►
From YouTube: SimPEG meeting Jan 24
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).
B
A
B
C
A
A
B
A
Guess
is
because
Ron
started
to
freak
out,
because
there
are
so
many
things
on
that
pull
request
its
getting
bigger
and
bigger
and
as
I
Cali
yeah,
it's
kind
of
a
PR,
that's
getting
out
of
control,
but
there's
cool
stuff.
So
most
of
it
you
guys
I
received
already
like
I
presented
it
so
many
times,
but
it's
the
MDI
Cartesian
amplitude
inversion
that
one
is
a
nonlinear.
So
now
we're
getting
in
the
nonlinear
part
of
the
inversions,
which
is
I.
A
G
A
Minggu
worry:
okay,
I'm,
not
enough!
I'm
mel
out
to
talk
about
this,
but
this
is
the
dataset
and
when
you
showed
it
to
me
this
is
the
TMI
and
then
obviously
there
is
like
super
intern
program
announced.
You
know
this
is
close
to
Mount
Baker,
so
it's
almost
vertical
field
and
and
then
this
is
amplitude
the
amplitude
of
the
data.
A
So
we
want
like
when
we
saw
that
looks
like
I
care
for
sure
we'll
need
the
code
so
started.
Writing
the
code
for
her
job.
So
now
we
have
the
env
I
and
the
Magnetic
inversion,
I'm
in
for
Mira
and
you
get
the
gravity
gradiometer
and
then,
since
soggy
already
had
written
the
other
PD
mag
added
the
gravity
in
pd
and
then
the
nice
thing
is
that
we
get
the
grade
geometry
of
furry
friends.
H
I
A
J
H
A
H
A
H
A
A
H
What
was
suspicious
about
like
ours,
all
this,
because
I
mean,
is
a
little
different
like
anyway
you're
generating
this
is
TV
and
then
doing
jking.
Keep
that
like.
We
do
similar
way
like
this
in
exact
how
sweet
imagine
things,
but
the
PE
you're
not
generating
it's
still
again
and
I.
Think
like
there's
a
little
bit
of
difference,
I,
regenerate
or
not.
A
H
The
process
are
you
using?
Actually
the
flight
actus
is
a
TV
or
not
so
one
of
the
other
way
we're
actually
pretty
poor
accuracy,
60,
b
function
and
the
other
way
the
other
one
actually
write
a
Christmas
didn't
function,
so
is
that
it's
a
different
IR
IRA
but
like
I,
was
a
busa
species
about
that
perfect.
A
A
And
then
I
just
wrote
the
examples
and
tests
for
to
increase
the
coverage,
be
able
to
pass
coverage
and
then
yesterday
was
more
on
the
spherical
formulation,
which
is
super
highly
nonlinear
and
then
ended
up
now
to
the
point
where
we
need
a
compostable
objective
functions,
because
everything
needs
to
scale.
You
know
differently
depending
of
angles
or
or
amplitude,
and
then
the
test.
H
A
A
H
E
H
A
A
H
A
So
for
my
in
my
problem,
I
just
tested
it
that
just
like
kept
Kevin
cruising
babblings
I'm
delve
into
my
my
fields.
We
were
changing
to
the
decimal
places
that
I
didn't
care
about,
and
then
that's
that's
good
enough.
Then
I
would
just
apply
that
same
match
everywhere.
He
would
one
match.
I
was
just
like
we
recycling,
yeah
yeah.
K
A
A
A
Elsewhere,
auntie,
or
maybe
just
run
the
example
via
so
now,
I
know
I'm
sure,
right
coverage.
It
looks
at
examples
too,
not
just
the
test
with
the
examples
too
yeah,
yeah
and
yeah
since
most
of
them
it's
like
inversion
related
and
that
I'm
covering
with
examples,
because
another
man
touching
and
touching
everything
it's.
B
A
L
Am
I
on
the
right
binge?
Oh,
you
might
be
on
the
objective
functional
stuff
because
we
switch
to
that.
Oh.
A
K
J
K
A
K
A
Yeah,
I
was
just
saying
for
the
fall
for
the
nonlinear
problems.
The
important
part
is
the:
is
the
directive
to
update
between
each
each
iteration
for
like
I,
think
eventually,
I
want
to
make
examples,
and
we
start
talking
about
it
as
a
group,
because,
at
least
with
a
problem
on
your
bottom
edge,
you
definitely
need
to
re
weight
with
sensitivity.
Every
time
you
change
your
model
functions
are
changing
as
we
go
for
the
amplitude
inversion.
It
depends
on
the
on
the
data.
A
So
the
approximate
sensitivity,
sorry
here,
but
the
amplitude
problem
it
changes
and
as
function
of
the
of
the
data
for
the
MV
is
vertical.
It
changes
the
function
of
the
model,
but
in
either
case
J
transpose
J
is,
is
iteration
depend
them,
and
if
we
don't
do
that,
I'll
create
an
example
to
show
you
what
happens
and
the
in
both
cases.
But
if
you
don't
update
that,
often
enough,
you
end
up
with
the
in
a
weird
space
and.
A
B
A
Oh
yeah,
what
sensitivities
it's
taking
yeah
the
entire
gradations,
taking
all
the
way,
it's
okay,
there's
a
special
case
for
the
for
the
spherical
model.
Parameters
are
the
same
right.
Two
of
them
are
angles.
One
of
them
is
is
isn't
as
a
model
model
value
ring,
so
you
can
treat
them
the
same
way,
so
you
need
also
to
compensate,
but
that's
that's
another
level
of
it.
Bexley.
A
A
A
It
asked
to
create
that
negative
lobe
needs
to
put
stuff
from
on
either
side
or
when
we
run
that
example,
it
created
like
stuff
like
on
the
wrong
places
and
ice
is
that
you're,
a
dyke
that
runs
right
right
in
the
middle
here.
So
that's
why
I
amplitude
is
is
important
and
then
finally,
what
managed
to
work
last
night
is
the
is
a
spherical
mbi
in
spherical.
I
guess
I
can
show
you
also
example
for
cartesian
is
again
the
same
block.
A
E
A
A
F
A
Still
0
because
in
this
case,
you're
going
for
the
smallest
smallest
model,
but
on
XY
and
z,
independently
yeah,
so
on,
whichever
is
easiest
to
create
the
model
is
going
to
go,
who's
gonna
go
towards
towards
there
and
in
this
case
it's
just
easier
to
create
two
anomalies.
Because
then
you
have
kind
of
like
a
double
lobe
yeah.
B
A
A
A
A
K
A
A
A
Guess
I
changed
my
colors:
it's
not
it's
not
nice,
but
this
is
the
the
Cartesian
formulation.
Yeah
like
a
double
double
loop
here,
but
then,
when
you
go
to
circle,
is
just
a
simpler
space
and
now
we're
starting
to
get
stuff
that
just
much
more
like
a
mag
mag
inversion,
barcoding
come
come,
come
back
nests
on
them.
That's
the
that's
Daniel
I'm,
not
there!
Yet
I
have
a
lot
of
problems
with
the
the
line
search.
It
keeps
breaking
on
me.
Yeah.
H
H
A
B
B
I
B
B
H
H
B
A
B
J
H
B
I
B
A
B
B
F
A
Okay,
whatever
one
thinks
for
some
reason,
this
guy
has
the
you
remember
you
like
I
me:
I
needed
them
using
it,
but
it's
India
in
the
maps
this
guy
mesh
mesh
tofu,
oh
I,.
I
I
B
Yeah,
so
what
it
does,
because
most
maps
are
very
simple
and
that
all
you
need
to
instantiate
them
is
a
mesh,
and
so,
in
this
case
anything
that's
not
in
via
this
list.
It's
going
to
try
to
create
the
map
and
run
a
test
on
it
in
this
case,
because
the
match
too
much
top
oh
I,
think
you
need
to
give
it
a
couple.
Other
parameters
right,
it'll,
fail.
You.
A
H
A
I,
your
your
top
ahmadi
to
pull
air
is
also
very
and
was
very
important.
This
was
a
nice
nice
contribution,
your
surface
to
end
the
just
to
get
that
layer,
because
that
then,
for
the
equivalent
source
that
was
just
a
mapping.
I
was
just
calling
your
function
during.
That's
just
sorry,
I
might
arrive
once
this.
A
A
A
H
H
B
A
K
F
B
Yeah,
it
should
be
just
as
simple
as
changing
the
regularization
in
theory,
good
someone.
If
we've
done
everything
right,
it
should
be
that
simple,
that's
good!
So.
B
So,
in
this
case,
all
I
would
do
for
that
problem
is
construct.
We
have
a
single
Phi
D
and
then
my
regularization
just
consists
of
a
Phi
Sigma
and
a
fine.
You
and
I'll
just
add
them
together
and
they
could
pose
for
you
Tegan
offs.
They
can
be
whatever
and
just
weight
them
relative
to
one
another
yeah.
So
actually
it
should
be
pretty
straightforward
with
Lee
yeah.
B
Okay,
yeah
Emily
I'll
get
that
fixed
today,
even.
A
A
B
And
then
the
other
thing
hopefully
coming
in
today
or
tomorrow,
I've
now
got
the
time
domain,
evh
j,
all
running,
not
HJ,.
I
B
So
and
then
there's
a
couple
things
that
just
like
I,
don't
have
they're,
not
all
cross
compatible.
Yet
like
you
can't
get
age
from
eb
yet,
but
that
we
can.
G
B
B
B
Know
it
doesn't
just
naturally
actually
eliminate
because
the
thing
that
you
eliminate
to
get
eat
formulations-
you
eliminate
DV
DT,
so
you
can
naturally
get
that
out,
but
be
you
can
either
discrete
like
you
can
integrate,
which
that
was
what
I
was
doing,
but
you're
not
going
to
get
the
exact
same
results
is
doing
the
other
formulation
like
it
is
better
just
just
if
you
want
to
be.
You
should
just
solve
for
me
right.
So
that's
how
that's
implemented
right
now,
just
to
keep
it
simple.
H
B
D
B
H
B
H
H
B
H
B
Know
that's
the
something
we
should.
We
can
look
at
like
this.
First
iteration
won't
have
that,
but
we
can
try
to
pull
that
in
later
on
and
then
Rowan
I
just
mentioned
that
ok,
so
the
discretized
package
that
is
now
under
the
sim
peg
organization.
Eventually
that's
going
to
be
used
to
pull
out
the
mesh
class
so
that
code
will
get
removed.
So.
B
There's
a
package
called
discretize
that
basically
is
synthetic
mesh.
Ok,
so
we'll
just
use
that
that'll
simplify
testing
and
things
like
that.
Oh.
B
A
good
question,
I
think
it
depends,
depends
on
what
you
want
to
do.
You
I
mean
I.
Think
you
tells
just
needs
an
overhaul
so
right
now
what
I
would
guess.
The
first
thing
that
will
do
is
basically
just
leave
them
in
simple
for
now
and
then,
when
we
see
things
that
are
actually
sort
of
useful
just
on
the
mesh
will
just
pull
it
over.
So.
H
E
H
A
H
What
you
did
something
to
me:
Sam,
oh
yeah,
that's
what
no
yeah
he
was
asking
where
it
could
be
youtube
youtube,
video
or
some
of
that
yeah
I.
Wasn't
sure
I'm
not
like
that,
but
you
need
to
be
good
stuff,
yeah
yeah.
It
was
actually
like
that
guess
we
got
the
tutorial
that
page
still
get
our
Yes
Mother
star
ya.
J
A
J
A
G
B
H
I
H
I
H
I
H
B
H
B
The
pic
and
stole
the
pivotal
will
be
fixed
with
dis
critize,
because
a
lot
of
the
stuff
that
was
causing
issues
with
Windows
was
this
site
on
code,
and
actually
all
that
like
we
want
simpang
will
be
purely
Python.
Once
we
pulled
the
mesh
out
and
then
Joe
fixed
how
to
install
disc
or
ties
on
Windows,
he
was
able
to
get
that
like
cleanly
working,
so
that
actually
should
be
that'll.
Hopefully
remove
that
overhead
completely.
So
people
will
be
able
to
like
pip,
install
SCIM
peg
and
that
will
reliably
working
does.
B
B
B
One
of
the
other
things
is,
is
I'm
going
to
try
and
get
a
couple
examples
deployed
on
the
Microsoft
Azure
system
and
that's
great
because
then
you
can
actually
just
we
have
control
over
the
install.
So
then
people
can
just
start
using
it
go
because
then
you
can
play
with
it
and
realize
that,
like
this
is
actually
useful
and
then
fight
with
intel
on
your
own.
D
G
B
B
H
B
B
G
B
Bruins
comment
and
like
Jupiter,
does
this
and
a
few
packages
to
do
this.
They
probably
gave
a
Google
Group.
That
just
is
going
that's
one
thing
we
could
try
yeah.