►
From YouTube: SimPEG Meeting March 4
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
B
C
B
B
B
E
B
B
B
F
B
Through
an
update
all
the
point
at
which
they
will
be
deprecated,
so
that's
something
I'd
like
to
talk
about
or
discuss
a
little
bit.
They
saw
all
the
changes
from
simulation
I'm
doing
my
best
to
kind
of
keep
functionality
of
what
was
previously
there,
but
it's
like
we
have
to
kind
of
decide
on
the
point
at
which
it
will
break
like.
B
B
B
E
G
F
B
F
H
That's
great
jobs,
kind
of
been
talking
about
the
simulation,
branch-like
cool
changes,
I
think
I,
know
I've
heard
some
of
them
just
floating
around
in
these
meetings.
That's
one
and
I'm
at
the
point,
where
I'm
going
to
start
doing
a
paramedic
conversion,
but
I,
don't
know
I'm
going
to
actually
impact
as
driver
for
that
or
I'm
just
going
to
code
it
up
in
kind
of
its
own
thing,
just
because
it's
a
it's
a
very
simple
problem
and
I,
don't
think
it's
worth
going
into
anything
synthetic
for
all
of
that.
E
C
A
A
H
A
Marini
and
exam
all
know
he's
got
like
see.
He
seems
to
be
happy
with
step
off,
but
he
wants
to
model
impulse
function
which
I'm
not
as
I.
Let
him
know
how
do
you
do
that,
but
that
I
think
you
are
having
a
little
bit
of
issue
so
and
the
other
other
example
like
as
it
is
Italian
girl
working
on
the
loop
system
but
yeah.
So
how
like
how
the
current
time
domain
code
works
it's
working,
but
as
an
outside
user,
it's
very
hard
to
understand.
A
A
Exactly
so
sorry,
yeah
I'm
expecting
to
them
to
have
a
hard
time,
but
this
girl
was
expanding
a
year,
trying
to
use
simpang
and
not
getting
a
single
inversion
result
right.
So
yeah
I
thought
that's
like
that's
kind
of
pessimistic,
I,
don't
know
yes,
I
mean
it's
in
so
many
ways
like
writing.
A
good
tutorial.
He
making
a
good
example
yeah
other
than
that
I
was
not
quite
quite
sure.
Yes,.
E
There
was
an
example
that
Peter
sent
to
lying
that
he
was
getting
really
frustrated
with
and
there
were
just
two
things
that
were
slightly
wrong
as
his
padding
didn't
go
far
enough,
so
there
were
boundary
condition
effects
and
then
his
time
is
critias
asian.
Of
course
it
was
just
like
too
quickly
at
be
at
the
end
and
yeah.
G
A
I'm
trying
to
enforce
them
to
post
on
this
course
yeah,
but
I,
think
about
not
all
of
them
are
coming
into
to
the
surface,
I
think
and
sometimes
I'm,
just
giving
like
a
simple
solutions
like
oh
I
fixed.
There
was
parameter
in
the
notebooks,
try
that
so
hopefully
at
the
end
like
if
they
can
put
together
like
at
least
like
a
single
notebook
that
runs
and
shows
how
they
solve
the
problem.
That'll
be
that'll,
be
nice,
so
we're
asking
people
to
do
it,
but
yeah
not
exactly
sure
but
yeah.
Like
not
sure.
E
F
Guys
I
just
I
just
pushed
it
test
and
clean
up
an
example
for
the
tile
math.
So
the
for
requests
on
to
summation
and
you
asked
for
it
so
I
I
did
this.
I
can
I
can
demo
if
you
guys
are
clear
at
you
want
what
the
example
looks
like
it's
very
simple,
but
that's
a
trick
and
then
the
test
should
cover
basically
the
cover
that
that's
simple,
math
yeah.
H
B
F
It's
alright,
there's
zero
iterations
on
them,
so
you
will
feel
good
right
right
away.
Okay,
never
mind
man
yeah!
It's
just
I'll
do
other
than
this
I'm
doing
more
buoys
for
Wendy
frequency
and
time
domain
EMS
for
for
our
sponsors
that
works.
That
works
really
well
I'm,
still
getting
a
lot
of
mesh
mesh
effects,
though,
on
the
on
the
on
the
connectivity
models.
So
that's
kind
of
like
what
I'm
working
on
right
now
rather
than
in
Gandhi
yeah
the
stitch
this
one,
the
every
time
the
mesh
increases
in
size.
A
A
F
A
So
what
kind
of
what's
what's
a
typical
use
case
in
here
is
just
like
using
like
a
some
factor
like
one
point
one
you
start
from,
like
whatever
you
define,
whatever
the
smaller
cell
size
and
then
you
slowly
increase
like
like
one
point.
One
is
a
factor
that
I'm
using
but
yeah
with
that
I
haven't
really
seen
too
much
ship
there.
That's
effect,
yeah
I,
don't
know,
but
that's
that's
what
we're
doing
yeah
now,
maybe.
F
A
F
Other
than
this
there
is
a
we
starting
to
get
attraction
here.
There
is,
there
was
desire
to
push
the
SEM
T
is
like
I
asked
you
right
front,
we're
gonna
empty
as
well,
so
I
think
the
next
couple
months,
we'll
need
well
I'm,
going
to
first
people
of
working
hours
on
those
codes,
and
if
anybody,
if
no
one
wants
to
get
involved,
they
could
be
like
no
financial
support
for
us.
A
A
H
A
F
E
I
Again,
working
on
empty
side
of
things
here,
actually
Joe
posted
there
that
little
over
look
they're
just
doing
the
one
matrix
multiplication
and
yeah
works
well
on
the
test
or
from
the
test
side.
It
was
like
it
usually
runs
at
about
a
hundred
seconds,
and
now
it's
down
to
about
56
seconds
nice.
It
works
pretty!
Well,
maybe
I'll
try
to
do
that
on
the
DC
side
as
well.
There
I.
B
H
B
A
B
It's
gonna
be
most
apparent
and
like
okay,
the
MT
codes,
where
it's
like
you
have
real
imaginary,
like
I,
said
real
and
imaginary
components
for
each
I.
Think
impedance
and
Tipper
is
steeper
rate
yeah,
like
12
yeah,
like
frequency
domain.
If
you
have
real
and
imaginary
receivers
or
or
different
components,
orientations
sometimes
right.
D
I
I
D
Yeah
I
mean
I,
guess
ultimately,
where
my
mind
is
going,
is
I'd
love
to
be
able
to
do
with
through
comparison
of
the
the
3d
MT
through
simple
and
then
the
3d
MT
from
the
jiff
codes,
because
I've
been
using
them
a
lot
and
I
love
them.
A
lot
and
I'd
like
to
be
able
to
find
some
kind
of
comparison
or
fine,
finally
be
able
to
go
back
and
forth.
No,
so.
D
I
D
You,
when
you
define
a
receiver,
would
you
define
the
component
say
like
I,
want
the
real
component
of
Z,
X,
Y
and
or
the
imaginary
component
of
said,
Y
X
and
that's
how
you're
defining
receivers?
That's
correct,
yeah,
okay
and
you
haven't
you're,
not
thinking
about
doing
any
anything
in
terms
of
an
arbitrary
orientation
like
you
could
have
your
your
let's
say
your
X
being
along
the
northwest
or
something
like
that
or
no.
E
H
D
I
mean
with
with
the
MT
stuff
that
raw
data
seems
to
come
in
and
it's
been
put
in
a
coordinate
convention
that
has
an
alignment-
that's
north,
south
east
west.
You
may
have
to
flip
things
around,
but
at
least
it's
it's
a
low-lying
north
north
south
east
west
for
Zed
team,
though
it
can
it's
a
long
line
in
cross
line
and
that
that
was
a
bit
of
an
eye-opener,
getting
it
into
a
format
that
you
can
invert
and
interpret
so
I.
Guess
to
make
these
things
practical,
which
I
guess
is
the
point.
I
C
A
Right,
it's
a
back
to
GWB
I,
just
got
kind
of
that
couple
of
question.
Well,
what's
the
exactly
problem
that
you're,
having
like
it's
a
memory
shoe,
are.
D
You
asking
me
yeah,
yeah
yeah
I
mean
I'd
I.
C
B
B
However,
we
can't
it's
not.
We
can't
really
get
around
it
if
we're
storing.
If
we
want
to
compute
the
whole
J
transpose
or
the
full
J
matrix
right,
because
no
matter
what
you
have
to
do,
a
solution
for
each
data
point,
but
at
least
it
makes
the
other
things
faster
and
we
can
take
advantage
of
the
reduced
space
of
storing
the
fields
because,
okay,
the
issue
is
that
there,
like
there's
1500
fields,
20
thousand
model
cells
and
then
there's
16s
wave
numbers
right.
B
So
it's
it's
all
of
those
like
it's
16
times,
fifteen
hundred
twenty
thousand
and
it
gets
I.
We
were
like
looking
at
the
fields,
object
itself
and
I
was
taking
up
like
probably
about
four
gigabytes
on
its
own,
but
that
was
that
the
fields
object
itself
was
what
16
times
larger
than
the
sensitivity
matrix
right.
B
B
A
Yeah,
so
there
there
are
like
yeah.
There
are
number
of
ways
to
kind
of
make
a
speed-up
and
also
reduce
the,
but,
like
I,
saw
this
playing
with
just
the
object,
but
most
of
cases
I
don't
need
that
many
sales,
like
twenty
thousand,
is
a
lot
so
like
so
like
I
was
working
on
that
onsen
dcd
de,
like
if
I
like,
generate
deep,
like
relatively
kind
of
small
auntie
sales,
but
I
think
I'm
not
doing
a
and
not
sacrificing
my
accuracy
70
like
still
doing
an
okay
job
generating
the
Yaquis
match.
A
Type
like
I
just
need
couple
dollars
in
sales,
and
that's
usually
enough
for
handling
his
2d
data
and
in
such
a
case
like
I
yeah,
so
even
like
so
this
window,
Schlumberger
takes
longer
because
we
got
much
more
sources,
but
yeah
memory
was
not
an
issue
for
me.
So,
yes,
it
depends
up
on
us
like
Oh.
What
do
we
really
want?
A
D
A
A
B
A
D
I
got
a
I,
have
a
few,
so
I
went
and
looked
at
the
sim
peg
tutorials.
We
have
ones
on
gravity,
magnetics
and
dcpip,
so
this
would
cover
all
of
I
would
say
the
standard
things
you
would
want
to
do.
You'd
want
to
forward
model
in
3d
you'd
like
to
try
out
tensor
meshes
and
octree
meshes
you'd
like
to
think
about
typography
you'd
like
to
do
an
l2
inversion,
maybe
you'd
like
to
do
related,
least
squares.
D
So
all
of
these
kind
of
standard
things
that
you
would
do
for
those
three
methods
we
have
tutorials
for
and
I
went
in,
and
I
I
fixed
a
couple
bugs
in
the
examples.
So
there
is
there's
just
a
whole
gallery
of
examples
on
simpang,
so
I've
been
trying
to
fix
any
kind
of
bugs
or
problems
in
those,
so
that
pull
request
is
going.
Travis's.
D
Timing
out
and
I've
got
a
couple
other
little
things
in
the
dark
test
that
fails,
but
they're
more
related
to
the
not
keeping
up
with
the
examples
gallery
as
opposed
to
the
tutorials
I've
made.
So
that's
basically
ready
and
once
that's
put
in,
we
can
always
add
to
it
and
yeah
I've
got
some
interest
in
the
MT
stuff.
I,
don't
know
how
soon
I'll
get
to
it,
but
I
definitely
like
to
be
involved
in
the
final
product
and
making
tutorials
for
that.
D
With
regards
to
the
gwb
stuff,
the
inversion
tutorial
video
has
been
posted
so
now
they
can
basically
do
everything,
except
for
that
electro
pro
2
thing,
so
they
bought
the
resources
to
to
take
data
from
the
virus
process
it
and
then
invert
it
and
get
a
reasonable
result
and
I
made
one
of
the
t.
Ot
s
an
administrator
on
the
Facebook
webpage
just
long
term.
We
need
I,
mean
we're
not
going
to
be
involved
forever.
C
D
D
E
You
don't
do
that
automatically
for
you,
because
we
know
like
cuz.
It
knows
which
organisation
here
a
part
of
yeah
fair
enough.
So,
if
like
when
it
looks
for
relevant
material,
the
one
of
the
first
places
that'll
Marcus,
is
within
within
that
same
YouTube
organization.
So
if
you're
comfortable
with
it
it'll
make
it
much
easier
for
other
people
to
find
to
find
these.
E
D
I
guess
the
other
reason
for
this
is
right.
Now
it
doesn't
look
like
any
of
the
content
of
the
video
will
need
to
be
changed,
but
I
can't
replace
an
old
video
with
a
new
video.
You
have
to
fully
get
rid
of
the
video
on
youtube
and
then
post
a
new
one
and
update
the
link
and
if
people
are
getting
used
to
putting
the
link
in
to
YouTube
and
something
gets
updated
and
it
changes,
maybe
in
the
end
I'm
just
kind
of
encouraging
them
to
go
to
the
Facebook
group
as
their
starting
point.
B
D
G
D
The
dcpip
is
going
to
be
quite
an
undertaking
and
then
there's
a
lot
of
utilities
that
are
defined
from
scratch
in
geo
Scilab,
but
these
utilities
have
actually
been
added
to
this
impact
framework.
So
I
don't
know
if
I
want
to
clean
things
up
as
I'm
doing
the
stuff
now
or
just
get
everything
to
run
on
simulation
branch
and
then
think
about
cleaning
it
up
after
the
fact.
E
There's
some
stuff
that,
like
it's
a
bit
unclear
or
like
it
doesn't
work
out
of
the
box,
I
think
started.
Getting
getting
getting
it
onto
simulation
in
somewhat
timely
fashion
is
probably
the
most
important
piece
because
there
are
people
there.
Other
folks,
interested
in
the
education
side
of
things
who
were
gonna
rely
on
them.
B
B
I
was
like
okay,
just
you
know
batch
rename,
this
switchable
B's
just
to
get
it
done
quickly,
but
sometimes
I'm,
like
oh
I,
see
what
they're
doing
here
and
it's
a
utility
I
just
switched
out
really
quickly,
like
just
I,
would
say:
first
priority,
just
get
it
working
and
if
you
notice
things
when
you're
going,
you
can
either
like
make
a
comment
in
the
page.
Anything
like.
Oh,
this
has
been
infected
in
synthetic
and
then
just
go
on
or
something
or
just
changing
yourself.
It's
if
it's
easy,
yeah.
D
Fair
enough,
but
yeah,
that's
so
that's
about
it
for
me,
I'm
really
keen
on
getting
the
tutorials
merged
to
based
emulation
branch
as
soon
as
possible,
so
that
it
will
be
updated
as
stimulation
branch
of
balls,
yeah
stuff
put
some
time
into
GOC
labs
as
I.
Do
other
work
that
I'm
on
the
hook
for
I
have.
C
B
D
D
You
know
showing
somebody
how
to
use
simple
to
solve
a
specific
problem,
whether
or
not
it
is
a
like
synthetic
Ford
modelling
example
and
in
the
end
you'd
like
to
create
this
survey
and
plot
your
your
results,
or
you
know,
I'm
going
to
start
with
some
data
and
some
topography
and
ultimately
I'd
like
to
get
an
inversion
result.
So
these
are
like
very
deliberate
tutorials
that
are
going
to
show
off
some
very
specific
things.
It's
really
practical.
B
It's
like,
if
you
like
some
examples,
should
be
more
like
that,
and
essentially
it's
like
okay
well
I
mean
this
example
shows
how
to
do
a
magnetic
inversion.
You
know
this
example
shows
how
you
would
do
a
general
iterative
least
squares,
inversion
and
then
I
don't
know
it
would
be
like
they
they're
almost
like
tone
down
examples
and
more
concise
or
the
tutorials
are
generally
waste.
I
would
think
that
yeah.
D
D
G
E
D
And
I
think
one
of
the
things
about
those
that
examples
gallery
is.
It
was
also
an
opportunity
to
put
the
results
of
codes
that
we
published
or
we
used
to
publish
papers,
it's
kind
of
an
advertising
and
say
look
at
the
actual
practical
work.
We've
done
published
work
that
was
done
using
simple
I,
think
that
was
an
ideas
as
well
as.
G
D
I
guess
there
was
this
idea
at
the
beginning
of
the
website,
was
gonna
have
kind
of
an
introductory
section.
You
know
getting
started
what
is
simple,
but
also
something
that
showed
examples
to
hook
people
in
that
way.
There
was
a
few
people
that
expressed
that
kind
of
interest
on
wanting
to
just
show
something
up
front
right
away.
That
would
keep
people's
interest
and
show
that
you
can
do
a
lot
of
stuff
with
this
so
that
they
decide
to
keep
reading
and
actually
invest
their
time.
That
was
the
thought
you
know
say.
B
G
G
H
G
F
C
C
C
E
E
G
B
H
B
Fields,
object
takes
a
long
time
to
access
in
general,
like
especially,
if
you're
accessing
it
a
single
source
at
a
time
even
like,
if
you
get
have
a
list
of
all
sources,
so
like
this
change,
I'm
like
doing
this
you're,
just
storing
all
of
them
at
once
with
the
operator
like
that,
was
actually
pretty
significant.
Yeah.
G
B
That
single
change,
like
save
a
lot
of
time,
the
right-hand
side
in
general,
was
being
constructed
as
a
be
ordering,
whereas
this
solver,
like
the
party
stow
solver,
expect,
was
wanting
it
to
be
Fortran
ordered,
so
it
can
fertilize
it
efficiently
over
columns.
Oh
right,
I
mean
there's
no
like
we
can
be
there
and
just
kinda.
B
Yeah
what
else
oh
I
was
kind
of
talking
with
soggy
a
little
bit
about
this.
The
I
made
some
simplifications
in
the
operation
for
the
sensitivity
matrix.
So
instead
of
like
doing
the
computing
all
the
trapezoidal
integration
stuff,
like
recursively,
almost
like
one
step
at
a
time
right
for
each
KY,
compute
the
weights
for
the
trapezoidal
integration
ahead
of
time,
and
that
makes
this
a
vector
like
so
much
easier
right,
I
guess
a
lot
here
than
what
it
used
to
look
like,
and
that
also
made
it
makes
it
easier
to
kind
of
see.
Okay.
B
This
is
the
kind
of
thing
that
I
was
talking
about
like
how
like
a
really
quick
way
to
implement
that
kind
of,
so
that
J
transpose
vector
operation
of
thumbing
them
and
then
doing
this
solved.
So
you
just
keep
track
of
these
and
sum
them
up
like
that,
pretty
quickly,
actually
as
numpy
race
and
then
do
the
solution
from
what
I
was
seeing.
It
took
actually
longer
to
do
this
operation
Jenna.
Wouldn't
you
generally
solve
things
yeah,
so
just
kind
of
looking
at
that.
B
So
then
I
started
looking
at
that
as
well
and
there's
I
mean
there's
a
few
little
things
that
that
week,
that
I
could
do
to
speed
it
up
in
general,
but
generally
it's
like
it's
avoiding
constructing
a
sparse
matrix.
If
we
don't
actually
need
to
it
takes
time
to
build.
It
takes
time
to
like
constructed.
Even
if
it's
a
diagonal,
it
takes
additional
time
and
we're
not
really
using
it
to
do
anything.
Yeah.
B
G
B
E
B
B
B
B
B
B
Like
just
kind
of
like
things
like
it's
a
night
like
I,
said
it's
me
old
trick
there.
That
was
basically
about
it.
B
B
The
port
was
looping
over
a
bunch
of
stuff
and
would
I
think
there
was
a
small
error
in
it,
because
when
y
was
not
equal
to
zero,
I
would
get
vastly
different
was,
and
the
like.
This
operation
matches
what
what
none
PI's
trapezoidal
integration
would
do.
Okay
for
the
expected
function
so
like
the
expected
function
that
we're
trying
to
do
the
integral
or
the
integrate
right.
Oh,
this
is
a
PF
as
a
function
of
cosine
or
a
function
of
the
things
or
trying
to
integrate
with
trapezoid.
B
Is
this
RF
x
cosine
like
K
times
y
right?
That's
the
function,
we're
trying
to
integrate,
and-
and
this
here
matches
like
I
said,
dump
ice
and
numpy
is
trapezoidal
weights
when
I
couldn't
get
it
to
match
with
some
of
the
other
one
like
with
what
it
was
before,
but
it
was
like
I
said
it
was
only
Bess.
It
only
came
about
when
y
was
not
equal
to
zero
because
obviously
cosine
times
cosine
of
zero
wasn't
gonna
change.
Anything
yeah.
A
B
G
F
H
F
G
F
H
F
H
F
So
first
it
said
loop:
do
you
need
to
create
your
measures?
First,
okay
and
then,
once
you
have
your
mentions,
you
create
your
global
mesh
and
then
this
is
pretty
important,
because
that
will
guarantee
that
your
global
mesh
will
always
gonna
have
the
smallest
cell
size.
You
always
want
to
imagine
eyes
down
the
global
mesh,
not
not
your
way
around.
F
F
You're
gonna
use
like
let's
do
typography
as
your
as
your
initiator
to
initiate
the
full
extent
of
the
global
match,
but
you're
going
to
refine
only
on
the
on
the
local
surveys
right.
Okay,
that
gives
your
local
measures
and
then
you're
gonna
create
your
your
global
match.
By
just
grabbing
the
cells.
The
cells
I
beat
your
goal
mesh
to
refine.
F
A
F
G
F
No,
like
yeah,
it's
gonna
I
joined
with
a
few
problems
and
splitter
like
yeah,
an
optimization
curve.
You
know
at
the
beginning,
you
know
you're
gonna
save
a
lot
of
time
and
lot
of
memory,
but
eventually,
as
you,
you
have
a
minimum
number
of
cell
that
you
need
or
each
each
points
so
you're
gonna
end
up
by
just
costing
you
more
time
to
to
generate
and
store
those
matches.
F
A
So
I
mean
like
what
kind
of
the
design
of
this
tile
map
is,
like
probably
handling
we
say
several
meshes
and
that's
like
is
that
sort
of
the
scope?
Okay,
when
you're
running
the
a
single
inversion
you
may
generate,
like
not
thousands
of
the
match,
but
like
ten
Mitch,
where
six
to
seven
mesh
or
something
like
that,
yeah.