►
From YouTube: SimPEG meeting Dec 8, 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
So
I
guess
one
quick
item
and
I'll
probably
circle
back
to
this,
if
there's
more
folks
who
pop
in
later,
wanted
to
check
and
see
if
the
simpe
governance
meeting
conversation
tomorrow,
if
that
works
for
folks
at
11am
pacific,
I
got
a
few
there's
a
few
thumbs
ups
on
slack,
just
a
question
of
like.
Is
there
anyone
who
it
doesn't
work
for?
A
Okay,
it
looks
like
mostly
thumbs
up
so,
let's
plan
for
that,
and
I
will
send
out
a
calendar
entry
with
the
zoom
link
and
we
can
go
from.
We
can
go
from
there
and
I'll
try
and
send
up
like
a
couple
sort
of
agenda
items
with
that
too.
To
give
us
a
bit
of
structure,
cool
okay,
who
wants
to
kick
things
off.
B
That
shows
that's
a
kind
of
my
first
pass
at
removing
properties
from
discretize
the
properties
package,
just
to
be
able
to
have
a
little
bit
more
control
over
the
documentation,
like
that.
That's
a
primary
motivation
as
well
as
it's
not
super
reliant
on
it
anyway.
So
it's
an
easier
ask
to
be
able
to
remove
it
from
that.
So
we
can
control
the
documentation.
B
There's
a
few
small
little
things
in
there
that
I've
been
that
I
need
to
catch
a
few
small
little
bugs
that
are
still
there,
mostly
with
how
you
can't
so
some
of
the
properties
on
meshes
are
like
numpy
arrays
and
you
can't
serialize
a
numpy
array
directly
or
at
least
do
a
json
to
json
directly.
B
The
pi
data
theme
for
the
numpy
or
for
the
documentation
on
discretize,
so
I'd
like
to
show
that
really
quick.
If
I
can
absolutely
to
kind
of
get,
I
showed
a
little
screenshot
before
on
I'm
slack,
but
I'll
share
this
really
quick.
B
So
this
is
kind
of
what
it
what
I
was
thinking.
The
phi
data
theme
was
working
looking
like
there's
a
few
things
that
I'd
like
to
change
around
here,
still
I'm
not
quite
sure
yet
how
to
switch
the
twitter
icon
to
be
like
a
black,
black
and
or
grayscale
one
I'd
like
to
make
up
a
grayscale
image
of
the
simpeg
logo
with
discretized
next
to
it
just
to
be
able
to
have
it
on
there
as
well,
because
grayscale
looks
better
on
this
black
on
this
background.
B
B
B
B
I
I
just
particularly
like
that
that
that
the
it
looks
the
same
similar
as
the
you
know,
simple
xyz
format,
website.
C
Yeah,
it's
really
clean
man
and
it's
it
looks
quite
modern
right.
It's
aired
out
space
or
it's
not
too
cluttered
with
a
bunch
of
banners
and
stuff.
So
yeah.
B
B
It
kind
of
it's
really
interesting
like
how
it
works
as
you're
scrolling
through
things.
So,
if
you
look
up
over
on
the
right
hand,
side
as
I'm
scrolling
through
this
list,
it's
updating
itself
where
it's
at.
B
Can
you
go
and
click
on
base
match
yeah,
that's
fine
yeah,
so
base
mesh
I've
gone
and
kind
of
updated
a
little
bit
of
the
internal
documentation
on
it
to
look
more
numpy
like
or
to
get
numpy
style
documentation
for
the
bass
mesh
as
an
example,
that's
beauty,
so
they
all
cross
ref
like
shape
cells,
gonna
switch
to
so
yeah
now
so
before
the
so.
This
is
what
this
documentation
looks
like
for
this.
A
B
Yeah
it
doesn't
it
kind
of
it's
a
like
a
table,
content
scroller,
so
you
only
really
see
it
in
sometimes
so
when
you're
over
here
in
this
function.
So
this
is
as
you're
scrolling
through
it.
It
tells
you
where
you're
at
in
the
rst
document.
Basically
it
doesn't
work
very
well.
It
doesn't
really
pop
up
in
the
numpy
doc
generated
things,
so
this
page
was
generated
by
like
an
auto
summary
and
auto
generate.
B
B
A
B
That's
about
it
other
than
that
you
can't
see
because
she
is
quarantining,
but
my
wife
is
in
there.
C
B
B
A
All
I
got
excellent
well,
that
is
that's
really
exciting
news,
joe
I'm
so
pleased
she's
made
it
and
you'll
be
done
quarantining
by
christmas.
So
that's
that's
excellent.
A
Perfect,
I
see
dom
next
on
my
screen
tom.
Would
you
be
willing
to
go
next.
A
Yeah
sure
no
problem
tebow.
G
Yeah,
hello,
everyone
just
one
thing
not
simply
related
to
start
with
I've,
finally
updated
the
ubc
gif
website
for
publication
thesis
and
a
portion
of
the
proceedings
with
presentations.
So
if
you
want
to
check,
if
I
didn't
forget
your
publication
or
that,
if
you
want
to
send
me
a
presentation
that
you
gave
and
I
I
can
put
that
on
the
proceedings,
just
yeah,
just
let
me
know-
but
yes,
the
ubc
gif
website
had
not
been
updated
since
2017.
G
But
yeah
I
might
have
forgot
one
or
two
a
a
few.
Let
me
know
I
put
everything
I
was
aware
of,
but
I
don't
guarantee
there
is
no
omission
more
on
simpag
I've
added
yet
another
pr
to
the
to
the
to
the
pile
this
one
is
I've
actually
added
like
a
tutorials
for
a
joint
grab
and
mag
inversion
with
petrophysical
coupling
and
on
the
noctree
mesh,
and
there
is
two
inversions
in
that.
G
Like
I
discussed
with
devon,
we
might
see
parents
that
in
two
tutorials
and
one
is
like
when
you
provide
petrophysical
information
and
the
other
one
is
when
you
don't
have
prior
petrophysical
information
and
how
how
you
you,
you
use
both
how
you
can
do
that
with
with
the
code
and
it's
actually
fairly
fast.
So
it's
like
both
inver,
like
both
inversion
together,
takes
six
minutes
on
my
laptop
to
run.
G
So
so
that's
so
I'm
happy
with
I'm
happy
with
the
result
and
it's
all
it's
again
like
the
synthetic
dkc
but
yeah
so
like
it's
close
to
the
publication
too.
So
that's
that's.
Gonna
be
useful
and
other
than
that
working
to
get
all
my
other
players
moving
forward.
So
joe
has
been
helping
me
a
lot,
starting
with
moving
the
the
beta
estimate
one.
G
So
one
thing
for
everyone,
too,
is
like
I've,
checked
all
the
examples
and
tutorial
and
updated
the
beta
is
made
values
so
that
it's
very
similar
to
what
was
there
before.
So
so
like
this
updating
that
directive
should
not
change
strategy
drastically.
What's
the
example
that
was
already
in
there,
that's.
A
I
A
Yeah
tebow,
would
you
mind
maybe
connecting
with
xiao
long
and
just
pointing
him
to
sort
of
like
what
sorts
of
things
you
changed
so
that
we
can
kind
of
think
through
how
to
because
I
think
I
suspect
that
the
solution
that
we
needed
for
for
either
scenario
for
either
of
your
work
should
should
serve
the
other,
so
just
kind
of
making
sure
that
what
what's
been
done
on
your
pull
request
hopefully
serves
his
pull
request.
G
Well,
I
from
what
I
remember
looking
through
the
cross
gradient,
pull
request.
I
I
saw
that
directive
and
he
was
basically
using
the
same
like
the
same
thing
as
a
previous
beta
estimate.
G
So
in
my
opinion,
it
will
need
to
be
updated
to
at
least
like
use
that
that
new
eigenvalue
utilities-
that's
that
would
be
the
first
step
and
after
that
we
can
see,
I'm
I'm
not
sure
yeah,
I'm
not.
I.
I
don't
remember
what
was
a
big
big
difference
so
after
that
I
think
when
we
work
through
the
the
the
course
gradient
pull
request,
see
what
can?
G
What
can
we
merge
with
other
directives
or
how
to
create
a
new
directive?
It
really
depends
on.
I
guess,
on
how
we
we
formulate
the
inverse
problem
and
what
we
consider
as
a
trading
as
a
trade
of
parameters
or
alpha
scaling.
So
I
think
that
was
a
main
difference,
but
yeah
like
just
as
of
now.
What
I
remember
is
that
it
was
very
close
to
the
former
formulation,
so
updating
it
to
use
the
new
utility
will
be
a
first.
A
Step
yes,
so
in
terms
of
maybe
a
bit
of
a
suggestion
is
how
to
make
that
a
bit
concrete
and
joe
feel
free
to
jump
in
if
you've
got
other
thoughts
on
this
too.
But
when
tebow,
when
you're
happy
with
the
beta
estimate
pull
request,
would
it
be
possible
to
maybe
start
a
pull
request
onto
xiao
long's
work
to
show
or
like
start
to
work
towards
bringing
like
bringing
just
that
one
piece
a
little
closer
together?
Would
that
make
sense
as
sort
of
the
next
a
next
step.
B
I
I
think
that
would
be
a
good
idea,
I'm
just
looking
through
it
again
on
there
as
well.
Right
now,
again,
just
to
remind
myself,
it
does
touch
a
lot
of
directives.
I
think
there's
a
few
things
that
are
commented
out
that
probably
shouldn't
like
just
it
looks
like
they
were
changes
that
were
meant
for
your
local
stuff
on
there
that
were
like
just
bracketed
commented
out.
That
probably
should
be
looked
at
again.
A
Yeah
and
at
least
to
me
this
seemed
like
a
good
sort
of
first
step,
because
at
least
this
one
is
well
scoped.
So
even
if
there's
other
stuff
that
was
commented
out-
and
things
like
that,
that
need
to
be
resolved
later,
at
least
like
getting
onto
sort
of
a
common
beta
estimator,
is
perhaps
a
good
step
step
forward
in
bringing
bringing
those
works.
Closer.
C
It
looks
really
really
good
results.
It's
pretty
amazing.
I
have
one
question
I
guess
related
to
this.
For
for
joe
or
lindsay,
is
it
possible
to
still
have
those
tutorials?
You
know
built
from
a
pie
file,
but
just
try
to
move
all
the
flooding
functions
out
of
it
or
somehow
hide
it,
because,
right
now,
like
you
know
those
tutorials,
they
they
look
super
long,
but
there's
so
much
of
it.
There's
just
flooding
functions
just
unrelated
to
what
we
want
to
talk
about.
I
don't
know
if
it's
possible
to
somehow
hide
those.
J
Yeah
I
mean
some
of
it
is
just
being
a
bit
clever
and
finding
ways
to
kind
of
compress
it
or
write
the
code
more
compactly.
Another
thing
that
can
be
done
is
with
the
way
that
you
comment
it.
You
can
partition
out
the
parts
of
that
pi
file
into
sub
sections
in
the
website,
so
you
can
put
a
heading.
That's
like
all
right!
We're
plotting
this
now.
J
J
Yeah,
that's
true.
I
find
it
sort
of
useful
to
see
some
of
I
mean
yeah.
You
don't
want
to
have
some.
You
don't
want
to
have
like
80
percent
of
the
the
pie
file
being
it's
completely
unnecessary
utilities,
and
if
we
are
reusing
some
stuff,
we
might
be
able
to
just
add
that
plot
style
into
plotting
utilities.
That
might
be
a
better
solution,
but
I
find
it
kind
of
useful
to
be
able
to
go
and
see
how
certain
plots
were
were
generated.
A
L
J
Well,
that
would
that
would
be
super
awesome
too,
because
the
way
it
generates
on
the
website
you'd
have
some
kind
of
heading
or
subheading.
You
would
have
text
that
gets
generated
there
and
then
usually
below
you
have
the
portion
of
code.
However
long
it
is,
and
if
you
just
had
a
button
to
expand
that
would
actually
really
really
shorten
those
pages.
That
would
be
pretty
sweet.
G
But
yeah,
I
agree
with
you
dom.
It
was
one
of
my
concerns.
Writing
is
the
tutorial
because,
like
for
example,
when
I
want
to
plot
the
the
gaussian
mixture's
probability
distribution,
we
don't
have
any
utils
for
that.
So
I
I
end
up
with
having
like
a
lot
of
lines
to
trying
to
plot
that-
and
I
put
some
concern
right
now
to
like
try
to
have
it
compact
but
yeah.
It's
true
that
it's
very
long
and
one
like
one
quick
solution.
G
I
can
see
right
now,
will
be
like
I
already
download
like
the
mesh
and
the
data
etc.
So
it's
not
in
there.
I
can
probably
imagine
that
we
can
download
like
a
dot
by
file
and
then
add
it
to
your
pass
and
have
utilities
in
that
python
file
that
you
download
that's.
That
could
be
one
option
too.
I
could
try
that
this
afternoon
and
see
what
people
think
of
that
as
a
as
a
quick
solution.
If
that's
that's
that
sounds
good
to
people.
A
Soggy,
just
posted
a
link
that
I
think
is
worth
checking
out
and
joe.
I
don't
know
if
you've
actually
seen
any
solutions
like
this,
but
I
think
with
this
before
we
come
up
with
sort
of
a
complicated
solution
of
like
downloading
python
files
that
just
contain
like
a
tiny
utility.
Let's,
let's
do
a
bit
of
a
search
on
this,
because
I'm
sure
somebody
solved
this
problem
and
so
yeah.
Let's
not,
let's
not
invent
something
before
before.
B
It
looks
pretty
similar
as
I've
been
working
through
css
stuff
and
then
customization
stuff
for
the
the
documentation
already.
D
Yeah
one
thing
that
I'm
not
sure
like
when
you
general
auto
generate
the
doc
from
the
notebooks.
I
F
All
good
now
yeah,
nothing
too
too
much
to
report
other
than
yeah.
I've
got
the
dcip
parallelized
they're,
sorry,
the
tiling
parallelized,
so
it's
doing
similar
to
what
dom's
got
going
and
I'm
in
the
inversion
part.
So
there's
just
a
few
little
like,
I
think
it's
just
yeah,
there's
some
mismatches
between
dimensions,
I'm
getting
there
so
that
should
be
running
by
next
week
or
by
the
weekend.
Hopefully
yeah
and
I'm
gonna
help
out
with
the
2d
mt
and
and
get
into
the
amt
stuff
yeah.
F
I
don't
know
if
we
want
to
include
that
in
this
pr
or
should
I
should
we
maybe
wrap
up
that
first
pr
first
and
then
make
this
something
new,
I'm
open
for
suggestions
and
also
on
the
csa
mt
side
dealing
with
the
near
field.
How
important
is
that
and
yeah
how
to
tackle
it
if
anybody's
got
experience
or
or
good
recommendations.
D
F
Well,
yeah,
that
was
again,
I
was
going
to
help
out
with
the
2d
mt
so
that
well
because
most
of
the
applications
that
are,
I
guess,
the
applications
yeah
that
have
been
tendered
are
mostly
2d
cs
amt.
So
I
thought
getting
the
2d
mt
up
and
going
and
then
should
be
after
that
we
can
incorporate
the
cs.
Amt
receivers
was
the
idea
anything
I
see.
F
I
see
if
yeah
yeah,
lindsay's
kind
of
gotten
me
up
to
date,
I've
gotta
I've
got
some
reading
to
do,
but
yeah
we
should
be
able
to
work
out
those
boundary
conditions
for
the
2d
stuff
and
yeah.
I
think
that's
what
I'll
aim
for.
F
A
And
john
and
I
have
gone
back
and
forth
a
bit
on
a
notebook
and
on
my
list,
is
to
write
out
the
2d
derivation,
and
so
I
know
soggy's
done
a
notebook
too
so
between
between
all
of
us.
We
can
compare
and
hopefully
get
that
running.
I
think
it's
it's
really
close
right
now
I
was
showing
john
yesterday.
It's
really
close
like
we're
getting
the
right
apparent
resistivity
in
the
middle
of
the
mesh,
but
near
the
boundaries
it's
wrong
so
and
it's
also
expanding.
A
So
I
think
I
miss
we're
missing
a
volume
term
or
an
integration
term,
or
something
and
so
I'll
write
down
what
we've
done
and
send
that
around
and
if
we
can
get.
You
know
a
few
sets
of
eyes
on
that.
That
might
be
helpful.
So.
D
D
H
The
battery
conditions,
but
did
something
wrong
something.
A
On
that
note
for
csamt,
I
don't
know
if
anyone
else
has
recommendations
for
papers
that
we
should
look
at
or
that
modeling
nick
williams
had
asked
a
couple
questions
about
kind
of
getting
up
and
running.
I
mean
we
haven't
tested
any
of
that
or
sort
of
tried
that
out
with
simpeg.
But
if
there's
any
implementation
papers
that
are
worth.
D
If
the
assumption
is
as
long
as
the
assumption
is
hold,
I
don't
really
see
other
issues
of
just
using
the
2d
empty
code.
First
to
as
you
as
the
john
mentioned,
like
once
the
near
field
comes
in,
you
cannot
really
assume
that
as
a
boundary
condition,
you
just
need
to
put
the
source
in
to
the
mesh.
Then
like
it
is
like
a
typical
frequency
domain
problem,
so
you
need
to
use
the
actual
active
source
and
then
treat
the
data
as
an
impedance.
I
think
yeah
yeah.
F
The
we're
talking
with
some
of
the
geos
in
this
here
who
are
who've
done
mt
and
stuff.
They
say
that's
kind
of
like
what
the
toss-up
is.
They
don't
know
when
you're
going
to
be
near
field
and
far
field
like
you
can
you
can
assume
all
you
want,
but
sometimes
they're
still
like.
I
don't
know.
He
said
it's
very
hard
to
tell
so
it's
nice
to
have
the
the
near
source
or
sorry,
the
near
field,
kind
of
derived
to
see
what
it
what
it
does,
but
yeah
we're
like
the
most.
D
D
D
One
one
thing
to
sort
of
think
about
is
we
see
like
a
latest
ubc
code?
Does
the
projection
because,
what's
actually
kind
of
nice
in
the
way
ubc
code
works,
you
don't
need
the
magnetic
field,
so
you
can
just
use
the
electrical
field
to
derive
like
because
it
just
evaluate
the
voltages.
So
you
don't
need
to
actually
like
you,
don't
need
to
store
or
compute
the
h
from
e,
so
you
can
directly
from
a
project
that
from
electrical
field.
D
So,
if,
like
the
projection,
is
kind
of
cumbersome
in
the
current
frequency
code,
just
stick
to
just
use
the
electrical
field
and
develop
a
projection
from
electrical,
they
can
compute
the
voltages
and
then
compute
the
ratio.
I
think
it'll
be
it'll,
be
useful
and
that
sort
of
like
gives
you
a
little
bit
more
better
efficiency,
so
consider
that.
J
If
you
want
some
background
theory
for
that,
I
can
point
you
to
the
theories
section
of
this
mt
code
and
it
sort
of
explains
how,
in
general,
we
simulate
this.
M
M
So
we
have
some
ground
gravity
data
as
well,
and
we're
just
kind
of
playing
around
with
that
and
going
to
do
some
kind
of
upward
continuation
of
that
ground
gravity
data
which
has
already
been
terrain,
corrected
up
to
kind
of
the
elevation
of
the
airborne
data
set
and
see
how
those
compare
and
try
to
get
some
idea
of
the
impact
that
the
topography
is
having,
and
I
was
also
just
thinking
of
doing
some-
some
basic
forward
modeling
with
that
airborne
data
set
with
kind
of
different
sized
cells
and
different
refinement
rates
on
the
topography,
to
kind
of
see
how
that
changed.
M
Our
forward
model
data,
so
just
yeah-
I
was
talking
to
dom
yesterday
and
we
were
kind
of
working
on
putting
together.
I
was
just
kind
of
using
the
example
that
he
had
run
a
long
time
ago
to
put
together
a
an
equivalent
source
solution
for
upward
continuation
of
those
fields,
so
just
playing
around
with
that
on
a
3d,
octree
mesh
and
then
also
going
to
try
messing
around
with
it
on
a
kind
of
a
quad
tree
mesh
to
see
the
difference
in
those
two
results
and
just
kind
of
differences
in
efficiency.
M
M
Yeah
yeah
we're
we're
trying
to
figure
some
of
those
things
out
and
just
kind
of
get
a
handle
on
that.
I
was
also
talking
with
dom
about
that,
looking
at
the
way
that
they
had
kind
of
done.
Some
of
that
with
the
data
from
I
think
it
was
from
iran
that
mason
was
working
on.
A
Thanks
mike
devin.
J
Sure
so
I,
if
there's
anyone
who
has,
I
guess,
shorter
updates,
we
could
let
them
go
first.
I
remember
last
week
you
guys
asked
me
to
maybe
give
you
the
10
minute
tour
of
the
discretized
theory
sections.
So
I
don't
know
if
we
want
to
get
to
anything
else.
First
or
whether
or
not
I
should
just
sort
of
go
for
it.
A
Will
it
be
sort
of
ten
yeah
if
it's
gonna
be
like
yeah
ten
minutes,
then
yeah?
Let's,
let's
continue
around
and
we'll
we'll
circle
back.
If
that's,
okay,
yeah.
A
Perfect
thanks
dev
xiaolong
yeah.
I
Sure
I
I
don't
have
anything
to
report
these
times,
but
for
the
new
beta
estimation
method,
I
will
check
it
to
see
whether
I
can
use
this
part
of
the
code.
Here.
I
have
a
question
for
the
teapot.
So
can
the
new
method
estimate
the
two
beta
values
say
for
one
for
gravity,
one
for
magnetic?
At
the
same
time,
okay,.
G
So
right
now
like
in
that
pr,
I
wrote
like
directives,
that's
like
that
are
like
when
you
have
like
data
misfits
you
can
have
one
or
several
and
you
have
a
regularization
with
alphas,
so
it
and
one
bit
one
beta.
G
So
I
think,
what's
your
what's
what's
different
in
your
estimate
is
with
your
beta
is
that
you
have
different
beta
for
grab
and
mag,
but
the
way
it's
written
right
now,
it's
there
is
actually
actually
now
a
utils
in
mad.utils,
where
you
where,
when
you
give
it
any
objective,
function
term
like
a
data
misfit
or
regularization,
it's
gonna
output,
you
the
an
estimation
of
its
eigenvalue,
so
what
you
can
do,
and
so
that's
so
this
type
of
ratio.
G
You
actually
doing
it
right
now
within
your
directive,
but
I
showed
in
that
notebook
that
I
put
in
the
pr
that
when
you
do
a
when
you
do
that
estimation,
a
single
iteration,
it's
it's
actually
like
it's,
not
very
a
very
good
estimator.
G
So
what
you
can
do
is
that
you
can,
within
your
directive
instead
of
keep
computing
the
eigenvalue
within
the
directive
with
a
comp
with
a
with
a
relay
value.
You
actually
just
call
that
utility,
so
that's
going
to
be
the
change
to
happen
is
that
you're
gonna
instead
of
computing
writing
the
code
to
compute
the
eigenvalue
within
your
directive.
You
call
that
utility,
that's
gonna,
be
the
main
difference,
but.
I
G
G
But
yeah,
let
p
ping
me
if
you,
if
you
need
help
on
that.
I,
as
I
remember
your
the
way
you
wrote
this
directive.
It
was
very
close
to
the
one
before
so.
If
you
check
the
way
it's
written
now,
it
should
probably
be
fairly
straightforward
to
copy
paste
and
modify
to
your
need.
I
I
assume,
but
let
me
know
if
yeah,
if
you
have
any
issue.
L
A
Can
be
difficult
to
translate
if
if
somebody
needs
to
go
hunting
around
in
the
code,
so
if
you
could
give
like
the
you
know,
the
five
six
line
version
that
that
might
be
helpful.
E
Is
the
is
the
math
for
all
of
this
sort
of
captured
someplace
so
that
xiaolong
or
anybody
else
could
take,
take
a
look
and
see
exactly
what
is
being
done?.
G
Well,
yeah,
it's
the
same
math
as
before
and
in
the
documentation.
We
say
what
we're
using
that
that
power
iteration,
but
then
we
didn't
copy
paste
like
equation
from
the
wikipedia
article
or
anything.
So
like
say
there
was
no
math
description
before
and
there
is
no
math
description
now
really,
instead
of
we,
we
we
give
the
name
of
the
method.
That's
I
think.
What's
what
that
is
right
now,
yeah.
J
In
in
a
lot
of
ways,
doug,
what
I'm
working
on
right
now
in
discretize
is
kind
of
a
test
for
making
something
similar
on
the
simpeg
website.
So
when
we're
happy
with
theory
sections
for
discretize-
and
we
have
all
of
that-
someplace
that
becomes
sort
of
a
template
for
making
theory
sections
for
for
simpeg,
okay,
that's
sort
of
where
the
space
that
that's!
Where
my
vision
is.
J
A
A
Excellent
thanks
shalom
dieter.
L
Yes,
I
was
mostly
working
with
the
automatic
reading
sort
of
what
I
feared,
so
I
released
this
automatic
reading
for
csm
and
then
folks
used
it
and
it
break
apart
everywhere,
mainly
because
they
use
it
in
different
ways,
and
I
thought
so.
A
phd
student
is
using
it
for
up
to
a
million
hertz,
very
high
frequencies,
which
is
one
thing
I
didn't
really
test,
because
I
never
modeled
it
just
because
I
thought
in
csm,
one
meter
cell
with
should
be
far
enough
for
anything
you
ever
do.
L
But
then
it
isn't.
If
you
go
to
really
high
frequencies
and
the
other
problem
was
that
the
company
they
tried
to
have
a
huge
survey
and
then
it
ended
up
in
some
150
million
cells
and
they
had
understandably
some
memory
issue
with
that
yeah.
I
had
to
do
some
changes
to
that
code,
and
then
I
moved
the
github
org
to
another
name,
because
this
empire
thing
caused
a
lot
of
confusion.
A
Thanks
peter
soggy.
D
I
have
nothing
really
particular.
I
have
found
one
good
link
from
kerry's
webpage
talking
about
how
to
understand
the
data
uncertainty,
so
they
actually
wrote
a
paper
about
kind
of
doing
a
taylor,
expansion
and
kind
of
how
to
assign
the
proper
uncertainty
in
the
frequency
to
the
main
data
so
yeah.
It
would
would
be
quite
useful,
like
anybody
working
on
any
frequency
domain.
F
E
Yeah.
I
don't
have
anything.
J
All
right
a
lot
of
pressure
yeah,
so
I
guess
I'll
just
share
the
screen
and
give
you
guys
the
quick,
10
minute
tour.
J
Let
me
know
if
you're
hearing
buzzing
in
the
background,
apparently
this
neighborhood
just
loves
to
do
landscaping
at
this
time,
but
yeah
what
you're?
Okay.
So
what
you're
looking
at
now
is
the
is
the
discretize
website
I.
This
is
built
locally
on
my
machine
and
there's
two
sections
on
the
side
here:
there's
tutorials,
which
we've
we
we
already
have
and
then
there's
the
theory.
Sections
and
one
thing
you'll
notice-
is
that
the
sections
in
the
theory
correspond
to
a
section
in
the
tutorials.
J
So
I
want
these
to
work
really
well
together.
I
want
you
to
be
able
to
look
at
the
background
theory
about
like
what
is
a
discretized
curl
operator,
and
then
I
want
you
to
be
able
to
go
to
a
place
where
you
can
create
it
and
play
with
it
and
vice
versa.
If
you're
in
a
tutorial
you're
playing
around
with
stuff,
you
want
to
know
how
it
works.
J
I'd
like
you
to
be
able
to
link
back
to
the
theory,
so
maybe
we'll
just
go
to
operators
or
something
there'll
be
some
kind
of
landing
page.
I'm
really
trying
to
assume
that
people
don't
have
familiarity
with
this
so
that
base
level
of
what
is
what
is
finite
volume?
What
does
it
mean
to
discretize
something?
Where
do
things
live?
I
am
trying
to
start
at
a
place
where,
where
people
maybe
aren't
as
familiar
so
you
can
link
to
the
contents
or
associated
tutorials.
J
I
spent
a
lot
of
time
making
images
and
I'm
sure
I'll
get
some
feedback
on
on
whether
or
not
it
makes
sense,
or
it's
really
getting
the
point
across,
but
really
trying
to
do
a
lot
of
visual
aids
trying
to
keep
notation
really
consistent,
and
if
you
want
to
on
any
of
these
pages,
there
will
be
a
place
to
link
to
the
tutorial.
J
So
now
I
just
got
sent
to
a
different
place
in
the
website
and
now
I'm
in
the
tutorial
section,
and
so
now
I
can
play
around
with
going
in
and
taking
some
average
in
1d
talking
about
mappings
and
dimension.
You
know
what
did
these
sparse
matrices?
J
I
have
a
whole
host
of
all
of
the
theory
for
taking
inner
products
of
different
types
from
basic
to
the
tensor
physical
properties,
differential
operators
and
now
moved
into
the
boundary
conditions,
how
to
implement
boundary
conditions
on
these
differential
operators,
and
then
I
have
some
examples
of
actually
solving
all
this,
so
the
putting
it
all
together,
and
so
I
have
one
for
a
poisson
equation
where
we
choose
to
discretize
on
the
nodes
and
then
also
at
the
cell
centers,
and
so
we
we
derive
it.
We
come
up
with
some
examples.
J
We
start
with
maxwell's
equations-
or
I
guess
in
this
case
gauss's
law
of
electrostatics
and
we
take
the
inner
product
and
we
just
we.
We
talk
about
all
the
different
pieces
that
we
learned
and
follow
the
derivation
and
get
to
the
end.
J
J
Then
we
basically
give
you
the
final
result.
Just
a
synopsis
of
everything
go
through
and
you
know
there's
the
discretized
solution
for
one
choice
of
discretization
and
then
we
get
basically
the
identical
solution
here
and
in
one
case
the
boundary
conditions
were
natural.
In
another
case
we
actually
had
to
implement
them,
and
so
that's
that's
generally
the
setup
of
what
we
we
have
here.
The
latest
thing
I
worked
on
is
is
how
we
want
to
implement
general
boundary
conditions
on
any
of
these.
J
These
operators-
and
I
think
I've
got
one
way
of
implementing
it.
I
think
I
could
use
another
pair
of
eyes,
but
when
we're
happy
with
how
we
want
to
do
that,
then
we
can
update
the
code
to
really
just
implement
general
boundary
conditions
on
anything.
J
So
that's
that's
generally
what
I've
been
working
on
and
I
think
it's
now
kind
of
time
to
get
reviewed.
Do
people
like
the
notation
do
people
like
the
figures?
Does
it
make
sense?
Is
there
anything?
Is
there
anything
that
people
question
is
correct?
It's
always
good
to
find
any
kind
of
typo
in
there
if
it
exists,
but
that's
basically
that's.
Basically
it.
E
Devin
that
looks
really
nice.
Nice.
B
Do
we
have
somewhere
where
you
can
basically
dump
the
built
html
folder,
so
we
can
take
a
look
around
at
it.
J
Yeah,
that's
kind
of
the
next
step.
I
know
doug
was
really
interested
in
reading
through
it
and
giving
his
thoughts.
So
I
was
gonna
zip
that
and
send
it
to
him,
but
I'm
open
to
any
kind
of
feedback
so
curious.
If
we
can.
A
Yeah
we
could
easily,
we
could
easily
do
that.
We
could
just
like
create
a
sub
domain
somewhere.
That's
like
temporary
documentation
or
something
like
that
to
take
a
look
at.
C
How
much
are
you
referencing
back
to
the
api,
because
I
think
you're
fencing
a
lot
between
tutorials
and
theory,
but
are
we
linking
anything
back
to
to
the
actual
code
codebase.
J
Not
right
now,
because
I
guess
joe
has
been
working
on
how
the
api
is
generated
and
making
lots
of
progress.
Yeah
yeah.
B
C
J
It
seems
like
it
wants
to
wear
a
lot
of
hats.
I
mean
we
yeah.
We
need
a
theory
section
and
anything
that
you
learn.
You
would
like
to
be
able
to
play
around
with
and
then
for
any
functionality
that
you
have.
J
So
I
don't
I
guess
what
what's
specifically
would
I
be
like?
I
don't
want
to
try
and
make
links
to
the
api
with
all
the
functions
that
I'm
using.
It
seemed
like
that's.
If
you
had
a
question
about
a
particular
function,
you
would
just
go
to
the
api
section
of
the
website
and
and
find
it
or
you
would
call
help
and
get
it
to
spit
out.
The
dog
string,
yeah.
C
Usually
I
can
see
like
numpy
is
the
other
way
around
right.
Let's
say
you're
gonna
do
a
search
on
k-means,
so
you
go
to
k-means,
that's
the
api,
and
then,
after
this
you
can
have
links
to
like
theory,
examples
and
all
that
stuff.
But
it's
kind
of
the
other
way
around,
like
you
kind
of
like
put
the
tutorials
and
the
the
theory
on
top
and
the
api
is
kind
of
like
hanging
below
like
with
it
with
no
link
with
it.
But
it
should
be
the
other
way
around.
C
J
Yeah
and
we
we
can
sort
of
bring
it
up
to
the
top
to
sort
of
prioritize
it
a
bit
more.
One
thing
that
I
did
when
I
worked
on
some
of
the
api
stuff
was
you
can
you
can
create
examples,
basically
in
in
the
docs
string,
so
that
when
you're
in
the
api
section
and
you're
looking
up
a
particular
function,
you
get
an
example,
that's
specific
to
that
function.
J
B
If
you're
like
okay,
well,
let's
build
a
tensor
mesh.
You
can
literally
like
make
that
word.
Tensor
mesh
link
to
the
sensor
mesh
documentation
like
the
api
documentation,
there's
no
sense
to
like.
If,
if
you
find
yourself
talking
about
a
class,
you
should
like
just
make
it
a
link,
yeah
or
or
something
like
yeah.
If
you're
talking
about
a
function.
Okay,
let's
use
the
you
know
the
plot
you
or
plot
whatever,
or
we're
going
to
use
this
utility
function
to
do
this
or
something
I
could
just
make
it
a
link.
A
One
other
question
that
comes
to
mind
is
also
external
references,
as
we
can
connect
like
a
bibliography
to
the
documentation,
and
it
would
be
great,
I
dropped
a
link
for
the
pixels
and
their
neighbors.
That
rowan
led
the
tutorial,
which
is
a
nice
tutorial
on
on
finite
volume.
There's
zelda's
book,
there's
other
resources
out
there
that
we
should
be
pointing
to
because
a
lot
of
this
you
know
we
didn't.
We
didn't
invent
finite
volume,
so
it
would
be
good
to
point
point
to
other
useful
resources
out
there.
A
I
don't
know
devin
if
that's
something
you've
kind
of
been
thinking
through
as
writing.
This,
because
there's
also
like
uri,
asher
and
hens,
work
like
their
their
textbooks
too.
That
would
be.
That
would
be
good
to
point
to
because
we
don't
want
to
end
up
like
you
know,
on
the
hook.
For
for
really
you
know
explaining
every
every
detail
of
finite
volume,
we
should
be
able
to
point
folks
other
places.
J
J
We
could
maybe
do
something
that
says
you
know
this
work
has
been
adapted
from
the
works
of
these
people,
see
these
books
for
a
full
treatment
of
of
finite
volume.
I
mean
really.
J
A
A
L
L
J
Yeah
yeah
we've
done
that
with
some
of
our
our
other
websites
yeah,
I
guess
it's,
it
sounds
like
people
really
want
to
embed
the
specific
references
for
each
specific
thing
in
the
in
the
text.
I
yeah
it's
just
quicker
if
it
can,
if
we
just
take
the
things
that
we
have
have
used,
and
it's
all
sort
of
up
front
and
says
everything
after
this
has
been
adapted
from.
C
Because
because
we
can't
we
cannot
grow.
If
we
do
this
right,
we
cannot
start
abandoning
new
stuff.
I
think
if
we
look
at
the
psychic
learn
the
psychic
learning
website.
You
know
it's
really
well
laid
out
each
each.
You
know
theory
section
has
its
own
references,
it's
all
linked
with
some
papers.
I
really
like
this.
I
think
it's
something
we
should
emulate
and
there's
really
good
cross-reference
between
the
api
and
the
tutorials
and
the
examples,
and
they
all
you
know
they
all
back
back
click.
So
I
think
we
should
try
to
emulate
this.
J
Okay,
well,
I
guess
I
will
find
a
way
to
package
what
I
have
now
so
that
I
can
get
feedback
and
then
whoever
wants
to
be
involved
in
that.
We
can
just
summarize
all
of
the
sort
of
changes
that
we
think
maybe
need
to
get
made.
And
when
we
come
to
consensus
we
can.
I
can
make
the
changes.
A
J
A
Yeah
that
can
sort
of
be
the
the
touch
point
for
ideas
and
documenting
things,
yeah
excellent,
but
this
is
really
exciting
to
see
devin.
I
think
that,
like
this
is
really
going
to
elevate
the
utility
of
the
documentation
and
all
you
know,
ease
so
many
of
those
questions
that
we
get
around,
why
you
know
why
different
operators
are
used
in
different
places,
and
so
this
is
a
really
important
contribution.
So
congrats
on
on
all
your
hard
work
on
this
well.
A
Excellent
perfect:
does
anyone
else
have
any
other
items
comments,
questions.
C
For
those
interested,
the
bcgs
is
doing
a
special
issue
on
on
airborne
surveys
over
superconductor
targets,
3mt
nut
so
check
it
out
tomorrow,
at
4,
30
bc,
time
I'll
be
I'll,
be
the
moderator,
but
james
mcnay
and
others
will
be
there.