this
module
and
I
can
see
all
the
things
that
it
does
like
you
know
right
away,
so
that's
and
then
it's
just
kind
of
like
listed
like
okay.
We
want
this
this
this
this,
so
that's
kind
of
the
way
that
we
I
I
see
this
all
going
forward.
D
As
far
as
the
api
documentation
documenting
each
of
the
individual
modules
themselves,.
G
Yeah,
I
think
that's
that
works
really
well,
and
the
only
thing
I
I
wanted
to
make
sure
is
that
there
was
some
separation
of,
I
guess
theory.
So
obviously
it'd
be
nice.
If
we're
solving.
Like
the
I
don't
know
the
electric
field
formulation
of
maxwell's
equations,
it
would
be
nice
to
just
say
this
is
what
you're
solving
this
is
the
system
you're
solving,
but
I
do
want
to
avoid
really
lengthy
mathematical,
derivations.
G
G
Yeah
yeah-
I
guess
now
that
you're
bringing
this
stuff
up.
I've
done
some
stuff
about
kind
of
yeah,
improving
the
organization
of
the
website
and
thinking
about
the
organization
of
auto
generating
the
api,
and
I
guess
one
thing
I
thought
for
that
would
be
beneficial
for
the
end
user.
G
Is
that
if
you
call
us
a
particular
survey
method
like
frequency
domain
electromagnetics,
you
should
have
all
you
should
be
able
to
import
all
of
the
relevant
stuff
from
there
and
so
right
now
things
like
gravity
and
magnetics,
the
the
analytic
dot
pi
exists
there
and
it's
all
kind
of
packaged
in
one
one
spot.
But
if
you
go
to
electromagnetics,
there's
just
a
separate
directory
that
houses
all
the
different
analytics,
but
they're
still
partitioned
by
survey
methods.
G
H
D
A
One
thought
there
too
is,
I
think,
and
we
can
maybe
sort
of
build
off
of
devin's
notes
here,
but
I
think
a
lot
of
the
analytics.
It
would
be
nice
if
we
actually
had
those
in
sitting
in
gom
is
because
they
don't
actually
depend
on
discretize
and
are
sort
of.
A
lot
of
them
are
useful
on
their
own
and
you
might
want
to
use
that
without
having
installed
simpeg
at
all.
And
so,
if
we
can
sort
of
maybe
take
take
a
look
at
all
the
analytics
that
we
have
and
and
see.
A
What
fits
over
in
in
geoana,
but
I
think
overall,
is
would
be
a
good
direction
for
where
analytics
might
wanna
might
wanna
live.
G
See
who's,
so
this
is
geoanalyst
or
just
joanna.
A
A
Our
source
definitions-
and
things
like
that
are
so
at
least
a
lot
of
the
very
simple
things
like
dipoles,
should
all
be
hidden
in
geoana,
some
of
the
more
complex
analytics
that
we
have
like.
I
threw
a
bunch
of
stuff
in
there
for
the
casing
stuff
like
that.
That
should
probably
get
moved
over,
but
those
are
still
sitting
in
sync.
G
H
G
Okay,
that
makes
perfect
sense
to
put
all
that
in
there
yeah
then.
The
other
thing
I
was
kind
of
curious
about
was
we
had
so
we
have
this
thing
called
io
and
we
use
that
for
for
dc.
Does
anyone
have
a
one
or
two
sentence,
rigorous
definition
of
what
io
is.
F
G
Too
yeah,
so
so
as
of
right
now,
what
it
is
and
what
it
can
do
isn't
really
well
established.
We
don't
have
to
get
to
it
right
now,
but
it
seems
like
we
want
something
that
that
im
imports
and
exports
things
and
at
a
high
level
interacts
with
the
simpeg
code.
Would
let
us
do
like
a
workflow
almost.
F
Yeah,
so
I
think
that
we
need
some,
I
kind
of
like
that,
having
that
kind
of
class
that
you
can
in
the
basically
outside
of
the
problem
or
simulation
class
and
that
can
readily
generate
input,
parameters
and
visualize,
simple
data
or
inversion
models
or
whatever,
but
we
need
something
different
name.
I
guess.
A
Yeah,
I
think
a
lot
of
that
is
part
of
the
bigger
conversation
and
probably
the
next
phase
of
development
in
a
sense
of
like
what
are
the
high-level
apis
that
sit
on
top
of
simpeg.
Yes,
in
potential
fields,
we
have
some
of
the
concept
of
drivers,
which,
I
think
is
you
know
it's
in
a
similar
vein
as
to
what
the
I
o
currently
is,
and
so
yeah,
it's
another
another
big
conversation,
but
absolutely
an
important
one,
because
that
is
actually
what
drives
usability
for
a
much
broader
audience.
G
Okay,
so
then
I
I
guess
in
the
meantime,
then
I
guess
I
should
be
part
of
the
the
gio
anna
group
and
I
should
be
able
to
go
and
I'm
not
sure
if
I'm,
if
I'm
not,
I
need
to
get
added
to
that
and
then
I
could
contribute
towards
moving
the
analytics
solutions
into
joanna
so
that
we
can.
We
can
call
them
instead
of
redefining
them.
A
Yeah,
if
it
works,
maybe
I
don't
know
joe
and
devin.
If
you
both
have
time
tomorrow,
we
can
sort
of
maybe
go
through,
because
you
we've
brought
up
like
a
whole
bunch
of
different
things.
Some
of
them
are
related
or
not
and
go
through
and
maybe
prioritize
a
few
items
with
that.
Would
that
work
either
tomorrow
or
friday.
We
can
figure
out
a
time
on
on
slack,
but
that
seems
like
maybe
a
productive
step
yeah.
I.
G
Need
to
work
around
our
normal
meeting
with
doug
and
roman,
but
that's
fine.
Tomorrow's
good
sooner,
the
better
yeah
free.
A
Excellent
that
sounds
good.
Okay,
well,
we'll
coordinate
afterwards
on
on
a
time,
excellent,
cool
soggy.
Would
you
like
to
go
next.
F
Very
I
have
nothing
to
really
update
that,
although
I
actually
interact
with
peter
a
little
bit
about
how
to
implementing
jvec
deeper
has
a
jt
value
in
his
adjoint
problem.
But
I
think
I
got
pretty
good
idea
of
actually
getting
jay
back
from
pieces
of
his
code,
so
yeah
actually
implementing
both
jtv
and
jpeg
into
simpac
running.
F
With
actually
using
emg3
as
a
backend,
then
sort
of
syntax.
A
F
Rapper
would
be
really
cool,
it
will
be
a
cool
application
anyway.
Other
than
that,
I
was
actually
in
an
interesting
conversation
in
these
days
actually
recommending
how
to
use
airborne
geophysics
in
central
valley
of
california,
so
this
california
government
actually
now
launching
much
larger
project,
basically
entire
central
valley.
So
it's
about
12
million
dollar
project
so
and
they
are
actually
taking
the
recommendation
from
this
project
and
putting
that
into
their
sort
of
rfp.
F
A
Excellent
one
thought
just
to
put
on
the
radar
with
the
emg
3d
connection
this
could
be.
I
think
this
would
be
a
really
cool
paper
at
some
point.
I
don't
know
where
it
would
go,
but
like
sort
of
that
that
level
of
interoperability
between
between
different
code
bases,
I
think,
is
not
super
common,
especially
like,
because
they
did-
and
you
know,
start
from
different
streams
or
I
know
it
could
be
kind
of
a
cool
cool
one
to
put
on
the
radar.
E
E
D
A
Thing
excellent
tebow.
I
Oh
yeah,
basically
I'm
almost
done
with
this
this
correction,
so
I'm
just
going
through
one
last
time
and
get
that
out
of
the
way
by
like
end
of
the
week
or
early
next
week,
so
excellent
to
be
like
done.
A
Excellent
dumb,
are
you
able
to
chat
with
you.
C
Yeah
not
much
to
say
either
we
had
the
whole
conversation
right
with
the
with
the
subsurface
members,
and
that
made
me
thinking
a
little
bit
of
how.
How
could
we
integrate
it?
A
little
bit
closer
with
the
with
our
that,
for
instance,
like
data
slash
survey
objects.
C
If
we
could,
you
know
serialize
them
and
as
a
subsurface
object.
That
would
be,
I
think,
pretty
efficient
way
to
do
it
and
other
than
that
yeah,
it's
just
the
tiling.
That
is
still
on
still
on
my
on
my
list
and
it's
just
a
pretty
busy
week.
So
I'm
gonna
get
back
to
it.
Only
I
guess
mid
next
week,
then
I'll
be
a
little
bit
freer.
That's
about
it!.
A
Excellent
thanks
john
yeah
also
go
ahead.
Zaggy.
F
E
And,
and
also
because
what
joe
mentioned
about
the
discretized
serialization,
ideally
is,
they
could
then
also
serialize
to
a
subsurface
element
or
object
or
whatever
they
call
it.
Then
yeah.
D
E
Out
yes,
well
with
payne
and
miguel,
we
might
post
it
rather
sooner
than
later,
for
the
people
who
are
a
bit
interested
in
in
the
details,
details
of
the
format
and
then
in
the
biggest
round,
probably
early
december.
We
said
before
the
end
of
the
year,
something
like
this,
but
yeah.
It
will
be
in
the
channel
over
at
svang
in
the
channel
subsurface
for
anyone
who's
interested.
A
Okay,
yeah,
if
you
could
ping
a
few
of
us,
I
try
and
keep
an
eye
on
that.
But
if,
if
my
name
is
mentioned,
I
will
definitely
take
a
look
because
yeah
these
are
exciting
conversations.
Yeah.
E
A
Cool,
yes,
I
guess
maybe
a
couple
updates
from
me
is
last
week
I
had
a
chat
with
the
coils
folks,
so
matt,
rocklin
and
hugo,
and
so
got
the
buchanong
inversion
up
and
running
with
desk
on
shine,
which
was
really
cool.
Is
that
I
was
able
to
run
all
thousand
soundings
in
parallel
and
it
was
done
in
like
two
or
three
minutes,
which
was
just
really
satisfying,
and
that
was
using
sort
of
the
the
sill
mesh
on
simpeg.
A
So
we
had
written
that
example
before
em1d
existed,
but
it's
kind
of
encouraging
to
see.
That
is
that
it
is
then
easy
to
sort
of
set
up
a
simple
inversion
with
simpeg
and
get
that
up
and
running.
There's
some
subtleties
with
how
you
set
up
the
cluster
so
that
you
don't
get
the
threads
running
over
each
other.
There
were
a
few
different
layers
of
things
to
sort
out
there,
so
I
did
actually
talk
a
bit
about
that
in
the
in
the
chat
with
them.
I
think
it's
all
on
youtube.
A
If
you
are
interested
in
in
watching
I'm
going
to
try
and
clean
up
that
notebook
so
that
it
actually
is
a
little
more
usable
on
other
machines
as
well
like
right
now,
it's
tied
exactly
to
cheyenne
but
I'll,
try
and
get
an
example
running
where
you
can
run
it
locally
and
then
maybe
on
nurse
as
well,
which
is
the
lbl
computer.
So
yeah
we'll,
hopefully
get
a
few
examples
out
of
it's.
That.
C
A
A
Yeah
well,
this
one
is
easier
because
it
just
is
totally
separable
yeah
yeah,
and
then
I
guess
so.
The
other
thing
is
I'm
starting
to
do
a
bit
of
work
with
teddy
who's
at
ufc,
working
with
adam
pidlaseki
she's
interested
in
mostly
dc
examples
for
monitoring
and
she's
got
kind
of
a
cool
problem.
It's
looking
at
monitoring
beneath
permafrost,
and
so
you
have
this
surface
layer
that
you're
pretty
sensitive
to
that
that
changes
depending
on
the
season.
A
So
if
it's
in
the
summer
it's
going
to
be
more
melted
and
if
it's
in
the
winter,
it's
going
to
be
much
more
resistive
because
it's
all
frozen
and
then
trying
to
monitor
a
contaminant
beneath
that,
and
so
we're
going
to
play
around
with
combining
some
of
the
work.
Dom
did
with
the
sum
map
to
basically
set
up
a
like
parametric,
plus,
smooth
inversion
components
so
invert
for
a
parametric
description
of
the
background
and
then
have
the
contaminant
be
a
smooth,
smooth
piece.
A
So
we'll
keep
you
posted,
I'm
going
to
try
and
get
a
minimal
example
up
and
running
with
her
in
the
next
couple
of
days,
so
should
be,
should
be
a
fun
example
to
play
around
with
kind
of
yeah
showing
off
the
mappings
and
so
it'll
be
cool
yeah.
A
And
then
I
think
that's
the
other
thing,
and
this
is
a
longer
conversation
than
two
minutes,
but
the
other
thing
that
I've
started
to
be
thinking
about
again
is
thinking
about
what
we
want
to
sort
of
have
written
down
in
terms
of
governance
of
simpeg.
As
just
like
you
know
right
now.
A
I
think
a
lot
of
our
decision
making
process
happens
at
the
meetings,
which
is
great,
but
I
think
we
want
to
sort
of
maybe
at
least
document
that
somewhere
so
I'll,
try
and
get
a
few
thoughts
down,
but
then-
and
we
can
chat
about
it
and
I'll
share
some
links
and
stuff
around.
But
if
there's
other
things
that
you
think
we
should
have
sort
of
somewhat
formalized
with
simpeg,
I
mean
I
don't
think
we
need
to
be
like
implementing
new
processes.
A
At
this
point
in
time
of
like
yeah,
I
mean
I
think
overall,
things
are
operating
well,
but
it's
if
there's
things
that
you
see
that
we
could
be
doing
better
just
in
terms
of
how
we
manage
community
and
manage
changes,
and
things
like
that
or
if
there's
stuff,
that
you
think
we
should
be
formalizing
in
terms
of
people's
roles
and
things
like
that.
A
I
could
see
that
being
useful
as
having
a
bit
of
role,
definition
and
things
like
that,
both
for
for
ourselves,
but
also,
I
think,
well,
for
ourselves
sort
of
within
the
community.
But
I
also
think
for
folks
to
be
able
to
advertise
and
speak
about
what
you
do
and
what
you
contribute
to
simpeg.
It
would
be
nice
if
we
have
a
bit
of
language
around
that
and
kind
of
definition
as
to
as
to
what
that
means.
A
Yeah
so
give
something
some
thought
if
you
have
ideas,
feel
free
to
just
send
things
around,
but
I'll
try
and
get
a
bit
of
a
document
or
some
issues
and
stuff
like
that
sketched
out.
So
we
can
have
a
bit
of
a
more
structured
conversation
around
that
in
the
coming
weeks.
A
All
right:
well,
I
guess
we're
on
the
hour
for
half
hour.
Does
anyone
else
have
other
burning
items
things
you'd
like
to
bring
up
before
we
wrap
up.
H
Is
anyone
interested
in
doing
another
coding
hacking
thing
at
some
point
soon?
It's.
D
E
Wednesday
afterwards,
so
we
find
friday
night
is
a
bit
late
to
me,
but
yeah.
I
thought,
and
it
probably
also
goes
into
lindsay's
thing.
I
don't
know
what
like
what
would
be
the
approach
in
discretize
and
simpak.
There
are
like
pull
requests
from
people,
sometimes
even
outstanding.
No
one
has
replied,
and
I
don't
know
myself.
Should
I
not?
What
is
the
government?
Should
we
do
it?
Do
we
leave
it
to
see
what
happens
so
these
yeah?