►
From YouTube: SimPEG meeting May 6
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
Alright,
oh
I'll
go
for
it.
So
yeah
just
remind
myself
of
these
notes.
Yeah,
so
I
went
through
the
GOC
labs
and
updated
that
to
run
on
base
simulation
branch.
So
that's
that's
finished
and
while
I
was
doing
and
I
found
a
little
bug
for
how
the
2d
fields
are
being
computed
and
kind
of
fixed
that
for
dcpip,
so
glad
that
got
pulled
in
and
then
looked
at
some
suggestions
for
the
simulation
tutorials
review.
B
So
thanks
guys
for
for
giving
a
review
on
that
so
and
then
yesterday,
I
got
into
contact
with
soggy
and
dumb
and
started
to
look
at
transferring
the
em1
D
stuff
into
simple
it's
living
in
a
separate
repo.
So
basically
right
now,
I'm
working
in
that
em1,
D
repository
and
and
and
yeah
making
it
work
on
on
simulation
and
simulation
branch
and
then
I'm
going
to
kind
of
finalize
the
class
structure
and
get
it
more
in
line
with
what
simpang
does
and
when
that's
complete
then
actually
move
it
into
the
sim
peg
repository
I.
B
C
D
C
B
So
I
mean
the
time
domain
in
the
frequency
domain
code.
They
used
so
much
of
the
same
stuff
right
like
the
time
domain
code
is
just
solving
at
a
bunch
of
frequencies
and
then
using
filtering
to
give
you
the
time
domain
response,
so
yeah
I
wasn't
to
separate
them,
didn't
make
much
sense
if
they're
using
so
much
of
the
same
stuff.
So
there
was
sort
of
some
decisions
on
well.
B
D
C
B
That
was
yeah.
That
was
another
option,
is
make
like
a
base
em1
D
inside
electromagnetics,
so
that
you
could
just
use
that
you
could
call
that
cuz.
It
would
be
nice
if
you,
you
know
you
go
into
frequency
domain
in
simp
aghhh,
and
then
it's
one
of
the
options
in
there
right
and
try
to
follow
the
same,
the
same
kind
of
way
of
getting
at
stuff
that
we
have
with
one
DMT
and
the
the
one
D
DC
resistive,
so
that
when
you're
looking
for
stuff
you're,
always
looking
for
it
kind
of
the
same
way.
This.
A
Is
one
to
think
through
to
is
I'm
like
I'm,
certainly
not
settled
in
my
mind,
actually
as
to
where
we
should
move
p.m.
windy
like
I
in
some
ways,
I
wonder
if
it
makes
more
sense
inside
of
something
like
kyo
anna
and
we
could
still
I
mean
simpang
depends
on
to
you
Anna,
so
we
could
still
have
it
basically
be
importing
like
importable
from
sim
pic,
but
I
I'm
not
entirely
sure
where
we
should
have.
A
Basically
more
of
these
like
semi
analytical
approaches
living
is,
do
we
want
to
keep
growing
sort
of
the
simpler
posit
ori,
or
do
we
want
to
have
them
stored,
some
like
that
sort
of
family
of
solutions
stored
somewhere
else,
and
so
I
think
it's
probably
worth
like
sketching
out
a
few
different
scenarios
before
we
think
about
exactly
where
it
should
move
cuz
yeah,
I'm
certainly
not
like.
I
don't
have
those
pros
and
cons
kind
of
clear
in
my
head
yet,
and
I
don't
know
if
anybody
else
does.
C
A
A
C
Right
so
I
think
there's
quite
a
few
things,
because
general
code,
kind
of
handling,
all
the
components
and
no
problem
like
changing
to
my
mode
is
much
better
way
and
it's
actually
much
general.
And
then
we
can
like
just
use
to
find
a
difference
for
sensitivity,
computation,
one
pathway,
but
I
think
this
code
was
really
designed
to
handle
like
airborne
yen,
like
a
large
scale,
I'm.
Finally,
like
the
berries
to
be
very
specific
because
it
serves
a
purpose,
so
I
don't
know
and
yeah
I'm,
not
sure
that
I
really
want.
C
B
Ike,
it's
trying
to
serve
two
or
three
purposes
right:
it's
an
excellent
learning
tool
we
want.
We
want
that
for
people
and
then
academic
research
and
then
yeah,
practical
usage,
so
I
understand
the
motivation
behind
using
this.
This
em1
decode
is
that
you
want
to
be
able
to
to
get
a
result
to
interpret
very
large
scale,
airborne
data,
but
at
the
same
time
we
also
don't
want
to
I
guess
avoid
catering
to
the
other
up
to
audience
like
the
academic,
research
and
and
just
being
able
to
use
it
for
learning.
C
Mostly
horizontal
loop
and
basically
right
at
the
center
at
the
roof,
and
that's
the
majority
of
the
use
case
so
I
mean
if
you
can
hit
those
most
of
the
use
case,
you're
not
only
like
that
kind
of
hitting
like
the
water
skill
problem,
you
also
kind
of
provide
a
research
tool
that
they
can
play
with
with.
There
are
simple
tools,
so
yeah.
E
A
A
C
Pathway,
I
think
because
it
has
a
two
parts
like
basically
handling
a
single
sounding,
they
sucked
a
truly
1d
problem
and
then
there's
the
other
level
calling
that
one
be
function
and
I
guess
like
lots
of
1d
problem.
So
we
can
break
the
break
that
apart
and
then
put
that
truly
1d
solution,
either
like
or
geo
Anna,
and
then
just
break
that
apart,
like
those
like
inversion
functionality
to
handle
large-scale
theta
or
something
like
that.
F
To
me,
it
sounds
a
little
bit
like
just
patching
I
mean
ideally
I.
Think
we
just
want
to
bring
it
down,
because
installation
is
gonna
be
complicated
and
people
want
to
do
one
decode
and
they
need
to
do.
You
need
their
friend
of
Forge
for
synthetic,
but
not
for
synthetic.
My
MD
just
do
it,
I
mean
why
don't
you
just
bring
it
in
it's
not
that
much
work.
B
F
If
the
latest
and
brightest
of
synthetics
going
to
be
the
simulation,
you
know
the
Sioux
Nation
class
I,
don't
know
we
still
need
we're
still
any
good.
If
simple
moves
on
and
we
get
rid
of
the
problem
then
SIF
again
when
they
will
need
to
be
updated
anyway
to
be
able
to
keep
to
keep
track
of.
So
it
needs
to
be
updated
regardless
right,
yeah.
A
Absolutely
no
I
wasn't
advocating
that
we
we
don't
update
it.
It's
just
my
question
is
more
so
what
what's
the
home
that
makes
sense
for
it,
but
that's
all,
but
yeah
I
completely
agree
that
we
need
to
go
forward
in
any
case
with
was
getting
it
updated.
A
Okay-
and
maybe
with
that,
we
can
even
start
either
like
a
discourse
post
or
even
a
github
issue,
just
to
like
think
through
some
of
where,
where
its
home
might
be,
then
we
can
follow
that
up.
A
D
B
Of
the
benefits
of
bringing
it
all
in
under
the
simple
umbrella
is
it's
just
it's
a
central
location
where
somebody
can
go
and
find
everything
you
know
you
want
to
learn
how
to
do
some
modeling
in
1d
or
something
like
that,
and
then
you
you're
in
the
sim
peg
website,
you're
trying
to
look
at
that
and
then
use
it
doesn't
exist
in
simp
egg.
So
then
you're
gonna
get
thrown
to
another
website.
That's
going
to
have
to
explain
the
1d
stuff
you're
bouncing
around
a
little
bit
for
information.
B
A
A
So
I
think
there's
there's
potentially
multiple
trajectories
through
this,
because,
like
one
thing,
that
I
think
has
been
really
positive,
with
breaking
this
critize
out
of
Simpa
independently
on
disk
retires
and
then
how
about
have
that
update
and
so,
like
there's
I,
think
there's
ways
we
can
think
through
it
without
fragmenting
too
much.
But
I
think
that
that's
a
that's
a
very
valid
and
very
fair
concern
that
we
should
like
seriously
consider
as
we
figure
out
where
we're
as
home
should
be.
Okay,.
B
It
would
be
good
to
continue
what
I'm
doing
to
be
working
in
the
em1
D
repository
to
basically
refactor
this
into
a
class
system
that
that
mirrors,
what
is
being
done
in
simp
edge
and
really
the
the
decision
that
we're
gonna
be
faced
with
eventually,
is
how
we
want
to
call
it,
and
where
does
it
live
in
the
end?
But
right
now
what
I'm
working
on
is
still
useful
and
it's
still
what
we
need
to
do
absolutely.
A
B
A
G
A
quick
update,
yeah
I've,
been
tied
up
with
a
little
bit
of
field
processing
debug
in
that
interface,
for
synthetic
and
whatnot
for
organs
in
house
stuff,
but
I
have
been
tackling
the
sensitivity
matrix
or
the
empty
stuff.
I've
just
been
taking
what
what
we
talked
about
there
on
Thursday,
how
to
break
up
the
sensitivity,
calculation,
storing
peat
or
storing
the
key
matrixes
to
a
hard
disk
and
then
going
through
just
in
chunks
and
calculating
everything.
It's
not
necessarily
say
the
fastest,
but
it
is
certainly
a
lot
cheaper.
G
G
So
pretty
much
debugging
that
I
think
I
just
got
something
wrong
with
when
I'm,
putting
all
the
matrices
back
together,
but
everything's.
Looking
pretty
good
should
have
that
finish,
hopefully
by
next
weekend,
all
right
something
to
look
at
and
see
what
yeah
see
what
Joe
thinks
he
wants
to
take
a
look.
Please.
C
C
In
a
Zard
like
a
because,
like
a
the
worst
case,
if
you
think
about
it
like
as
far
as
me,
that
stores
a
dense
matrix
that
gender
is
a
huge
memory
or
so
I
was
just
curious
because,
like
I
mean
we
could
be
pretty
smart
about
storing
this
force
matrix.
Just
like
not
like
store
the
nonzero
elements
and
indices.
G
A
good
point
I'll
have
to
look
into
that
a
little
bit
more
of
I
was
looking
at
so
saving
all
the
P
matrices
and
even
the
J
sub
matrices,
my
my
folders
above
three
gigs,
or
something
like
that.
So
it
does
it
either
compresses
really
good
or
maybe
it
is
just
saving
the
the
nonzero
entries
I'll
have
to
take
a
look
at
that,
but
that's
a
good
point
but
yeah
yeah,
whatever
I'm,
storing
to
hard
drive
or
the
hard
disk.
It's
significant.
G
C
Actually
related
to
sympathy,
m1b
I
think
there's
a
potential
Avenue
that
we
can
really
sell
hard
to
this
California.
They
are
like
in
the
planning
stage
that
actually
scale
up
here,
one
survey
to
entire
Central
Valley
so
and
that
we
are
actually
just
discussed
Oh.
What
are
the
tools
that
we're
going
to
use
to
handle
and
process
the
data
and
so
hope
we're
a
little
bit
competing
with
the
arduous
is
that
they
are
I,
keep
pushing
to
use
their
commercial
software
and
commercial
database.
C
Mentioned,
oh,
we
can
potentially
hire
a
software
programmer
to
make
that
code
more
useable
to
kind
of
end
users,
so
yeah
I'm,
not
for
sure
at
the
moment,
but
I
think
there's
a
there's
a
potential
so
doing
like
it
is
water
for
safe,
like
bringing
into
simulation
branch
and
making
it
clear.
There
it'll
be
really
kind
of
helpful
to
move
that.
C
Because
the
idea
here
like
to
acquire
debate
getting
clear,
they
probably
have
a
chance
to
revisit
their
data
again,
so
I
think
they
want
to
be
prepared
to
have
equipped
with
tools
and
also
personnel
that
can
run
the
code
to
have
three
colors
from
everyone.
Yeah
we're
native
they've
gone
so
yeah
I
think
here
there's
a
real
opportunity
that
we
can
kind
of
bringing
in
something
into
this
water
problem
in
California,
so
yeah
a.
C
C
F
F
I
think
that's
that's
kind
of
an
easy,
easy
path
forward
for
you
guys.
G
G
F
C
F
It's
it's.
It's
basically
an
HD
f5,
it's
an
HD
f5
file,
but
it's
just
structured
so
that
you
can
store.
You
know
metadata
and
different
type
of
geophysical
data.
You
know
like
it
knows
what
the
line
is
and
also
the
pointers
with
the
black
mesh
is
that
kind
of
stuff
right.
So
it's
basically
just
a
prepackaged
hdf
file,
yeah.
C
C
F
The
beauty
is
that
you
don't
need
anyone
after
this,
you
could
use
like
third-party.
You
know
database
management
that
can
handle
h5,
so
you're,
not
you're,
not
you
know,
bind
to
like
any
any
specific
service
provider.
Once
it's
written,
it's
just
an
open
format,
so
you
can
still.
You
can
still
read
and
write
to
it.
Okay,.
C
A
Here's
the
plan
for
it
to
be
so
just
one
like
extension
to
consider
I,
think
h5
is
probably
quite
quite
sensible,
but
also
if
they
do
want
to
actually
have
this
running
on
the
cloud
and
store
it
on
the
clouds
or
is
a
better
format
for
that.
But
there's
there's
some
sort
of
like
in
a
sense
you
can
we
can
you
can
write
an
API
that
can
access
our
or
hdf5.
So
as
long
as
you're,
using
the
same
like
internal
structure
for
the
for
the
data
file,
then
that's
fine,
but
that's
just
something.
A
C
F
C
C
Working
with
them
in
I
think
so,
actually,
writing
writing
up
the
documentation
and
how
things
working
very
helpful,
because,
frankly,
it's
almost
a
black
box
code
that
I
can
only
see
or
a
few
other
people
who
knows
so
yes,
it'll,
be
it'll,
be
great
to
have
examples
and
how
things
are
working
and
three
pilot
studies
that
we've
done
so
I'm
moving
up
by
the
time,
I'm
hoping
to
kind
of
make
that
as
an
example
and
that
other
data
format
storing
as
an
odo,
then
kind
of.
Why
point
that?
C
A
Yeah
I
mean,
if
there's
other
things
that
come
up
and
even
if
you
think
of
stuff
that
would
be
helpful
to
have
unlike
simple
X
Y
Z,
like
the
main
landing
page.
That
would
help
tell
the
story.
We
can
definitely
prioritize
some
of
those
things
because
it
would
be
fantastic.
It's
in
Vegas
at
least
one
one
of
the
potential
avenues
in
this
project.
C
C
The
other
thing
was
also
the
parameter
like
default,
parameter,
maximum
CG
and
the
tolerance
of
CG,
so
that,
like
I,
think
I
haven't
recognized
like
this
L
like
this
LP
norm.
Inversion
is
like
combining
on
all
of
those
parameter
when
I
think
probably
just
are
set
properly.
It's
ins
working
quite
well
anyway,
so
but
like
if
you're
just
simply
using
default,
parameter.
That's
quite
that's,
but
anyway.
F
C
So
I'm
interesting
this,
what
they
call
corchran
claim.
That's
if
you
to
a
smoother
version,
that's
what
you
would
recover,
but
like
it's
100
meters,
Hana
metre
gap,
but
it's
not
going
to
be
a
hundred
meter
thick,
but
what
they
really
interested
was
effectiveness,
what
the
extent,
because
it's
actually
it,
which
dissect
upper
and
lower
aquifer,
it's
really
thick
clay
which
impede
the
water
so
at
annoying
it's
competing
unit,
is
very
important.
So
if
you
are
supporting
that's.
F
C
And
you're
what
I'd
have
to
do
was
I
just
like
spected
this
portion
here
and
just
invert
it.
Those
part
like
just
impose
the
sparsity
on
the
white.
So,
okay,
it's
clean,
it
is
as
force
that's
what
you
can
get
right
at
the
end,
it's
actually
matching
almost
perfectly
so
I
haven't
used
the
you
log,
it
the
right
top
boundary
of
the
core
and
clay,
and
you
can
actually
get
that
that's
a
nice
thickness
of
the
clay
later.
So
that
was
quite
quite
happy
with
that.
And
finally,
things
are
working
so.
F
C
F
F
C
I
think
they
are
actually
serious
about
kind
of
to
be
like
it
being
converting
those
data
per
system.
I'm
not
sure
I
got
how
much
before
they
have
pen
to
get
all
this
calibration
right,
so
they
actually
have
a
test
site.
So
what
they
do
is
they
have
like
a
make
it
test
site
man,
you
got,
it
reads:
TV
model,
you
statically,
generate
a
synthetic
data
and
they
flew
over
the
test
site
and
see
if
you
can
actually
match
there
was
decay
curve.
Man
calibrate
first
point
calibration,
so
yeah.
G
C
F
D
F
F
I
F
I
F
D
F
Are
we
I
guess
I
can
do
it
whatever,
but
I
think
one
thing
that
I
would
like
to
really
like
to
bring
in
is
the
is
the
director
for
the
spherical
formulation,
because
right
now
we're
kind
of
running
two
inversions
back-to-back,
but
I
have
a
directive
to
do
the
other
spherical.
So
I
think
my
plan
this
week
is
create
a
PR
for
for
that
guy
you'll
clean
things
up
a
lot
more
for
the
inversion
part
thanks.
I
I
I
It
look
nicer
I
generally
moving
forward.
I
would
like
to
get
them
all
in
terms
of
numpy,
Docs,
documentation,
style,
I,
I.
Think
it's
much
cleaner.
It's
a
much
easier
way
to
write
the
documentation
actually
and
so
now,
I,
don't
know
it
looks
cleaner.
Only
within
like
the
like
the
documentation
is
the
Sphinx
generated
stuff.
Like
that's
separate.
What
I'm
talking
about
is
that
it
looks
better
if
you
like,
just
look
at
the
documentation,
it's
readable,
like
human
readable.
Instead
of
all
this,
like
you
know,
semicolon
per,
am
Coen
something
coming.
I
Yeah
I'm,
looking
into
a
few
tools
to
kind
of
help
us
automate
that
the
biggest
issue
we
will
have
is
the
documentation,
that's
generated
from
the
properties
and
I.
Just
haven't
thought
about
how
to
get
on
top
of
that.
Yet
it'll
have
to
you'll
have
to
submit
some
pull
requests
to
properties
to
get
it
working
have.
I
So
the
things
that
we
get
from
properties
right,
we
get
automatic,
getters
and
setters
and
type
validation.
Are
you
and
then
we
also
get
the
serialization
from
memory
yeah,
okay,
so
I'm
just
kind
of
curious
like
if
it
if
it
would
be
worth
it
for
us
to
like
write
some
simpler
ones
that
we
just
to
do
what
we
need
to
do.
I.
B
One
thing:
I
guess
since
we're
talking
about
this.
As
for
for
the
codes,
let's
say
the
frequency
domain
eeehm
like
the
3d
one
in
the
doc
string
under
that
class,
you
put
how
the
problem
is
discretized
and
then,
when
that,
when
you
Auto
generate
the
API
you're
gonna
have
a
lot
of
a
lot
of
math
kind
of
mixed.
B
In
with
what
the
manners
are
to
do
that
I
guess
I
wondered
if
people
were
really
keen
on
having
the
formulation
always
be
put
in
the
dock
string
for
that
simulation
or
whether
or
not
it's
better
to
put
those
in
in
something
more
of
like
a
theory,
section
I
kind
of
prefer.
The
latter
I
think
it
makes
the
documentation
for
the
API
shorter
and
easier
to
find
stuff.
You.
I
D
B
I
agree
and
then
yeah
after
looking
it
up
and
you
just
click
and
it'll.
Take
you
take
you
to
that,
but
I
guess
the
first
time
going
around
and
trying
to
learn
simple.
That
was
all
auto-generated.
There
was
like
there
was
math
her
patience
and
theory
in
there
and
then
and
I
guess:
I
found
it
harder
to
get
started
with
simple,
so
yeah,
ultimately
I'm
hoping
we
can.
We
can
move
those
the
information
about
the
formulation
to
another
place
and
then
yeah
link.
It.
A
I
I
I
So
there's
ways
you
can
at
least
for,
like
the
API,
like
the
API
documentation,
I
personally
like
to
have
that
as
part
of
like
almost
within
the
code
so
like
within
the
module
initialization
like
module
in
it
dot
pies,
we
can
put
a
bunch
of
stuff
in
there,
so
it's
like
also
not
lost
if
someone's
running
it
locally,
without
having
a
look
at
the
documentation
without
having
to
keep
it
up.
I
would
like
to
reorganize
how
like
the
landing
page
with
it
cuz
right
now.
I
A
Was
just
gonna
say:
I
think
that
this
could
be
timely,
I
mean
I.
Don't
think
you
should
like
address
this
with
the
pulling
in
of
the
simulation,
but
it
definitely
should
be
on
the
list
for
shortly
afterwards,
but
I
can
see
this
being
quite
fitting
with
sinkings
through
basically
some
of
the
updates
we're
making
TEM
geo
sigh
and
also
potentially
thinking
about
like
what
an
inversion
style
resource
should
look
like
and
what
things
we
want
to
link
out
to.
A
A
I
I
put
in
the
notification
like
I
think
that
we
had
at
least
the
last
time
we
had
talked
about
the
relative
error
standard
deviation
and
all
that
stuff.
Think
that
means
that
we
had
like
kind
of
converged
on
it
commented
on
into
the
thread
in
the
in
the
book
thread.
I,
think
we're
comfortable
settling
on
relative
error
and
noise
floor
and
renaming
uncertainty
to
standard
deviation.
At
least
that's
what
I
think
we
had
converged
on.
So
let's
do
any
other
thing
that
I
I
would
like
to
get
in.
A
One
question:
I,
don't
know,
folks
think
this
is
confusing.
Maybe
we're
actually
fine,
because
we've
working
on
the
simulation
branch,
the
whole
time
anyways,
but
is
we
do
actually
have
the
term
standard,
deviation
on
the
data
class
right
now
and
we're
we're
thinking
about
like
giving
a
new
meaning.
Do
you
think
that
that's
potentially
confusing
like
do?
I
I
mean
iid
I
mean
I,
think
I
would
I
would
be
okay
with
just
kind
of
going
forward
with
it
doesn't
rename
a
little
and
then
put
in
deprecation
for
what
it
used
to
be
called
in
the
original
master
or
in
the
master
branch
that
says,
I,
don't
know
what
else
other
people
who've
been
working
with
the
simulation
branch
would
think
I.
Don't
think
it
should
be
like
I'm
not
like
I'd,
be
fine
with
jumpin.
All
in
that
sounds
good.
E
E
I
E
E
F
To
both
sock
for
for
patching
sympathy,
I'm
waiting
so
quickly,
that
was
that
was
super.
Helpful
people
started
trying
to
install
the
different
machines
and
then
everything
broke
like.
C
E
H
A
I'll
share
a
couple:
updates
have
been
trying
to
actually
do
some
coding
again,
which
has
been
really
nice,
so
I've
been
jumping
in
a
little
bits
on
just
finding.
You
know
there's
a
few
little
bugs
here
web
reports
that
you
know
folks
who
have
pushed
on
this
impact
issues
and
so
tried
to
just
address
a
couple
of
those
that
I
thought
would
be
easy
and
nice
to
include
with
the
simulation
refactor.
So
there's
a
couple:
pull
requests
on
there
that
have
been
started
but
are
pretty
lightweight
with
the
simulation.
A
Refactored
Joe
and
I
have
been
working
on
a
blog
post
and
I
put
a
Google
Doc
link
in
the
notes
there,
and
it's
definitely
still
like
a
sketch
and
there's
I.
Think
some
things
that
are
missing.
Joe
made
a
good
point
that
we
should
have
a
big
section
on
the
module
renaming
and
just
what
things
were
and
what
they
are
now.
But
if
there's.
A
A
If
you
want
to
basically
share
like
a
complete
setup
and
just
kind
of
get
up
and
running,
that
should
be
a
bit
easier,
but
it
does
take
just
a
bit
of
thought
like
a
lot
of
a
lot
of
it.
Just
is
you
know,
logic
around
how
things
are
instantiated,
so
I
might
be
firing
some
questions
here
and
there
just
to
make
sure
that
things
are
consistent.
So
it's
it's
both
how
things
are
instantiated
and
then
what
are
the
things
that
are
properties
with
the
properties
library,
because
that
is
basically
telling
us
what
is
being
serialized.
A
So
what
needs
to
be
like
the
the
logic
for
what
things
should
be
on
properties
or
basically,
what
are
the
things
that
we
need
to
be
able
to
recreate
this
class?
And
so,
if
there's
anything,
that's
computed,
we
shouldn't
be
serializing.
That's
like
we
shouldn't
necessarily
be
serializing
the
ingredient
operator
on
a
discretized
mesh,
but
we
should
be
serializing.
The
underlying
cancer
mesh
vectors
and
later
on.
We
can
think
through
I
mean
if
there
are
some
things
like
as
John
was
saying.
A
If
we
do
want
to
be
saving
things
like
projection
matrices,
if
it
actually
makes
sense
that
you
know,
we've
already
computed
this
once,
let's
just
serialize
it
and
so
I
have
them.
When
I
load
back
up.
We
can
start
to
put
some
of
those
hooks
in
place,
but
at
least
for
the
first
class.
It's
very
much
just
like
these
are
exactly
what
I
need
the
minimal
set
of
things
that
I
need
to
create
to
recreate
this
object.
A
So
I
said
yes,
so
we're
now
on
the
hook,
so
we've
got
the
three
hours
slots
and
so
what
it
would
be
helpful
and
I
can
start
some
of
this
and
we
can
iterate
together,
but
it's
just
to
kind
of
sketch
out
what
we
would
like
to
do
had
a
bit
of
a
slack
chat
with
some
folks
and
talking
about
basically
I
and
I,
really
like
the
suggestion
of
starting
with
real
data,
get
an
inversion
up
and
running
like
from
scratch,
go
and
then
sort
of.
A
Once
we've
worked
through
an
entire
inversion
and
hopefully
I
mean
something
simple:
we
can
start
with
potential
fields
the
lightweight
dataset
like
we
don't
need
to
do
anything
too
fancy
but
show
how
we
would
work
through
that,
maybe
for
two
datasets
or
something
like
that.
Do
DC
resistivity
and
potential
fields
to
give
folks
a
flavor
of
two
different
problems,
then,
basically,
after
that,
go
and
sort
of
unpack
what
what
happened
and
what
all
of
the
pieces
are
and
what
kind
of
knobs
and
things
there
are.
A
So
a
certain
document
where
we
can
brainstorm
these
things,
and
if
anybody
has
datasets
that
we
can
distribute
that
you
think
would
be
a
good
candidate
for
this.
That
would
be
great
if
we
can
sort
of
nail.
Those
down
early
on
I
was
in
some
senses
wondering
if
there's
some
like
for
DC
from
the
GWB
project
that
might
be
fitting,
but
other
than
that
we
can.
We
can
think
through
what
other
sites
would
make
sense.
A
C
D
C
H
C
H
H
A
So
you
can
sort
in
a
sense,
I
mean
it's
helpful,
actually
looking
through
the
schedule
and
to
see
what
other
tutorials
are
happening,
I
think
so
what
Matt
basically
said
is
we
should
aim
for
sort
of
a
beginner,
moderate
level
and
so
I
mean
I.
Imagine
folks,
who
are
gonna
show
up
to
a
simple
tutorial,
have
some
interest
and
hopefully
a
little
bit
of
background,
but
they
might
not
have
much
and
so
yeah.
We
should
plan
for
something
that's
reasonably
introductory.
D
E
I
C
B
Yeah,
this
is
I
guess
this
is
one
of
the
nice
things
about
the
style
of
tutorials.
That
I
was
trying
to
design
as
they
were.
They
were
done
for
in
a
style
where
someone
would
actually
be
trying
to
adapt
the
tutorial
to
complete
a
real
project.
So
you
know
if
you
want
the
one
for
for
like
a
2d
dcpip
inversion,
and
you
want
one
for
for
for
magnetics
or
something
like
that.
B
A
That's
great
yeah,
well
and
I
would
encourage
everyone
to
actually
take
a
look
at
the
lineup
there's.
Some
talks
that
I'm
in
tutorials
that
I'm
super
interested
in
showing
up
to
like
there's
some
geo
stats,
there's
also
gem
pie,
so
Miguel
is
gonna,
give
a
talk,
so
I
think
some
of
these
could
be
quite
useful.
So
hopefully
a
few
of
us
can
get
to
a
handful
of
them
and
yeah
sure
share
some
learnings.
A
Imagine
so,
but
I
can
ask
I
can
ask
Matt
I,
don't
actually
I,
don't
know
for
sure
Peter
do
you
know
no.