►
From YouTube: SimPEG meeting April 8
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
D
Quite
a
few
stuff,
and
in
Stanford
on
having
actually
more
time
now
to
actually
work
on
stuff
and
one
like
a
potential
Avenue
that
we're
working
on
so
we're
finalizing
all
right,
like
a
project
called
that
groundwater
architecture
project.
But
at
the
end
of
the
day
we
need
to
provide
the
data
data
management
system
I'm.
Not.
We
were
not
exactly
sure
what
it
is
at
the
moment,
so
we're
brainstorming
in
these
days.
How
are
we
going
to
provide
all
of
the
urban
data?
Well,
data
water
level,.
B
B
C
What
you
think
about
doing
is
archiving
it
on
soon
odo
and
then
I
believe
that
sonoda
actually
has
an
API.
So
they
basically
then
like
set
up
a
notebook
that
calls
those
an
odo
API
that
then
uses
simpe
to
from
the
inversion,
okay
and
so
cuz.
What's
nice,
it
was
nodoz,
you
get
a
DOI
for
that.
So,
like
you
have
a
permanent,
you
have
a
permanent
arc
of
the
data.
Yeah.
E
D
So
there
are
sort
of
two
things
going
on
on
our
and
it's
one
thing
is
basically
like
organizing
the
data
and
digitizing
in
such
a
way
that
we
can
easily
read
and
write
like
on
our
en,
and
also
like
a
distribute
some
of
the
interesting
kind
of
analytics
of
the
data
like
a.m.
resistivity
to
the
water
agencies
that
they
can
work
with,
where
like
they
can
use
that
to
better
develop
their
water
model.
D
So
yeah
we're
sort
of
thinking
about
developing
simple,
like
a
simple
code
base
that
they
can
add
a
query
to
beta
download
in
such
a
format
that
they
can
load
it
up
for
their
software
yeah,
yeah
I
know
this
seems
like
yeah
like
a
yeah.
We
were
thinking
about.
Where
are
we
going
to
put
the
beta?
But
that's
that's
sad.
That
seems
like
a
good
way
to
go.
D
C
C
D
And
what
just
to
add
like?
So
what
I
was?
We
were
thinking
we
just
needed
one.
Sometimes,
data
repository
and
archive
like
this
so
I
was
using,
was
trying
to
use
the
read
the
docs,
basically
archive
what
process
we've
done
and
my
just
kind
of
like
explaining
what
the
data
are
and
how
we
have
used
and
how
that
were
processed,
using
whatever
codes
so
which
is
like
a
metadata,
and
we
weren't
quite
sure
how
the
metadata
could
come
in
so
seems
like
just
archiving
and
to
read.
C
You
could
you
can
have
links
back
and
forth,
which
is
actually
kind
of
nice
is
to
keep
the
zoo,
Noto
record
like
reasonably
light,
but
basically
say
like
here's:
the
instructions
for
actually
working
with
these
data
and
then
just
point
to
the
docs
site
from
from
sonoda.
But
it's
it's
nice
having
a
DOI
for
your
data
and
knowing
that
it
is
like
archived
and
sizable
so
yeah
and
that's
what
you
know.
Doe
provides
okay.
A
So
that
would
also
be
maybe
a
good
home
then,
for
some
of
the
data
sets
nothing.
We
have
many,
but
you
know
we
do
have
a
few
data
sets
that
we've
been
traditionally
using
at
UBC,
pursue
some
of
the
papers
and
then
I
with
respect
to
the
Myanmar
data.
You
know
it's
really
those
first
four
or
five
case
histories
that
we
really
want
to
make
sure
kind
of
get
out
and
you
are
serving
as
assemble
it.
A
C
F
Yeah,
it's
pretty
simple,
so
Sarah
boss
sent
me
an
email.
She
was
it
an
issue
on
on
discretize
the
I
thought
we
had
fixed
it,
but
I
guess
for
the
refined
function
for
2d
tree
matches.
It
wasn't
doing
what
we
were
hoping
cyclope
are
fixed
the
issue.
At
the
same
time,
I
thought
I
talked
to
Joe
and
sorry.
F
I
wanted
to
try
to
work
on
the
surface
to
endow
function
because
there
was,
it
was
super
slow
on
three
meshes
and
then
it
kind
of
makes
more
sense
to
be
on
there
discretized
and
to
be
under
synthetic,
so
I
I,
move
that
function.
I
can't
give
that
function
over
to
this
guitars
and
then
fixed
it
simplify
it
up
and
it's
a
lot
faster.
Actually
it's
about.
It's
gonna
be
a
lot
faster
because
we're
doing
for
loops
and
what
nothing
in
that
function
so
and
it
works
for
tree
meshes
and
fencers.
C
F
E
G
G
Because
it's
it's
actually
a
really
important
decision.
You
know
we
want
to
keep
everything
like
xx
is
easting.
Why
is
northing
and
Zed?
It
is
positive
upwards.
I,
don't
know
if
people
want
that,
just
across
the
board,
sim
peg,
no
matter
what
that's
yeah,
that's
fine
and
then
also
what's
our
fourier
convention,
are
we
going
a
plus
or
a
minus
I
Omega
T
within.
C
G
So
right-handed,
okay,
right-handed
Zed,
positive
up
and
plus
I
Omega
T,
which
yeah
and
I
think
that
follows
the
the
analytic
forward
solution
that
I
found
on
EMG
Oh,
see.
That's
that's
where
I've
been
sort
of
developing
my
code
from
yeah
with
warden
important
Holmen,
okay,
that
makes
sense,
but
at
least
for
I
mean
for
like
the
1d
stuff,
the
Zed
was
positive.
Downward,
though
in
warden
home
a
was
a
left
handed
coordinate
system.
Wasn't
it
for
1d
stuff,
yeah.
C
That
that's
quite
possible
but
I
think
well,
unless
others
have
strong
opinions
so
far
everywhere
throughout
simple
we
try
and
keep
said
positive
up,
so
I
would
be
inclined
to
to
try
and
stick
with
that.
I
don't
know
if.
G
G
So
that's
what
I've
been
working
on
and
yeah
and
then
I
looked
through
some
of
the
stuff
that
we
had
for
MT
and
I.
Don't
know
if
John
you're
you're
working
on
the
the
3d
simulation
yeah,
but
when
I
looked
at
the
manual
for
how
the
jiff
codes
compute
the
1d
solution
on
the
boundaries
they
they
do,
a
like
a
1d
numerical
solve
of
a
wave
equation.
So
they
don't.
They
don't
use
this
propagator
solution.
G
G
E
G
G
Yeah
well,
I
think
what
we
wanted
to
do,
those
we
wanted.
We
didn't
want
to
use
a
primary
secondary.
We
wanted
to
forward
model
the
1d
solution
at
the
basically
do
a
1d,
discrete
discrete
solution
on
the
boundaries,
and
we
know
what
the
we
know.
What
the
solution,
or
you
know
what
the
value
is
on
the
top
right.
G
D
G
G
B
D
G
G
You
have
flat
topography.
You'd
only
have
to
solve
the
problem
once
because
then
that
would
be
the
that
would
be
the
you
everywhere
on
the
edges
of
that
that
domain
and
then
you
have
to
solve
it
at
different.
You
have
to
basically
solve
it
a
different
different
each
time.
If
you
had
significant
topography,
I,
don't
know
how
the
interpolation
happens
or
what
goes
on
with
with
that
I.
D
D
C
I'll,
just
stick
that
into
the
reading
notes.
Perfect,
any
any
other
updates.
B
Sorry,
yeah,
okay,
krummy,
nothing
too
major
other
than
I'm
getting
back
into
actually
inverting
empty,
a
real
empty
data
here.
So
this
might
play
into
a
little
bit
of
the
Aurora
thing.
So
I've
got
a
data.
Loader
started
and
I
took
a
look
at
their
elf
file
and
it
is
a
similar
to
what
I'm
loading
so
I.
Don't
know
it
wouldn't
be
much
of
a
stretch
to
get
their
data
in
if
we
want.
However,
we
want
to
go
about
this.
B
C
Would
be
great
one,
and
this
is
up
to
you-
how
we
want
to
sort
of
go
through
things,
but
if
it's
pretty
clear-cut
how
they
could
actually
change
that
yeah.
F
C
Might
be
actually
worth
us
just
spending
time,
actually
writing
the
issue
and
then
sending
that
to
Alexander
I'm
sayin
Melton
this
week
and
he
seemed
very
keen
to
try
and
get
involved.
So
if
this
is
like
a
pretty
clear-cut,
you
know
first
place,
it
could
be
a
great
way
to
sort
of
get
them
involved
and
contributing
a
bit
but
yeah.
B
Okay,
yeah!
Well,
let
me
finish
that
up
and
I'll
send
just
send
on
what
I
got
and
just
take
it,
quick
all
of
them,
because
really
it's
just
making
this
impact
objects,
so
it
just
a
little
few
lines
of
code
just
being
able
to
decipher
where
everything
goes.
Really
is
all
it
needs,
but
yeah
I'll
have
that
finished
up
here
by
the
end
of
the
week.
Expect
some
more
data
coming
in
here,
Friday
so
out
there
they'll
be
ready
by
then.
C
E
Okay,
so
yeah
just
still
been
looking
to
a
bunch
of
the
MT
code.
If
I'm
trying
at
this
point,
I've
been
trying
to
identify
more
areas
where
we
can
take
advantage
of
code,
repetition
and
finite
and
the
frequency
domain
stuff,
that's
already
there
you're
just
looking
through
it,
but
there's
a
lot
of
things
that
are
still
repeated
even
after
the
move
like
this
simulation
refactor,
so
I've
been
trying
to
look,
go
through
and
make
notes
of
that
and
see
where
I
can,
where
we
can
get
rid
of
lines
of
code.
E
It's
always
a
good
thing,
plus
there's
a
few
things
and
the
so
like
I
said
before.
I
have
a
working
kind
of
first
pass
at
creating
the
sensitivity
matrices
for
the
MT
codes
and
I
call
it
a
first
pass
because
basically
kind
of
does
like
a
mini
analytical
solution.
So
what
there's?
No
way
there
wasn't
a
good
way
to
get
the
projection
matrices
out
of
the
freak
out
of
the
mte
receivers
like
that
they
used
to
go
from
the
sources
to
the
series
properly.
So
can
I
just
had
to
do
a
quick
little.
E
You
know
1
and
0
multiplication
thing
to
get
him,
but
at
least
it
was
Isaac.
Okay.
We
need
to
focus
on
these
areas
and
that's
kind
of
why
I
went
back
to
the
e/m
stuff
are
the
frequency
domain
stuff
because
I
think
I
think
that
was
better
suited
to
him,
like
that
was
already
kind
of
coded
to
accept
matrices
or
output
matrices
as
well
or
so
yeah?
That's
why
I
started
looking
at
the
code
Reis.
E
I'm
going
to
be
we're
gonna
and
then
we're
gonna
make
a
conceded
effort
to
get
this
simulation
branch
out
as
part
of
that
and
we're
gonna
start
working
on
the
the
blog
post
and
get
everything
with
that.
I'm
also
Dom
and
John
we're
gonna
and
Lindsey
rainy
tomorrow
to
discuss
simulation
or
the
task
features
and
see
so
basically,
because
I
have
a
few
questions
on
them
and
a
few
things
I'd
like
to
go
over
with
them
specifically
areas
in
the
code,
just
ok.
Why
did
you
do
this?
E
E
But
it
would
be
something
that
if
people
wanted
to
use
to
ask,
they
would
say:
okay
imports
in
peg
desk
and
then
that
would
replace
everything
that
would
be
in
simp,
ed
with
tasks
versions
that
we've
implemented
so
far,
just
as
an
option
I've
been
kind
of
thinking
about
that
a
little
bit
more
and
just
kind
of
general
thoughts
is
that
so,
if
let's
say
someone
was
to
implement
their
own
sort
of
like
open
air
MPI
type
code
approach
to
the
simpad
paralyzation,
how
well
does
that
play
if
it's
already
paralyzed
with
desk
was
my
question
right.
E
So
that
would
be
one
of
the
reasons
why
I'd
kind
of
lean
to
doing
this
kind
of
like
approach.
So
it's
like
okay.
Well,
if
we
do
want
a
desk
simulation-
and
we
specifically,
you
know,
ask
for
that
like
so
that
was
kind
of
one
of
the
other
things
that
was
kind
of
in
my
head
as
I
started.
Thinking
about
this
a
little
bit
more
I.
C
E
Just
those
are
thoughts
I've
had
in
my
head
and
that's
something
I
like
is
I
want
to
get
that
ironed
out
before
this
gets
put
onto
this
release
really
quickly
and
I
think
it
should
be
pretty
cleanly
to
be
able
to
implement
that
as
well,
and
it's
the
same
fit
same
way
that
if
you've
ever
looked
into
the
auto
grad
package,
if
you
it's
a
thumb,
it's
a
package
that
automatically
calculates
numerical
different,
Derek
or
intiation
using
Auto
gradients,
and
things
like
that.
C
E
D
Cool
okay
that'll
be
great
if
you
could
do
that,
but
I
mean
sometimes
like
that
implementation
is
like
convolved,
with
the
with
the
application
and
sort
of
situation.
So
I
mean
like
that,
like
kind
of
embarrassing
ly,
parallel
problem.
So
all
the
problem
like
let's
say
we're
just
spell:
izing
simple
for
then
looping
over
I
mean
it
is
absolutely
possible
to
do
that.
E
F
All
be
great
because
it
was
something
we're
missing
on
the
simulation
right
now
is
to
give
all
to
you
the
to
compute
the
best
compute
on
the
entire
misfit
or
realization.
It
calls,
and
we
probably
don't
want
to
have
to
embed
it
so
so
deeply
intercourse.
If
we
can
step
right
up,
you
know
if
you
can
have
two
two
versions:
basically,
that
will
be
the
optimal
yeah.
E
C
Guess
is
like
one
thing:
I
would
appreciate
your
thoughts
on
a
night
I'd
understand
too,
if
you
don't
have
an
answer
yet,
because
we
haven't
sort
of
dug
into
this,
but
how
much
code
repetition
do
you
think
that
that
would
introduce
like?
Would
we
need
to
then
sort
of
maintain
two
versions
of
Jay
Beck
for
the
frequency
domain
code
and.
E
E
E
We
might
be
able
to
do
in
an
intelligent
way
that
kind
of
reuses
a
lot
of
it.
So
sometime,
okay,
well,
it's
like
if
we're
just
parallelizing
over
sources
and
or
for
paralyzing
over
frequencies,
then
do
a
specific
operation.
That
literally
makes
a
survey
up
by
sources
and
then
just
generates
a
bunch
of
them.
You
know,
and
then
we
uses
that
code
or
something
my.
F
E
C
E
A
C
So
I
took
the
document
that
we
were
sort
of
brainstorming
in
on
the
natural
source
methods
and
just
turned
that
into
a
discourse
post
so
that
it's
something
that
we
can
loop
folks
like
at
Aurora
and
on
as
well
as
the
Jupiter
meets
the
earth
project,
so
they'll
be
doing
some
like
tubular
development
kind
of
based
on
feedback
that
we
provide,
and
things
like
that
so
wanted
to
loop
them
in.
There
might
actually
be
so
I
know
we're
holding
a
meeting
next
next
week.
C
So
I
think
Fernando's
gonna,
try
and
join
for
a
bit
and
Eric
who
is
a
Jupiter
hub,
developer
and
he's
contracting
through
Berkeley
right
now
and
doing
work
specifically
for
this
project.
He
might
drop
in
and
say
hello,
but
down
the
line.
I.
Imagine
that
yeah
they
might
get
a
bit
more
looped
in
just
as
we
get
some
ideas.
I
think
our
parallelization
is
going
to
be
a
cool
place
to
connect
with
them
on,
because
Eric
has
a
really
strong
knowledge
of
kubernetes
and
knows
how
to
work
that
whole
space
very,
very
well.
C
C
A
couple
of
oh
yeah,
so
I
was
just
chatting
with
dieter
on
slack
over
the
thing
over
that
weekend
and
he
was
asking
if
we
might
be
willing
to
have
a
meeting,
maybe
once
a
month
that
is
a
bit
earlier,
so
that
he
could
join
he's
interested
in
getting
more
and
I
know.
He
watches
I,
think
all
of
the
meetings
and
provides
feedback,
and
things
like
that
on
slack,
so
wanted
to
float
that
by
folks
and
maybe
see
if
I
don't
know
something
like,
maybe
the
first
first
Wednesday
of
the
month.
C
A
C
C
C
Yes,
so
what
we
can
do
is
all
first
confirm
that
with
him
to
see
if
that
works
and-
and
we
can
see,
yeah
cuz
I
think
there
might
actually
be
a
couple
other
people
who
who
would
tune
in
from
from
Europe
if,
if
that
works,
should
be
cool,
yeah
and
then
the
last
thing
I
just
put
a
link.
I've
been
working
with
Doug
Fernando
and
with
Lauren's
Ferran
on
a
sec
abstract
for
the
machine
learning
UXO
work,
so
I
just
put
a
Dropbox
link
there.
C
C
Oh
and
with
a
discourse
post
the
way
that
I
set
that
up
so
this
one
is
set
up
as
a
wiki,
which
means
that
anyone
can
edit,
like
the
main
thread
that
I
put
in
there,
so
feel
free
to
add
stuff.
If
you
think
I
miss
things
and
then
we
can
sort
of
just
have
that
be
like
you
know,
one
coherent
document,
rather
than
sort
of
having
a
bunch
of
threads
and
posts
from
other
people,
so
don't
hesitate
to
go
in
and
edit
that
text.