►
From YouTube: SimPEG meeting April 29
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
A
B
Can
start
off
yeah
I?
Don't
not
nothing
too
much
new
I've
actually
been
doing
a
lot
of
work
on
our
I
guess
the
app
for
the
backend
so
that
we
can
use
simple
like
interactively
kinda
thing
with
a
GUI
yeah.
Almost
there
there
we've
got
four
nodes
connected
to
a
main
resource
manager
and
an
interface
that
talks
to
it
from
any
computer.
Just
so
yeah
start
it
up
in
Python
and
gonna
run,
run
simpang,
conversions
and
then
it
like
feeds
back
information,
I've,
modified
or
I
have
a
script
for
the
inversion.
Yeah.
A
B
B
Kind
of
a
work
in
progress,
I,
don't
know
what
everybody
wants
like
now
that
I've
got
users
starting
now,
I'm
gonna
get
the
request
ago.
Can
I
see
this
and
see
that
so
it's
not
quite
finalized,
but
I'm
just
putting
out
what's
in
the
text
file,
as
is
it's
the
poly
m5f
I
am
Phi
D,
the
smooth
alphas
and
all
that
kind
of
stuff.
Just
after
right
now
and
then,
whatever
request
I'll,
do
that
in
later,
okay,.
A
Cool
yeah
I
mean
that'd,
be
interesting
and
just
useful
insight
as
to
what
what
users
want
to
see
and
yeah,
because
when
sort
of
thinking
about
like
that
next
level
of
wrappers
and
things
like
that
being
able
to
have
basically
like
a
set
of
defaults
that
has
been
reasonably
tested
by
others
be
great.
Yeah.
B
I'll
have
all
the
information
whatever
and
I,
try
to
Joe's
sensitivity
code
for
the
MT
and
write
on
it
seems
to
be
working
pretty
good,
yeah
I'm
just
trying
to
run
a
few
simulations
using
it
as
the
weight,
but
I
keep
getting
a
line,
search,
break
so
I'm,
also
using
the
new
JT
JT
Beck
and
J
vex
that
I'm
using
so
maybe
they're,
causing
the
problem.
There's
got
to
check
into
that
and
yeah
just
want
to
dig
into
the
fields
object
a
bit
too
because
I
really
like
to
make
that
an
x-ray.
B
B
D
Say,
like
I
say
like
if
they
just
have
an
overarching
class
wrapper
that
decides
like
okay.
Well,
this
is
it
we'd
have
a
fields
class
and
then
I
would
say
that
we
would
have
like
a
you
know
like
the
regular
style
field
and
then
a
X
or
a
field
or
something
they
both
access.
The
same
way,
it
doesn't
matter
yeah.
B
B
D
D
I've
been
I've,
been
going
through
the
desk
stuff
again
and
trying
to
separate
it
out
it's
working
nicely,
but
I
just
have
and
I
think
I
looking
at
it
closely.
I
still
have
more
questions
like
so,
if
you'd
like,
if
I
don't
know
down,
if
you
can
stay
on
afterwards
a
little
bit
and
come
try
out
a
few
things
just
to
make
sure
I
understand
that
it
was
doing
what
I
think
it
was
doing.
Yeah.
B
D
Charge
abilities
velocities,
essentially
everything
for
those
nice
mining
example
target,
but
hopefully
I
can
push
them
to
make
the
model
open.
It
was
but
I've
been
using
that
to
test
a
bunch
of
some
MT
stuff
on
as
well,
because
we're
using
emptied
modeling
on
it.
So
it's
using
something
on
that,
hopefully
give
us
an
idea,
or
at
least
give
me
some
sort
of
sense
of
where,
like
where
things
are
going.
D
A
Yeah
I
mean
Jo
and
I
I
think
talked
about
this
on
slack.
Maybe
we
do
have
a
github
repo
that
like
has
the
intention
of
being
a
space
where
we
can
store
models
and
things
like
that,
and
it
would
be
fun
to
start
to
try
and
sink
through
then
I'm
like
I,
think
T
Casey
could
be
a
good
example.
A
The
synthetic
these
don't
simply
have
the
rights
to
distribute
the
data
yet,
but
we
probably
we
might
be
able
to
at
some
point
down
the
line,
the
L
block
and
basically
like
what
I
would
think
would
be
kind
of
cool
to
have
is
basically
just
like
a
model
class,
and
then
you
can
give
it
a
mesh
and
it
just
like
sticks,
sticks
the
model
on
image
and
so
something
that
basically
has
like
effectively
a
parametric
representation.
A
However,
you
define
that
of
what
that
model
is,
and
then
it
could
take
any
sort
of
discretized
mesh
and
put
the
model
on
there.
And
then
you
could
run
whatever
simulations
you
want,
yeah
and
so
something
to
like
think
about,
and
if
we
have
excuses
or
reasons
to
do
that,
it
might
be
nice
to
try
and
get
a
few
examples
that
could
sort
of
be
benchmarks.
D
D
E
I'm
not
L
dad
harbors
book
I
got
more
confused
about
differences.
How
things
are
about
notations,
but
but
I
finally
managed
it
somehow
I
think
your
chai
tea
bag,
even
though
it's
an
odd
it
uses
the
adjoint.
It's
not
the
adjoint
stays
method
right.
You
still
have
an
AE
in
so
you
have
somehow
still
an
implicit
shake
:.
Is
that
correct.
A
D
E
D
D
D
Ever
only
ever
find
it
online
I
never
been
able
to
find
it
again.
Ok,
it
was
at
some
conference
proceedings
or
something
somewhere.
It
was
like
this
list,
long
list
of
PDD
constraint,
optimization
stuff
and
that's
I.
I
have
a
feel
like.
If
you
look,
if
you
start
googling
like
PD
constrained
optimization,
you
might
find
something
similar
or
you
might
be
able
to
find
something
more
along
the
path,
because
that's
essentially
what
Sims
egg
is
doing
or
like
for
most
of
this
problem.
It's
a
PDE
constrained,
optimization.
E
So
yeah
I
mean
it
works,
it's
I
think
d-did.
The
outcome
is
the
same,
so
I'm
happy
with
it,
but
it
did
got
me
thinking
about
looking
a
bit
more
about
simple,
because
one
of
my
hopes
was
to
kind
of
make
a
wrapper
for
for
my
code
to
use
it
as
a
solar
for
force
impact
at
some
point
once
the
simulation
branches
in
then
we
could
have
maybe
want
to
call
about
that.
What
would
have
to
be
done
so
we
could
try
it.
E
Obviously
it
would
just
be
for
a
very
specific
task
of
say
freak
to
the
Mayan
problem
study
it
would
work
and-
and
if
it's
possible
one
thing
would
be
that
the
solver
needs
much
more
information.
It's
just
a
directive
it
just
it
doesn't
just
need
the
a
matrix
and
the
right-hand
side,
but
it
needs
the
whole
model
and
the
discretized
mesh.
So
some
but
I
guess
that's
not
a
big
blocker
that
should
work
somehow
they
couldn't
manager,
but
then
I
was
wondering
if
so,
the
forward
modeling
I
think
could
be
done
without
too
many
issues.
E
A
So
that
should
be
pretty
self-contained
to
this
simulation,
like
all
that
we
really
the
I
mean
and
we'll
see
if
we
run
into
any
issues,
but
with
the
data
misfit
like
all,
we
need
the
simulation
to
be
able
to
do
use
this
data
and
then
the
sensitivity
times
a
vector
and
the
adjoint
sensitivity
times
a
vector.
If
you
want
to
do
second-order
methods,
if
you
only
want
to
do
first
order
methods,
we
just
need
JT
times
a
vector
and
deep-red
out
of
the
simulation.
D
E
D
D
E
One
more
question:
I
stumbled
up
because
I'm
thinking,
oh
it's
the
whole
thing
of
data
management.
How
does
it
I
think
I?
Remember
you
talked
about
that
you
might
change
to
or
go
to
x-ray
or
how
do
you
handle
data
internally?
Is
it
the
data
class,
receive
the
survey
class
or
is
that
something
that
changes
in
the
simulation
branch.
A
Yeah
so
now
it's
the
data
class,
so
previously
it
would
have
been
the
simulator
seriously.
It
would
have
been
the
survey
class
that
hold
on,
but
now
it
will
be
the
data
class
and
right
now
it's
just
a
numpy
array,
but
that's
something
that,
like
we
can
look
into
and
see
if
makes
sense
for
that
yeah
for
now
it's
it's
just
simply
an
umpire
II
on.
C
A
E
A
E
E
D
E
C
Joined
inversion,
like
kind
of
like
also
like
a
drew,
did
something
similar
for
his
PTSD
of
like
including
petrol,
physical
information
in
George
information
in
your
inversions
and
that
trying
to
reproduce
that
petrol,
physique
and
urgent
information.
At
the
same
time
as
your
Joe
physics,
geophysical
data,
yeah,.
E
C
C
C
Thanks
things,
there
yeah
I
I'm,
just
trying
to
have
something:
that's
coherent
right
now,
just
like
so
I.
Thankfully
I
have
like
a
bunch
of
like
draft
for
bit
of
paper,
so
that's
I
can't
that
I
can
use.
But
then
it's
just
about
the
reflects.
Oh
I'm
moving
I
need
to
move
the
introductions
there
and
like
okay-
and
this
part
is
repeated
or
like.
Oh
now,
it's
a
daisy,
so
I
need
to
give
more
detail
about
that.
That
means
the
paper
ad
in
the
paper
that
was
just
like
pushing
on
the
side
and
Sapphire.
C
D
A
Well,
I
just
had
a
couple
updates:
they
dropped
some
links
in
the
Google
Doc,
the
first
one.
There's
a
the
software
underground,
transform
virtual
conference
happening
in
June.
It
looks
like
there's
still
potentially
space
on
the
schedule,
so
as
wondering,
if
folks
would
be
interested
in
like
helping
run
a
session
on
sort
of
inversion
with
simpang
there's
other
there's
other
similar
sessions
being
run
I'm
like
I,
don't
feel
like
I'm
sort
of
in
a
place
where
I
could
take
charge
of
this,
but
I
would
be
very
happy
to
support.
A
If
someone
wants
to
lead
this
or
if
we
have
sort
of
a
team
who's
gained,
then
I'm
happy
to
sort
of
be
someone
on
the
hook,
but
wouldn't
want
to
volunteer
just
just
on
my
own.
So
I
don't
know
if
folks
have
had
a
chance
to
look
at
that
or
we
can
think
on
it,
but
I
would
reach
out
to
Matt
Hall.
If,
if
this
is
of
interest
and
just
ask,
if
there's
a
slot
that
we
could
do
a
quick
overview
simply.
E
A
E
A
That's
good.
We
can
also
chat
so
I
know:
Dom
was
leading
the
charge
for
having
a
tutorial
s
e-g,
and
so
that
I
mean
this
could
be
basically
a
preliminary
run
of
that.
So
we
can.
We
can
ping
him
and
see
because
I
think
like
if,
if
we
get
like
three
of
us
and
one
person
who
we
know
will
be
able
to
present
like
I,
would
be
able
to
present
that
week.
But
if
there's
three
of
us
who
are
helping
generate
content,
three
or
four,
then
then
I'd
be
totally
comfortable.
A
A
B
A
Okay
and
then
the
last
thing
that
I
want
to
just
put
out
because
I
think
one
one
of
the
things
if
we
want
to
get
the
release
done
on
Friday,
it
would
be
nice
if
they
have
a
blog
post
to
your
company,
not
on
actually
I.
Think
one
of
the
most
important
pieces
of
that
blog
post
is
an
updated
sketch
of
the
framework,
but
I
just
took
a
crack
at
this
and
so
totally
open
to
comments
and
ideas.
I
basically
I
mean
so.
A
The
important
thing
to
realize
here
is
that
this
is
not
meant
to
be
like
fully
comprehensive,
even
the
previous
one
like
if
you
actually
sort
of
really
drill
into
it,
there's
things
that
are
missing
and
like
it's
not
a
full
complete
picture.
By
the
same
time,
it's
still
like
a
useful
conceptual
model,
so
I'll,
maybe
to
share
my
screen
and
I
can
show
you
what
that
looks
like.
So.
D
A
A
The
the
things
that
have
changed
here
and
we
can
play
around
with
the
visual
I'm
not
entirely
sure
if
this
is
the
best
layout,
but
so
we
can
see
now
that
we
no
longer
have
the
pair,
but
I
still
kept
the
survey
as
a
distinct
class,
because
it
is
a
distinct
class
that
goes
to
places.
It
goes
to
the
data
and
it
goes
to
the
simulation
I'm.
So
the
data
class
has
been
added.
That's
new
did
not
exist
here,
even
though
we
had
a
data
class.
A
Yeah
and
that's
basically
the
only
thing
that
I
changed
I
mean
we're
still
like
missing.
We
don't
have
directives
in
here
that
sort
saying,
but
we
didn't
before
so
I
sort
of
feel
like
that's
that's
fine.
Maybe
we
just
leave
it.
Leave
it
out,
yeah!
So
I,
don't
know
what
your
initial
reactions
are
coming
to,
that
I.
E
A
Folks,
I
mean
I.
Guess,
like
the
the
thing,
there's
a
couple
things
that
I
think
that
are
important
that
are
missing
but
would
perhaps
go
into
another
iteration
of
this.
For
example,
the
model
is
missing
like
the
model
and
mappings,
and
that's
actually
a
pretty
critical
piece
to
being
able
to
use
simple,
so
I
think
like
what
I
would
envision
is
perhaps
doing
is
maybe
publishing
sort
of
these
frameworks
side-by-side
in
the
blog
post
and
then
later
down
the
road
creating
a
new
one.
That's
a
bit
more
complete
I.
A
So
each
of
these
boxes
are
sort
of
things
that
you
would
set
up
in
a
Cell,
so
I
would
be
setting
up
like
you
need
to
set
up
your
mesh
sort
of
the
survey
and
the
data
are
the
first
like
sort
of
building
blocks
and
then
from
there
you
build
out
the
simulation
and
then
work
through
the
inversion
components.
So,
roughly,
if
you
start
sort
of
at
the
the
spots
where
you
know,
the
arrows
are
only
going
out
you're
in
reasonably
good
shape.
Yeah.
A
D
A
Yeah,
but
I
think
is
an
interesting
question
when
I
sketched
this
out,
actually
I
can
show
you
I
made
a
sketch,
just
like
on
my
iPad,
as
I
was
trying
to
think
through
this,
and
in
that
case
I
didn't
have
any
sort
of
particular
boxes
and
maybe
just
color
code
stuff
that
is
like
this
is
part
of
the
foreword.
This
is
part
of
the
inverse.
We
could
think
about
that.
A
E
E
E
D
D
What
happens
is
that
the
survey
is
defining
what
the
acquisition
geometry,
essentially,
whether
that
be
like,
obviously,
the
location
of
receivers,
also
like
their
components,
the
types
of
receivers
and
things
like
that.
So
it's
kind
of
like
the
geometry
of
the
survey
almost
and
then
simulation
handles
predicting
it
to
that.
Do.
E
A
A
Think
that's
a
bit
of
an
important
distinction
between
these
two
is
actually
the
role
of
the
survey
class
has
actually
gotten
smaller
in
this
refactor.
Okay,
the
survey
class
no
longer
all
it
does
is
hold
the
geometry
and
like,
if
you
ask
for
sources
or
receivers,
it
can
do
sort
of
the
interpolation
or
evaluating
itself,
but
it
doesn't.
It
no
longer
holds
the
data.
The
data
takes
the
survey,
so
we've
reversed.
Who
is
the
loser
the
container,
so
in
this
case,
for
a
for
an
inversion
where
we're
working
with
with
field
data?
A
Yeah
and
the
only
reason
I
may
be
a
baker.
Well,
there's
a
couple
reasons:
one
dish
for
visuals
is
I,
do
know
how
to
arrange
it
where
it
looked
reasonable,
where
it
was
fall,
but
the
other,
I
think
the
role
like
the
simulation
is
actually
a
pretty
big
class.
It
has
a
lot
more
guts
to
it.
Then
then,
some
of
the
rest
of
the
things,
and
so,
but
we
could
definitely
play
around
with
with
that
and
have
things
be
the
same
size
if
that
sort
of
prompts
too
many
questions
or
is
confusing.
C
Yeah,
like
one
comment:
I
like
I,
want
a
mental
I
like
that.
You
just
said
is
that
you,
like
you,
can
start
with
the
bots.
That's
a
volley
like
our
rule,
that
comes
out
of
it
so
mesh
and
survey
and
I
guess
it
depends
of
like
what's
what
with
the
target
and
what
you
want
to
converse
with
that
visual
framework.
A
I'll
start
from
this
is
just
kind
of
a
first
pass,
but
even
if
you
want
to
just
sketch
out
on
a
napkin
or
a
piece
of
paper
and
take
a
screenshot
and
throw
it
into
slack,
I
enjoy
seeing
what
other
conceptual
models
are,
and
we
can
iterate
on
this
or,
if
you
want
to
take
this
one
and
just
like,
create
a
new,
create
a
copy
of
it.
I
think
everyone
here
should
have
access
to
this
Google
Drive
Theatre.
A
B
A
A
You'd
be
willing
to
just
post
on
the
discourse
that
they
started
because
I
think
let.
C
A
E
D
E
A
D
E
A
Anything
else
all
right,
thanks
everyone
and
we'll
see
you
on
slack
and
yeah,
well,
sort
of
post
updates
on
the.
Hopefully
you
get
a
blog
draft
out
soonish,
and
if
people
want
to
jump
in
and
contribute
to
that,
that
would
be
great.
And
if
you
want
to
draw
a
scribble
on
napkins
I
enjoy
seeing
what
everyone's
perfect
all
right.