►
From YouTube: SimPEG meeting April 17
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
A
C
D
C
F
C
C
H
C
What
I
want
to
compensate?
We
don't
have
and
like
the
basic
idea
is
actually
like
doing
all
we
want
D,
but
when
you're
inverting
it
that
you
want
to
constrain
that
early.
So
that's
a
basic
idea,
but
there
are
quite
a
bit
of
details
inside
it
so
freaking
system
it
is
kind
of
easier
because
just
it
yeah
you
just
get
beta
and
invert
it.
That's
probably
that's.
Probably
it
processing
part
I'm,
not
sure
how
important
that
is,
but
time-domain
there
are
quite
a
bit
of
details
so
sometimes
that
you
need
to
know.
C
Let's
say
you
got
a
like
a
transmitter
here
and
I
didn't
put
a
receiver,
but
let's
assume
there
is
a
receiver
over
there
and
not
only
just
that
earth
receiver
itself
has
a
response
and
man
that
you
need
to
take
account
sometimes.
So
those
are
the
details
and
the
waveforms
that
can
be
fairly
complicated
and
the
altitude
you
don't
really
know
exactly
altitude.
C
Let's
say
you
got
a
piece
here,
then
you
use
that
like
a
can
radar,
then
you
don't
know
which
one
it's
too
hot,
that
that's
the
details
and
your
sampling
rate
is
much
higher
than
your
measure
of
time
channels.
They
do
an
integration
for
their
time
gates
and
also
at
the
end
of
day,
you
want
to
invert
that
spatially
constrain
the
bridge.
So
those
are
the
kind
of
missing
pieces
in
that
our.
D
F
C
Yeah,
so
those
are
the
details
and
like
so
when
we're
inverting.
Let's
say
it
specifically
for
time
to
mean
data.
We
need
to
take
a
cop
that
sometimes
or
not,
but
it's
very
important,
to
check
whether
that
effect
is
very
important
or
not
because
that,
if
like,
if
you
do,
if
you
did
not
consider
that,
like
guys
like
sky
10,
people
make
Connor.
Oh
you
should
consider
that
and
then
we
did
it
and
they
say
the
research
is
not
not
very
common.
C
I
didn't
mention
that
wave
one
like
that.
We
usually
consider
a
single
purse,
but
it's
actually
multiple
purse
and
if
we
stack
it
so
that
effect
could
be
important
depends
upon
the
case.
So
let's
so
that
was
the
idea.
And
now
yes,
I,
like
I,
wanted
to
consider
all
the
reason,
at
least
in
the
one,
because
that
was
that
most
of
the
development.
C
So
those
are
the
black
one.
Here
is
the
step
off
response,
and
then
you
can
consider
them
receiver
response.
That's
like
a
kind
of
wall,
low-pass
filter.
If
you
can
change
the
frequency
and
as
you
go
lower
frequency,
that's
the
blue,
so
that,
like
a
receiver
response
can
have
an
effect
on,
especially
at
the
LV
time.
So
if
you
know
that,
if
you
can
put
that
you
want,
you
may
want
to
put
that
in
so
at
least
sky
time
assistant
actually
provides
that
low-pass
filter.
So
we
can
put
that
it
put
that
information
in.
A
F
C
K
Because
we
do
tend
to
eat
under
bit,
Hey
we're
always
ready,
we're
always
a
little
bit.
So
it
might
be
that
a
lot
of
em
across
this
different
system.
To
me
it's
just
sort
of
one
of
those
general
thank
you,
let's
beat
em
later,
but
what
I
can't
wear
anything
yeah
bit
further,
then
it's
just
really
hard
if
those
early
times
and
we're
always
bias
and.
E
C
Right,
yeah
references,
I
think
that
kind
of
dip
pants
right,
like
it's
DBT
beta,
so
your
slope
can
be
play.
You
can
have
a
larger
responsive
at
the
only
time
I
can
get
depends
I,
guess
because
either
quit
that
conductor
on
resistor
depends
upon
your
team.
Your
again
remember
that
EBD
curve,
it's
all
like.
C
C
So
that's
that's
an
effect,
so
I
actually
put
that
in
so
the
little
bit
of
detail,
I'm
doing
in
doing
that
in
the
frequency
domain,
so
I
generate
the
low-pass
filter
and
the
freakish.
Let
me
multiply
that
and
then
use
that
to
transform
time.
So
that's
actually
kind
of
annoying
when
you're
solving
PD
Oh
like
a
you're
solving
indirectly
in
time,
yeah
right.
So
we
can
I
think
in
theory.
G
C
C
C
But
you
think
about
that,
like,
when
you're
competing
the
sensitivity
standing
into
convolve
every
sales
right,
yeah,
so
they're
like
a
yeah,
they
there's
nothing
hard,
but
that
there's
some
sort
of
computation
level.
So
that's
that's
a
detail
that
we
need
to
know
so
like
a
in
terms
of
3d.
If
you
can't,
if
you
don't
have
to
do
it,
that's
that's
great!
So
if
you
really
need
to
do
to
think
about
what
is
the
efficient
way
to
actually
cooperate
that
that's
a
question
I'm,
not
sure
what
is
the
real
solution?
C
So
first
cast
is
actually
just
step
off
and
ramp
off.
So
this
is
a
simple
exponential
function.
The
blue
one
is
just
true
exponential
function
and
that
orange
one,
if
I,
put
a
ramp
ramp
off
so
it
actually
pretty
different
and
for
real
Ian's
response.
That's
usually
different,
especially
at
the
only
time
so
putting
the
right
wave
form
is
important.
That's
actually
we're
doing
so.
That's
fine
and
the
other
one
is
the.
Let's
say
this
one.
C
E
C
Purse
you
have
two
responses
and
for
a
response
at
the
first
pulse
kind
of
made,
I
made
an
effect
at
the
second
purse,
so
you
need
to
take
a
combat
stacking
I.
Think
John.
You
may
I
think
you
know
what
what's
what
that
means
right,
like
a
VIP
response,
for
instance
right,
you
know
first
purse
and
then
you
have
an
off
time
all
right,
but
that
off
time
response
may
effect
on
the
second
person
responds
right
well,.
L
F
C
C
F
F
C
C
C
E
C
K
C
So,
okay,
I
think
that's
a
good
point
and
then
I
think
like
a
for
instance,
that
let's
say
this
is
a
simple
exponential
curve
and
then
I
was
using
a
point
o2
for
tau
time
constant.
But
if,
if
I
increase
that
power
to
0.01
we're
double
purses
like
much
less
that
so
I
think
that
the
the
blue
one
is
the
double
purse
and
the
black
one
is
a
single
course.
C
So
it
depending
upon
your
how
your
time,
decay,
that's
that's
different
right.
If
your
topic
fair,
like
a
response,
DT
slowly,
that
always
felt
that
always
make
an
impact,
so
I
think
it's
it's
kind
of
like
it.
A
comment
is
kind
of
like
bit:
tough
or
that
really
depends
on
your
system
parameters
and
also
I
can't.
E
Conic
you
be
okay,
so
in
that
sense
it
were
suspect.
More
pulses
gives
that
FX
gonna
it's
gonna,
be
fully
accumulative,
but
it's
gonna
keep
adding
up
and
adding
out
bright.
Theory.
Look,
let's
see
you
add,
like
ant
superconductive
over
your
response
really
drags
a
lot
of
time.
As
you
keep
stacking
more
and
more
thinking
cleats,
you
can
and
start
tapering
all
your
are
you
signal
yeah.
C
C
C
C
Secondary
or
a
product
like
a
total
field,
there
are
X
type
source
type.
Your
offset
is
7
to
8
meter.
You
need
to
pick
the
depth
and
that's
the
spatial,
discretization
and
I
need
to
put
the
frequency
in
to
your
survey,
and
here
I
want
to
invert
both
conductivity
and
the
height,
so
I'm
using
the
wire
map.
It's
like
a
it's,
like
kind
of
like
a
generating
projection
matrix
for
a
model
so
I.
C
My
mother
is
not
only
a
conductivity
and
also
the
height
of
my
transmitter
and
receiver
to
hear
the
transmitter
receiver
at
the
same
height,
so
I'm
just
inverted
for
one
but
depending
upon
the
case
like
okay,
we've
got
a
fixed
screen
system
where
you
got
a
transmitter
here
and
the
receiver
there,
and
both
of
them
could
be
wrong.
So
you
need
to
invert.
C
D
C
C
C
C
C
C
D
K
E
K
E
G
F
E
K
Not
it's
not
new,
but
yeah,
but
I,
guess
in
practice.
There's
a
question
of
whether
you
know
it
wouldn't
matter
very
much
if
your
source,
which
is
way
about,
is
it's
chasing
a
meter
or
to
compare,
because
that's
just
them
changing
how
you're
influencing
right
maintain
where
your
receiver
is
yeah.
That's
a.
D
F
C
My
analytic
sensitivity
actually
matches
with
the
finer
and
difference
acetate,
so
it's
like
for
this
one
still
doing
a
fine
difference
is
cheap,
but
analytic
is
cheaper,
so
Ana
legacy
is
one
solve
of
predict
like
a
core
problem
yeah,
but
that
final
difference
at
least
to
solve.
Yes,
so
if
you
could.
C
That's
the
data,
so
with
this
five
meter
we
got
pretty
high,
pretty
big
different
and
that
actually
depends
upon
the
conductivity
of
the
earth.
So
if
the
earth
is
very
conductive,
that
difference
is
bigger
so
that
you
may
have
a
better
chance
to
actually
kind
of
resupply
can
recover
that
to
height
I.
Think
that
my
conductivity
was
yeah
point.
C
A
C
But
that
regularization
it's
just
one
parameter,
so
it's
not
sure
like
a
gradient
trim.
E
K
C
E
E
C
What
I
thought,
but
it
seems
like
I,
don't
really
need
to
when
you
actually
got
this
attitude
so
frequency
domain
there
is
actually
working
pretty
well
because
if
you
think
about,
we
got
pretty
high
frequency
about
hundred
K
for
the
highest
frequency.
So
that
was
good,
but
for
time
there
may
yeah
I,
like
I.
Couldn't
I
couldn't
just
do:
okay,
inverted
them
together.
So
we
may
need
to
wait
so
I
think
it's
less
sensitive.
So
when
you
actually
got
the
sensitivity,
then
I
think
putting
them
in
and
herded
together
without
even
waiting,
I.
C
H
C
We
can
put
some
bounds
constraint
here.
I
didn't
put
ok,
but
yeah
I
was
I.
Think
so
here
I.
Actually
this
example
I
started
from
50
and
then
the
recovery
was
30,
so
Sims,
like
ok
but
dependent
I.
Think
if
you
don't
have
like
a
really
good
sensitivity
and
it
can
fluctuate
a
lot.
You
know.
In
that
case
we
may
want
to
put
some
sort
of
like
and
okay
Oh.
C
C
I
think
so
that
alpha
s,
one
means
it
meaning.
I
was
I
had
a
reference
model
that
I've
got
something
from
something:
okay,
so
I
think
that
was
about
it.
So
that's!
That's
all
again.
That's
considering
all
these
like
a
different
facts,
time
gate
there
I
haven't
right,
but
there
are
some
ways
that
you
can
actually
kind
of
just
correct.
That
I
think
I
haven't
tried
that
yet,
okay
and
just
show
you
one
more
example.
C
C
C
Kind
of
very
similar
to
what
our
who's
guys
does
so
like
they
have
at
each
sounding
has
the
same
number
of
letters
and
that
or
each
layer
we're
putting
some
sort
of
like
a
smooth
constraint,
maybe
a
virgin,
but
the
here
we're
putting
all
of
them
together.
At
least
so.
We
got
one
objective
function,
so
we're
minimizing
that
want
to
check
your
function,
it's
just
so
nice,
and
then
we
were
asking
in
like
for
each
node.
We
want
to
have
some
sort
of
smooth
model.
Okay,.
D
H
D
F
C
D
D
C
C
So
this
one
I'm
not
actually
putting
any
as
like
a
spatial
constraint
like
horizontal
constraint,
I'm
just
asking
for
a
smooth
modeling
for
the
clear
action.
Okay,
so
yeah
like
what's
okay
but
now,
let's
say:
okay
but
I
still
I
want
to
have
a
smooth,
smooth
solution
in
the
horizontal
direction
and
that's
it
looks
like
so
they
go
back
and
forth.
That's.
A
K
C
C
C
C
C
G
G
C
G
C
C
C
A
right
reference
model
that
gives
you
a
pretty
nice
like
a
sharp
boundary
and
then
below
that
you
may
not
be
very
sensitivities,
so
I
think
that's
good
to
have
least
like
I.
Sharing.
Two
images
of
smooth
than
alzira
in
virgin
gives
you
a
pretty
good
idea.
At
least
I
got
like
a
lower
end
like
a
top
and
bottom
Audrey's
a
like
this
conductor.
So
that's
a
I
thought.
E
E
A
C
A
C
D
C
C
F
C
But
what's
good
thing,
they
usually
usually
kind
of
effect,
classify
that
as
a
different
line,
so
in
a
single
line
index
I
think
it's
always
increasing.
So
that's
that's
a
good,
that's
a
good!
So
as
long
as
we
got
that
information
we
can
figure
out
that
which
stretched
it
was
so
I.
Actually
I
can
take
a
particular
count
and
assign
the
corresponding
receive
allocation.
But
one
thing.
C
E
C
E
C
F
C
C
Can
be
faster
and
yeah,
you
can
make
this
type
of
plot,
so
I'll
make
that
as
I/o,
so
that
that
once
you've
done
that
you
can
actually
plot
at
least
like
time
for
you
at
the
section
view,
so
that
you
can
so
those
are
the
resistor
right
like
I.
Would
those
guy
I
figured
that's
what
they're
looking
for
and
they
got
the
ground
a
chromium
data
DCP?
That's
actually.
C
A
C
C
A
A
C
I
C
C
C
This
is
my
like
a
1d
model,
then
I
put
just
an
out
in
here,
and
all
of
them
are
G.
Okay,
so
I
inverted
together.
The
look
at
here
got
pretty
high
resolution
really
resolving
of
the
resistor
pretty
well,
and
it
goes
that
kind
of
get
it
gets
worse,
but
I
think
still
helps
by
playing
some
sort
of
smoothness
constraint.
C
A
A
One
thing
I
would
appreciate
thoughts
on
is
I've,
been
working
on
cleaning
up
notebooks
to
go
with
casing
paper
over
and
so
I'm
trying
to
like
trying
to
sketch
out
how
we
actually
want
to
distribute
things
like
this
with
the
publication,
and
so
this
is
like
my
sketching
overflow,
but
I'd,
be
very
curious
to
hear
what
you
guys
think
so.
A
I've
got
tests
running
and
ones
failing
right
now,
because
what
I've
found
is
that
some
of
the
makeup
looks
at
you
long
to
run
and
so
I'm
using
the
test,
I
pie,
envy
and
within
that
you
can
actually
list
specific
ones
that
you
want
to
ignore.
So
here
I
know
that
these
ones
are
just
too
baked
to
run
so
I'll
just
ignore
those
and
then
what
I
was
doing
is
also
then,
because
I
just
had
too
many
that
are
like
slightly
too
long.
I
was
randomly
picking
like
three.
A
G
A
The
colony
yeah,
so
there's
sort
of
there's
two
dials
in
there
about
timing.
There's
one!
That's
the
execute
this
guy
at
the
time
of
so
this
will
run
the
notebook,
so
it
will
timeout
after
that's
in
seconds
so
it'll
timeout
after
that,
so
every
single
notebook
and
then
there's
the
Travis
timeout,
so
Travis
kicks
you
off
after
15
or
so
55
minutes.
So
if
you're
sort
of
like
integrated
number
of
tests
state
running
it's
too
long
on
a
single
note,
so
those
are
two
things
to
keep
in
mind.
A
A
So
that's
testing
and
then
binder
is
up
and
running
and
been
working
pretty.
Well,
though,
the
way
that
I've
found
best
to
interact
with
it
is
to
create
an
environment
that
it's
really
easy.
It's
basically
the
same
as
your
requirements,
but
you
specify
like.
What's
coming
in
through
Conda
and
then
what
your
Conda
dependencies
are
and
then
what
your
hip
dependencies
are.
So
I
found
this
to
be
a
little
more
stable
than
using
a
requirements
file
specifically
because
of
better
solvers,
and
we
want
mkl
install
dated
Conda.
A
A
A
A
A
I
A
The
last
thing
that
I
was
looking
at
doing
is
using
as
a
note
of
to
archive
it
and
give
it
be
a
lie
if
two
options
are
like
two
major
options.
So
there's
a
note,
oh
and
then
there's
fake,
share.
What
I
like
about
snow
dough
is
that
every
time
you
do
get
help
release
it
just
assigns
a
new
do
I.
Do
that.
So
actually,
if
you
look
at
its
in
page,
you
can
see
that.
A
Its
own
image,
you
will
be
alive
for
that
full
request.
Oh
and
then,
if
you
want
to
cite
like
the
whole
thing,
it
also
gives
you
a
DOI
for
version,
so
you
can
look
here
and
see
that
this
is
where
they
they
tell
you.
So
this
is
what
you
cited
us
and
then,
if
you
want
to
cite
all
of
the
versions
you
just
replace
the
DOI
here
with
this
deal.
Oh
that's
cool!
So
that's
kind
of
nice
because
then
everything
like
what's
important
about
getting
into
UI
is
that
it's
archived.
A
A
They're
both
suitable
for
different
things
like
I,
like
fake
share
for
poachers
and
those
kinds
of
things.
What
I
didn't
like
about
figshare
necessarily
was
they
restrict
what
kind
of
lately
they
rely
since
it
when
you
bring
it
in
which
is
strange,
and
then
here
like
to
actually
just
see
like
all
that
I
hope
to
do
to
turn
this
on
is
like
it's
the
on
button
and
that
you
are
released
and
then
it's
just
done,
and
then
it
monitors
so
like
now.