►
From YouTube: SimPEG Meeting March 3rd
Description
Weekly SimPEG Meeting from March 3rd, 2021.
B
B
C
B
Yeah,
I
did
testing
with
the
gradient
array
back
into
dc
and
the
you
know
the
cip
3d.
If
you
have
a
few
sources,
but
lots
of
data
or
a
big
mesh,
it's
so
slow
right.
The
realization
doesn't
really
kick
in,
so
we're
really
killing
it.
You
know
to
be
able
to
slice,
slice,
transmitters
and
yeah.
It's
we
see
the.
We
see
the
competitive
advantage
of
pushing
the
code,
so
it's
gonna
happen
for
sure
tom.
D
Thursday,
okay
yeah
tomorrow
at
8
year
time,
I
think,
yeah,
because
I
I
just
had
a
little
chat
with
miguel
and
it
looks
like
they're
far
enough
that
we
could
at
the
hackathon,
which
is
in
a
month
or
so
that
he
could
like
bang
our
heads
together
and
bring
it
into
different
libraries.
So
one
would
be
simpek,
but
the
other
one
would
also
be
interfaced
with
two
h5
pi
yeah.
E
D
E
D
A
B
Sounds
great
yeah
and
that
that's
the
tricky
part
right
now
we
have
with
with
tiling,
is
again
I'm
reacting
the
same
stuff,
but
the
receiver
locations
are
kind
of
hidden
down
down
the
pathway
and
it's
kind
of
hard
to
to
figure
out
the
spatial
correlation
between
transmitter
and
receivers.
You
kind
of
need
to
go
down
the
down
the
chain
and
come
back
anyway,
yeah
that
would
be
cool,
though
joe
go
for
it.
F
F
F
F
A
F
F
So
if
you
did
then
get
a
warning,
that
said,
you
need
to
assign
topography
and
you
would
those
points
are
above
the
surface
and
will
then
find
the
closest
surface.
I
think
for
tensor
meshes.
It
doesn't
matter
where
those
z
components
are,
if
you
say,
reset
the
z
value
based
on
topography.
It
will
find
the
surface,
but
I
think
with
tree
meshes,
it
might
actually
find
just
the
closest
active
cell
center
or
node.
F
But
the
general
idea
is
that
once
when
you
load
any
data
into
simpeg
for
dcip
either
whether
or
not
it's
2d
or
3d,
that
last
positional
value
always
has
to
be
the
elevation,
so
you
could
either
load
a
long
line
in
elevation
or
you
could
load
x
y
zed,
but
you
wouldn't
be
able
to
load
surface
data
with
like
a
null
elevation
location
and
then
play
around
with
it.
It
would
it
would
stop
that
from
happening.
F
A
The
I
know
that
stream
still
needs
to
be
updated,
dominic
put
an
issue
that
was
like.
Basically,
I
made
a
comment
on
an
issue.
I
made
an
issue
already
about
it
discussing
it,
but
it
was
like
so
what
the
tree
mesh
does
right
now
with
that
jade
circle
function
is
it
was
written
to
just
literally
find
the
closest
cell
and
move
them
to
that
cell.
F
It's
it
shifts
stuff,
and
I
noticed
that
in
some
of
the
scripts
I
was
writing
and
it
would
be
nice
if
it
actually
projected
downward
and
if
it
was
say
three
quarters
the
distance,
it
would
actually
go
straight
down
instead
of
just
because,
because
it
will
project
horizontally
at
times.
Yes,
I
noticed.
A
It's
still
abstract.
We
essentially
just
need
to
extract
the
boundary
cells
from
a
tree
mesh
like
along
that
topography
surface,
which
there
are
functions
to
do
that
right
now
and
then
like
generate
a
2d
tree
mesh
based
off
of
that
topography
and
then
project
it
onto
that,
because
you
can't
just
use
the
closest
like
a
nearest
neighbor
tree,
looked
up
for
because
it
won't
work
great
all
the
time.
F
F
F
F
F
C
Yeah
yeah
nothing
too
much
just
yeah
I
hooked
yeah
dom,
went
in
and
changed
the
receivers
a
little
bit
using
the
iron
sum
stuff
for
the
impedance
on
the
mt
receiver.
So
I
just
updated
that
for
the
for
the
apparent
res
and
phase
so
now
it's
working
with
tiling
made
a
couple
examples
with
it
and
yeah
now
I
actually
doubled
into
the
grounded
source
frequency
domain
stuff
playing
around
with
it.
Yeah
we're
doing
quite
a
bit
of
cs
amt.
Here.
C
A
A
A
You
know
volume
into
the
integrals
in
the
spherical
domain
on
a
crazy
measurement
right
as
far
as
as
opposed
to,
like
you
know,
creating
a
weird
shaped
mesh
when
you're
trying
to
approximate
a
volume
into
the
wall
in
a
weird
shape
measure
that
you
don't
know
the
analytical
you
know
answer
to
but
anyway.
So
that's
what
I'm
trying
to
do.
A
I
can't
figure
out
if
I'm,
if
the
errors
are
in
like
using
senpai
to
do
spherical
integral
integrals,
which
certainly
possible,
but
it
I
think
I've
double
checked
and
it's
giving
me
the
error
like
for
simple
functions.
It's
giving
me
back
what
I
expected
to,
but
I
I
can't
get
like
the
numerical
result
from
this
mesh
to
agree
with
the
analytic
result
from
the
integral
from
senpai
and
it's.
A
I
can't
tell
if
it's
happening
on
like
the
gradient
of
the
divergence
operators
when
they're
multiplied
by
things,
but
I've
been
trying
to
figure
that,
like
I'm,
not
concerning
myself
with
the
boundary
conditions
just
yet
I'm
just
trying
to
approximate.
You
know
a
gradient
of
a
function
times
a
vector.
A
I
think
we
can
get
those
out
first
and
then
keep
adding
on
some
of
those
automated
building
of
those
other
matrices
as
we're
going
on,
because
we
can
start
working
off
of
this
and
incorporating
it
into
like
the
because
the
mt
branch
that
has
the
2d
stuff
will
be
reliant
off
of
it.
And
I
want
to
get.
F
F
A
E
A
So
what
I
mean
is
like
we
don't
I
don't.
Let's
say
we
have
like
a
robin
boundary
condition
that
is
a
little
bit
different
for
each
type
of
those
like
the
weak
form.
So
right
now
I
only
have
the
robin
boundary
conditions:
weak
form
of
the
cell
gradient,
but
the
pieces
are
already
there
to
implement
them
for
the
other
parts.
F
Yeah,
I
I
mean
I,
I
think
that's
a
good
idea.
I
just
want
to
be
really
careful
about
what
it's
actually
called,
because
I
think
we
had
one
kind
of
situation
where
it
was
like
note.
When
you
ask
for
nodal
gradient,
it
makes
complete
sense.
It's
it's
actually
a
discrete
gradient
operator.
It
goes
from
nose
to
edges,
but
then,
historically,
when
you
were
calling
the
cell
gradient
operator,
it
wasn't
actually
a
cell
a
gradient
operator.
It
was
that
with
an
inner
product
matrix
because
it's
it
incorporated
the
boundary
conditions.
A
So
basically,
the
what
we
were
talking
about,
I
think
what
foggy
and
lindsey-
and
I
were
discussing
earlier-
is
that
so,
when
we're
doing
the
cell
gradient
right,
we're
doing,
we
use
the
face
divergence
for
the
weak
form
of
the
cell
gradient
yeah.
So
it's
named
after
okay
cell
gradient,
weak
form,
boundary
condition
yeah.
F
Yeah
yeah
and
I
guess,
when
you're
talking
about
putting
all
the
things
together
for
the
user,
I'm
wondering
like
to
what.
To
what
extent
are
you
you
doing
all
the
work
for
them,
because
I
I
would
like
the
I
the
idea
that
if
you
were,
if
you
went
through
the
derivation
did
the
week
formulation-
and
you
said
I
want
a
a
cell
gradient-
that
you
would
actually
be
saying.
Actually,
no,
I
want.
I
want
divergence,
transpose
plus
this
piece.
A
G
H
A
A
A
Obviously
that's
like
we
need
to
get
that
as
well
in
there
for
that.
I
guess
it's
something
that
could
be
useful
there
yeah,
so
it
shouldn't
it
should
be
closed.
It
should
be
able
to
do
it
pretty
quickly
because
it's
just
a
matter
of
you
know
putting
it
in
the
right
place
afterwards.
That's
not
so
what
ends
up
happen?
The
reason
we
just
was
talking
with
us.
So
if
you
have
a
forward
right
now,
we
have
the
cell
grading
operator
that
goes
from
cell
gradients
to
faces.
A
So
if
you
do
that
and
then
you
average
the
faces
back
to
cells,
it
gives
you
a
like
a
centered
cell
gradient
and
ends
up
with
like
a
weird
checkerboarding
pattern,
and
it's
just
kind
of
weird,
so
you
kind
of
have
to
use
it
for
the
cross.
Gradient
sometimes
like
that,
you
need
a
forward
slow
grading
operation.
A
And
yeah,
I'm
still
definitely
I'm
open
to
their
names
as
far
as
those
like
boundary
integral
products.
I
think
I've
talked
about
the
last
time.
I
was
showing
you
guys
looking
for
inputs
on
what
to
name
some
of
those
like
the
equivalent
of
those
inner
product
matrices
so
like
in
inner
products,
one's
called
the
edge
inner
product.
It's
just
that,
there's
a
face
in
our
product,
matrix
access
to
column,
boundary,
integrals,
boundary,
indigo
matrix.
A
A
F
I
mean
we
don't
want
to
be
like
saving,
factorizations
or
sensitivity,
matrices
and
stuff,
but
I
didn't
see
the
problem
with
how
I
guess
yeah.
If
you
have
a
larger
octree
mesh
like
it
kind
of
it,
adds
up
there
was
this
idea
of
putting
it
somewhere
on
a
cloud
or
somewhere
in
the
interwebs
and
downloading
a
file
and
unpacking
it.
F
But
we
can't
put
it
on
ubc's
own
cloud
because
then
only
ubc
employees
would
be
able
to
access
it
and
lindsay
put
it
on.
I
think
a
google
cluster
or
google
cloud
account,
which
meant
I
had
to
sign
up
for
that
and
pay
some
fee,
and
that
was
a
nightmare
to
get
set
up
with
so
where
things
are
living
now
is
maybe
fine
if
nothing
changes,
but
it's
been
a
nightmare
for
development
if
any
of
the
data
files
change,
but.
D
F
F
If
I
change
the
mesh,
if
I
decided
I
wanted
a
different
electrode
spacing
or
I
wanted
to
show
something
else
now
I
have
it's.
It's
been
yeah.
It's
been
frust.
I've
had
to
change
some
of
the
tutorial
scripts
for
what
I'm
developing
currently,
because
if
I
want
to
do
the
forward
problem
and
then
the
inverse
problem,
the
forward
problem
has
been
changed,
which
means
the
data's
changed,
and
if
the
inversion
script
is
now
downloading.
F
F
I
mean
I'm
kind
of
fine
with
anything
for
right
now,
like
I'm
just
keeping
lightweight
files
as
part
of
the
repository
and
when
all
developments
done
and
everyone's
happy,
we
can
find
a
place
to
to
put
it,
but
I'm
just
saying
as
of
right
now,
if
I
make
changes
to
the
tutorials
that
result
in
changes
to
the
data
being
output,
then
I
don't
actually
have
access
right
now
to
go
in
and
change
what's
up
on
on
the
internet.
So
that's
that's
a
problem
for
development.
A
F
F
F
F
C
A
F
A
F
I
got
access
to
it
and
I
tried
to
log
in,
and
I
mean
I
was
busy
and
it
was
just
taking
way
too
much
of
my
my
time
and
I
couldn't
figure
out
how
to
do
it
and
then
I
just
kind
of
gave
up
on
it,
but
it
wasn't
it
wasn't
in
here
I
didn't.
I
got
sent
the
link
to
where
it's
supposed
to
live
and
then
lindsay
gave
me
access
to
it.
F
A
F
One
of
the
things
that
I've
done,
which
is
something
that
actually
needs
to
needed
to
be
fixed,
is
when
I
was
doing
the
forward
simulation
and
basically
creating
the
assigning
noise
and
creating
the
data
that
would
be
used
to
in
the
inversion.
I
didn't
fix
this
seed
for
the
random
noise,
which
means
every
time
you
run
it.
It's
going
to
generate
a
new
data
file,
because
new
noise
is
being
added,
that's
something
that
needs
to
get
fixed.
F
A
F
F
Yeah
yeah,
I
don't
like
I
don't
because
then,
if
you
have
access
to
the
simpeg
repository,
you
would
automatically
have
access
to
everything
under
the
simpeg
umbrella,
which
means
you
would
have
access
to
this
already
it
just.
It
seemed
really
weird
to
store
simpeg
related
things
in
a
third-party
place.
I
F
D
A
F
I'm
not
sure,
okay.
Well,
I
guess.
For
now
I
was
hoping
to
just
finish
the
development
of
these
utilities.
When
everyone's
happy,
then
we'll
then
we'll
package
it
up
in
a
place
where
it
should
live,
and
I
I
can
take
another
stab
at
trying
to
put
it
where
things
are
now,
but
it
was
not
easy.
Maybe
somebody
could
give
me
a
tour.
Maybe
all
it
takes
is
like
a
five
or
ten
minute
show
and
then
I'll
know
what
the
heck
I'm
doing
up
there,
but
it
wasn't
inherently
obvious.
A
D
F
D
A
D
D
F
The
the
locations
would
be
easting,
northing
upwards,
the
inclination
and
declination
would
be
whatever
the
standard
definition
is
gravity
was
done.
Somebody
said
that
they
wanted.
Somebody
switched
the
gravity,
one
to
be
a
convention
where
gravity
anomalies
would
be
high
over
more
dense
structures.
F
A
F
A
A
F
I
I
G
A
G
Yeah
yeah
great
attendance.
I
think
there
were
about
170
people.
Alan
said
at
the
max,
which
was
excellent
and
then
yeah
lots
of
questions
at
the
end.
So
it
ended
up
sort
of
being
a
two-hour
event.
So
I'm
sure
doug
is
pretty,
but
it
was.
It
was
a
great
talk
and
I
think
would
be
a
really
useful
resource.
So
so
it's
good
so
now
he's
set
a
high
bar
for
my
talk
next
week,
which
needs
to
be
started
right
now,
but
that's
okay,
we'll
get
there.
G
Pretty
well
yeah
I
slightly
mixed
up
the
times,
so
thankfully
you
know
I
thought
I
thought
it
started
at
1
30,
but
it
actually
started
at
1
15,
but
they
said
in
the
email,
be
there
15
minutes
early.
So
thankfully
I
was
there
15..
I
thought
I
was
15
minutes
early.
I
was
on
time,
but
late
by
a
minute.
So
thankfully
you
know
they
started
with
another
speaker,
but
all
went
well.
It
was
a
good
session
and
it
was
a
lot
of
folks
from
sort
of
like
big
national
labs.
G
These
so
it's
kind
of
cool
to
see
a
bit
of
a
contrast,
or
that
we
kind
of
provided
a
bit
of
a
contrast
being
sort
of
a
very
a
bit
more
of
a
domain-focused
project.
That's
like
a
little
more
multi-stakeholder,
not
sort
of
the
you
know,
part
of
part
of
the
like
direction
or
scope
of
what
is
done
in
a
national
lab.
So
yeah.
A
G
One
thing
I
was
wondering
so
there's
like
a
few
avenues-
and
I
haven't
really
looked
into
any
of
these,
so
I'll
just
throw
this
out
as
if
anyone
has
time
or
interest
to
look
into
places.
So
I
did
mention
zenodo
and
fixture.
I
don't
know
if
it's
quite
the
right
fit,
because
it's
not
like
a
publication
it.
G
So
I
wasn't
sure
about
that.
The
other
thing
I
was
wondering
is
like
how
hard
it
would
be
to
set
up
a
little
web
forum
where
somebody
can
just
like
request
that
data
be
added
to
our
google
cloud
and
that
it's
a
one
click
button.
I
imagine
somebody's
made
something
like
that
in
the
past
that
we
might
be
able
to
pick
up.
G
Yeah,
that's
sort
of
I
don't
know
if
anyone
else
has
other
ideas
or
opinions
on
on
that,
because
it
would
be
nice,
I
guess
like
we
need
to
have
some
sort
of
a
little
bit
of
control
on
like
a
yes,
no
like
admin
authority.
If
it's
going
to
google
so
that
you
know
we
don't
just
become
somebody's
data
storage
warehouse,
but
it
would
be
nice
if
it
was
a
little
easier
than
it
than
it
currently
is
because
I
I
appreciate
it's
kind
of
it's
kind
of
clunky,
but
google
storage
is
a
good
place.
F
In
the
tutorial,
so
right,
right
now
for
the
tutorials
I've,
just
I've
chosen
to
work
with
synthetic
data
to
keep
the
problems
relatively
small,
to
keep
things
quite
simple
yeah.
But
in
the
the
gif
tools
area
of
things,
I
wrote
some
workflows
to
demonstrate
that
that
package's
ability
and
compared
it
on
on
real
data,
a
real,
a
workflow
from
start
to
finish
on
how
you
would
you
would
tackle
a
real
data
set
and
I'd
like
to
reproduce
those
examples
with
simpeg,
and
we
would
have
a
parallel
comparison
between
both
of
them.