►
From YouTube: SimPEG Meeting, January 11th, 2023
Description
Recording of the weekly meeting from January 11th.
Some new faces, and discussions on upcoming TEM developments for the codebase.
D
C
B
H
Yeah
and
then
Devin
I'm,
a
research
associate
I'll,
probably
see
you
in
person
tomorrow.
I
think
we
might
have
actually
met
once
before
and
I'm
also
doing
some
work
relevant
to
valet,
because
they've
got
these
plate-like
structures
and
they're
Imaging
things
with
utem
so
playing
around
with
some
sensitivity
waiting
and
some
things
that
we
can
maybe
model
plate
light
structures.
But
a
lot
of
what
I
do
is
I
sort
of
manage
the
proprietary
software
that
we
develop
and
and
give
to
our
sponsors.
A
A
B
A
Thought
anyway,
so
kind
of
go
around.
Can
I
just
appreciate
to
see
what
everyone's
been
up
to
as
far
as
impact
related
stuff.
So
my
side
had
a
book
report
here.
I
have
been
taken
care,
a
lot
like
the
new
year
code,
cleanup
and
stuff,
and
going
through
that
I'm
excited
kind
of
catch
up
on
all
these
Piers
I've
been
going
in
things
little
by
little,
and
we
updating
a
few
versions
of
things.
A
The
new
version
of
black
should
be
released,
hopefully
shortly
by
it's
going
to
reformat
a
lot
of
code
just
be
aware:
it'll
be
switched
up
and
open
something
there
to
be
updated.
This
morning,
I
was
working
on
a
random
little
issue.
It
seemed
like
there
was
an
issue
installing
mom
with
the
continuous
integration
stuff
that
was
trying
to
fix.
A
A
A
A
A
I
want
to
discipline
will
be
really
quick
and
small.
The
second
one
will
be
constant.
Peg
is
a
little
bit
more
thoughts
and
Don
might
appreciate
you
taking
a
look
at
it.
It's
related
to
so
it's
how
it
does
the
regularization
on
the
tetrahedral
and
triangular
measure.
It's
also
going
to
support
it.
At
the
same
time,
the
directionally
dependent
it's
like
in
the
antipsychoprotic
kind
of
rotated
gradients
things,
because
you
can
I'm
going
to
use
the
same
thing
to
do
that
that
I
would
do
for
the
tetrahedral
stuff,
because
it's
it.
B
A
A
Think
it's
pretty
straightforward,
but
it
just
some
thoughts
on
it
now.
That's
it!
That's
a
separate
thing
that
we
were
talking
about
that
is
related
to
the
vector
correctorization.
So
so
this
is
a
a
rotating
gradient
thing.
So
if
you
wanted
to
smooth
like
you
know
at
a
45
degree
angle,
as
opposed
to
like
the
uniform
in
a
circle.
B
A
A
F
A
B
B
B
D
And
yeah,
it's
been
working
really
good
as
long
as
you
haven't
got
to
make
a
domain
big
enough,
I'm
sure
that
this
kind
of
conditions
at
satisfied
and
yeah
I
made
a
function
just
like
the
DC
to
drape
everything
onto
the
mesh
because
yeah
it's
a
slightly
off
conversations.
It
just
yeah
completely
kills
the
inversion,
no
fitting.
So
it's
called
the
3D
conversion.
Yeah
3D
version
yeah.
So
this
is
all
3D
yeah
I
have
I
started
doing
a
little
bit
of
PGI
playing
with
the
3D,
but
that's
neither.
D
But
yeah
I
also
started
trying
to
blend
the
em
or
sorry
the
frequency
domain
and
the
natural
source
that
stuff
with
how
like
we
were
talking
about
before.
Just
like
you
know,
there's
a
get
J
function
in
normalcy.
We
should
be
trying
to
kind
of
follow
the
same
rules.
So
I
kind
of
I
got
that
stuff.
A
little
bit
cleaned
up.
I
haven't
touched
the
potential
field
side
yet,
but
yeah
just
a
little
few
hours
here
and
there
and
yeah
it
didn't
get
too
much
into
it.
F
B
D
B
D
B
H
B
C
C
Yeah,
but
every
time
you
move
a
receiver
like
the
worst
case
scenario,
is
going
to
be
moving
it
by
a
full
DX,
I,
guess:
half
a
DX
right.
Actually,
you
could
be
moving
it
from
a
full
DX
if
you're,
if
you're
bad
lucky,
in
which
case
that
means
that
you're
kind
of
Shifting,
your
your
your
poles
by
your
full
DX
and
that's
going
to
have
a
huge
impact
on
your
on
your
simulation,
so
you're,
better
off.
Just
thinking
not
moving
your
receivers
receivers
stay
where
they
are
relative
to
each
other.
H
H
C
D
D
And
one
other
thing:
I
do
have
to
go
a
little
bit
early
here
at
the
dark
class.
But
back
when
we
were
talking
before
Christmas
yeah
I
I,
wouldn't
mind,
I
could
take
alternative
to
the
newsletter
awesome
yeah.
You
guys
said
you
guys
had
some
contacts
for
your
talks.
Yeah.
If
you
guys
can
give
me
that
I
can
take
a
look.
I,
don't
know
if
we
want
to
like
we'll
release
something
this
month.
Maybe
talk
about
your
paper
or
your
presentation
and
stuff
and.
B
F
B
F
A
B
D
B
C
A
C
C
So
it's
going
to
be
a
bit
faster,
I
think
I
can
reduce
the
runtime
still
on
the
J
on
the
J
matrix
by
kind
of
combining
sources
together,
if
possible
and
yeah
otherwise
help
judge
how
to
write
a
paper.
You
submitted
this
paper
to
prospecting
geophysical
prospecting
or
something
about
the
the
L1
invariant.
You
know
the
TV
versus
hi
John,
the
whole
difference
between
TV
and
and
L1,
and
one
normal
gradient.
So
it's
going
to
be
in
review.
C
Yeah
I
only
look
at
the
JT
effect,
but
we
basically
had
like
a
full
time
Loop
and
then
doing
storing
all
the
derivatives
of
the
fields,
and
then
we
were
redoing
another
time
loop
after
backward,
but
really
we
only
needed
one
backward
Loop,
because
we
only
need
to
store
two
derivatives
at
a
time
and
just
step
backward
like
this
yeah.
You
know.
C
C
Oh
yeah
yeah,
so
it's
just
a
it's
a
dipping
plate
and
then
we
can
recover
it.
This
is
like
the
Canada
Sky
10
time
range,
so
from
0.1
to
10,
10,
milliseconds
and
then
yeah
convergence
looks
pretty
good.
Basically,
the
kick
here
is
that
if
you
put
a
reference
model,
you
get
the
ring.
If
you
don't
use
a
reference
model,
then
you
can.
You
can
converge
quite
nicely
like
I'm,
taking
two
gas
and
two
gas
and
steps
here,
but
you
know
five
is
coming
down
coming
down
quite
nicely.
F
C
Here
is
just
smoothing
and
then
with
just
the
smoothing
we
can,
we
can
fit
it
quite
nicely
already.
So
it's
a
simple,
it's
a
simple
fix,
but
that
means
that
we
can't
do
sparsity
because
obviously
alpha
s
is
zero.
So
that's
kind
of
like
where
I'm
kind
of
playing
trying
to
play
tricks
can
we
bring
it
back
later
in
life
or
yeah
I'm,
not
sure.
Yet
what.
C
H
C
B
C
C
Yeah,
that's
the
other
trick.
It
seems
like
you
can
only
want
to
go
slowly.
If
you
go
to
try
to
take
two
big
steps,
then
you
can
get
some
to
pretty
nasty
nasty
space
right,
so
it
takes
longer
to
iterate
down
the
down
the
target,
but
it's
kind
of
a
more
robust
way
to
to
do
it.
Just
take
multiple
gaussians
and
just
slowly,
you
know,
start
high
data
and
then
slowly
go
down
yeah.
This
might
be
a
way
to
accelerate
it,
but.
D
A
A
C
C
Timing
for
it,
how
long
did
it
take
to
run
right
now?
It's
a
pretty
small
problem,
it's
about
30
seconds
per
beta,
so
it
takes
about
like
five
or
six
minutes
for
this
problem,
but
that's
only
like
a
four
by
four
grid,
so
16
sources
and
then
a
Time
channels
right.
So
it's
a
it's
a
pretty
small
problem,
but
it's
it's
big
enough
to
be
able
to
test
up
at
least
be
able
to
see.
What's
what's
up.
H
I
would
like
maybe
the
three
of
us
to
get
together
at
some
point
and
get
an
explicit
list
of
tasks
for
this
I
kind
of
find
myself
going
through
and
making
improvements
for
what
I
need
to
do
my
work
and
then
finding
out
that
maybe
there's
something
additional
that
we
want
to
change
to
that
because
it
fits
into
something
bigger
picture
and
I'm
yeah
I'm
I'm
kind
of
finding.
Maybe
we
should
get
together,
get
a
global
perspective
of
what
needs
to
be
done,
make
explicit
tasks
and
then
partition
it
out.
A
Yeah,
that's
something
I
want
to
make
sure
that
we're
all
on
the
same
page
in
is
the
attraction
that
we
see
the
time
medium
stuff,
going
like
all
the
little
pieces
that
were
so.
If
I
think
I'd
like
to
see
Don
what
you
I'm,
not
I
know
you
shared
it
at
one
point
here:
ER
or
your
branch
that
you
have
the
like
your
Jacobian
operation
going
on
there.
So.
A
H
H
H
D
B
H
H
Yeah,
so
that's
pretty
much,
it
I
mean
I'm
yeah
I'm
running
my
toy
example,
unlike
Dom
I
can't
really
make
as
good
a
use
of
tiling
because
it's
large
loop,
but
it
seems
to
be
going
yeah,
not
not
so
bad.
So
I've
got
my
my
dipping
plate
that
I'm
trying
to
recover
with
utem
and
sort
of
recreate
the
example
that
I
had
with
the
still
mesh
code,
because
it's
just
more
impactful,
if
it's
in
3D
and
then
tomorrow,
I
guess
I'm
hoping.
H
B
G
I
know
Devon
you've
been
looking
on
in
place
like
cannot
help
brilliant
the
inversion,
but
if
you're
working
on
the
analytics
and
the
plate
modeling
happy
to
have
some
collaboration
on
that
or
like
further
discussion
at
some
point,
we'll
let
you
know
as
well
what
I
do
on
that
front
and
it's
not
really
an
update,
but
I
will
change.
The
chat.
I
was
an
issue
that
we
discussed
several
times
before
the
holiday,
so
I
just
want
to
re-put
it.
To
put
it
back
up
front
about
the
B
field.
G
G
F
G
G
G
G
Yeah
well,
I,
don't
remember
if
we
I
don't
remember,
for
that
example
that
issue
I
put
the
I
put
the
my
notebook
in
the
issue
so
like
we
can
reopen
it
I.
Just
don't
remember
if
it
was
a
circular,
Loop
or
or
a
line
current
but
yeah.
Well,
it's
a
it's
a
ground.
It's
it's!
A
it's!
A
ground!
Loop
like
I'm,
doing
a
lot
of
modeling
of
squid
systems
like
so
ground
super
Sienna.