►
From YouTube: SimPEG Meeting Feb 20
Description
Walk through of the FDEM upgrades as we move from the problem + survey to a simulation class
A
Okay,
yes,
so
the
things
that
I
wanted
to
chat
over
today
is
maybe
just
go
through
a
bit
of
where
the
simulation
class
is
I,
implemented.
The
frequency
domain
code
or
a
new
table
stimulation
and
I
was
thinking.
One
of
the
things
that
we
can
do
is
just
like
download
or
notebook
and
like
walk
through
what
kind
of
user.
B
E
E
A
Refactoring
is
always
a
good
place.
I've
been
doing
that
a
little
in
the
medium
code,
I'm
trying
to
like
be
careful
with
that,
because
I
still
want
to
it,
not
too
painful,
but
there's
a
lot
of
naming
things
that
I
look
at
and
it
just
drives
me
crazy,
but
I'm
gonna
leave
most
of
them
for
now,
okay,
so
all
download.
A
What
we'll
do
once
it's
sorted,
stabilize
and
I
put
all
of
the
examples
on
the
new
FDA
encode.
All
we'll
do
is
rename
it
so
that
new
FDM
is
now
at
VM.
The
other
one
is
old,
a
TM.
It
will
maybe
give
people
like
one
version
to
sort
of
get
off
the
old
and
then
the
week
that
code
well,
so
we
will
import
the
new
at
DM.
A
A
A
Okay,
so
this
is
where
things
actually
start
to
change
a
little
on
this
code.
Actually,
it
doesn't
need
to
change,
because
I
used
keywords
for
everything
already,
but
right
now,
so
in
the
past
we
actually
had
sort
of
a
separation
between
what
are
the
arguments
in
order
of
the
keyword
arguments
and
with
this
I'm
shifting
much
more
towards
putting
everything
on
keyword,
arguments
and
why
we
want
to
potentially
do
that
is
that
it
allows
you
to
develop
or
build
each
object
in
sort
of
an
order.
A
A
So,
if
we're
shifting
towards
using
keyword
arguments
that
also,
that
means
that
our
keywords
need
to
be
meaningful
I.
So
in
a
lot
of
places
for
you're
using
a
lot
of
shorthand
I
think
we
should
try
and
get
away
from
that,
especially
as
we're
trying
to
get
sick
by
adopting
like
sigh
shorthand
to
us,
doesn't
necessarily
translate
especially
people
speaking
different
language.
A
Yeah
so
like
Loucks
I
think
should
just
be
locations,
but
what
we'll
do
and
I've
done
this
with
the
solver,
for
example,
it's
like
we
used
to
have
solver
as
capital.
That's
so
I'd
still
let
you
set
that
and
I'll
still,
let
you
call
it,
but
we'll
throw
a
warning
for
now,
and
so
we
can
do
that
with
things
like
books
or
whatever
you
can
we'll.
Let
you
set
it
bubble
say:
don't
do
that!
A
A
Okay,
so
here
is
the
the
solver
that
I
want
you
to
say
and
then
what
I've
done
here
is
I've
got
a
soul
burn
like
this,
and
so
it's
a
property
and
as
soon
as
you
call
it
throws
a
warning
and
says,
use
this
one.
Instead,
oh
right
now,
I'm
racing
an
exception
because
that's
trying
to
find
where,
in
the
old
code
I
was
still
using
this,
but
then
they
will
return
the
solver
so
actually,
but
it'll
just
return
the
solver
and
then
here
on
the
setter.
We'll
do
the
same
thing.
A
I
K
F
A
You're
just
chatting
through
upgrading
the
co2
simulations
so
going
towards
using
keyword.
Arguments
for
everything.
Did
you
guys
trying
to
use
keyword
arguments
because
it
been
allows
you
to
construct
the
object?
You
know
much
more
flexible
way
so,
especially
when
you're
thinking
about
the
simulation
in
particular
like
maybe
you
want
to
create
the
survey
and
then
design
the
match
based
on
your
survey,
or
maybe
you
want
to
create
a
mesh
and
then
from
that
crates
are
great,
like
whatever
we
want
to
give
you
that
flexibility,
and
so
it's
a
lot
easier
to
do
that.
A
If
we're
setting
stuff
through
keyword
arguments
rather
than
saying
like
when
you
instantiate
the
class,
we
need
these
things
so
yeah
and
then
what
we'll
do
is
put
a
validation,
introduce
that
then,
basically
just
says
like
before
you
try
and
use
this
thing.
If
you
haven't
set
a
mesh
and
weedeater
match
to
solve
your
report
simulation
like
will
throw
an
error,
I'm
even
and
then
we've
talked
about
just
now.
A
If
we're
using
keyword
arguments
that
then
means
we
need
to
be
very
clear
with
what
those
keywords
are
also
things
like
Louis
I'm,
not
particularly
a
fan
of
just
cuz.
It's
not
necessarily
all
that
clear.
So
we'll
start
moving
that
towards
locations
and
just
spell
out
the
English
word,
but
we'll
do
that
in
a
way
that
lets
you
sort
of
set
these
things
in
and
throws
a
warning
on.
A
A
A
A
So
now
we
have
the
keyword
set
I.
Do
this
problem
about
pair
step
like
every
Ford
simulation
code
is
gonna,
have
in
it
and
right
now,
I.
Actually,
let
you
do
that.
So
what
we've
done
is
I
put
on
the
simulation.
A
pair
happen
at
all
its
gonna
do
is
say
like
chartiers,
it
just
says
yeah
so
a
pair
here,
it's
just
been
self,
sir,
it,
sir.
So
what
you
there's
I'll
show
you
from
the
top
level
what
you
could
do
instead,
so
here
we
could
have
done.
Then
we
had
set
the
survey
first.
A
A
A
A
Not
quite
so,
we
actually
there's
a
bit
of
confusion
going
on
I
think
in
this
notebook,
because
we're
still
using
the
word
problem
in
the
new
framework.
The
problem
is
gone,
it
thought
no
longer
meals,
so
we
have
the
simulation,
and
that
does
everything
that
the
problem
used
to
do.
But
now
it
actually
has
the
survey
as
a
property
of
that
class.
Okay,.
O
A
N
G
A
Data
is
actually
this
was.
This
is
something
interesting
and
so
data
is
actually
been
teased.
Apart
from
the
this
survey,
so
data
is
its
own
object.
It
takes
the
survey,
I'll
show
you
the
classes,
and
so
what
we've
tried
to
do
here
is
actually
sort
of
a
look
at
the
data
in
a
sense.
So
here
are
our
data
class.
A
We've
got
the
Oz
and
when
we
have
a
true
data
class,
what
we're
going
to
do
is
once
you've
sent
the
ops
not
allow
you
to
change
it,
and
then
this
is
something
we
can
debate,
but
I
think
that,
like
once
your
D
office
is
set,
we
can
take
actions
on
it
and
you
can
sort
of
choose.
You
know.
I
only
want
to
convert
it's
a
subset
of
data.
That's
all
fine,
but
I,
think
sort
of
once
you
get
to
this
step.
A
Your
observed
data
is
your
return
data,
and
if
you
want
to
do
some
sort
of
filtering
or
something
before
you
invert,
it
like
just
make
any
yeah
and
then
what
would
we
have
done
to
allow
a
bit
of
flexibility?
They've
also,
then
created
a
synthetic
data
objects,
and
we
do
then
lets
you.
You
can
go
in
and
change
your
yeah.
Are
you
observe
data
for
that.
A
So
the
where
this
changes
things
is
the
data
misfit.
So
the
data
misfit
will
now
take
a
data
object
and
a
simulation,
whereas
before
it
used
to
take
a
survey
because
the
survey
used
to
know
about
it,
but
so
we
sort
of
reversed,
if
that
makes
sense,
we've
reversed
the
role
of
survey
and
data
data
is
now
more
important
than
survey.
Data
is
the
class
and
it
has
the
surveys
of
property
and
then
the
simulation
is
a
class
that
has
the
surveys.
A
D
B
D
A
So
deep
red
is
just
a
vector:
oh
it's
just
a
vector
what
you
can
do
on
this
simulation
class
is
we
do
have
they
make
synthetic
data,
so
we've
use
that
in
a
lot
of
places,
and
so
this
will
give
you
back
a
data
object.
So
then,
if
you
wanted
to
actually
do
a
synthetic
conversion,
this
is
what
you
want
to
work
with.
So
here
we
have
our
clean
data.
It's
just
in
do
self-defense
so
that
it
keeps
you
vector
of
numbers
that
are
your
data.
A
C
A
A
Okay,
so
survey
is
used
to
basically
allow
us
to
stainle
index
your
absurd
data.
So
your
observed
data
is
just
a
vector
of
numbers,
and
the
order
of
those
numbers
depends
on
where
your
sources
in
a
supersonic
on
so
what
we
do
is
actually
build
under
the
this
is
all
under
the
hood.
But
what
we
do
is
we
build
a
dictionary
of
indices
based
on
your
source.
A
So
yours,
your
survey
has
a
source
list
and
each
source
has
a
receiver
list,
so
we're
gonna
go
through
and
for
each
source
in
the
survey
we're
looking
every
receiver
and
just
say:
okay,
what
indices
correspond
with
indices
in
Theon's
correspond
to
you
and
store
those?
So
then,
here
you
can
on
the
outside,
actually
access
the
data
like
a
dictionary,
so
you
can
do
data
of
source,
comma
receiver
and
get
it
back
data
corresponding
to
that
receiver.
D
Okay
Oh,
as
well
as
the
data
class
like
what
kind
of
fields
can
actually
contain.
Is
it
limited
like
only
data
the
data
fuss?
Can
you
only
take
the
ops
and
an
uncertainty?
It's
at
all
evident.
So,
let's
say
like
the
DC
data
you
go
out
and
like
there's
really
like
eight
to
ten
different
things
that
they
often
have
to
measure
for
single
like
measurements,
you
might
have
your
injection
current.
D
You
might
have
your
transmitter
voltage,
you
might
have
cost
cell
potential
or
like
so
you
might
have
all
these
different
things
that
are
measured
at
the
same
time
where
we
might
have
like
the
index
of
each
of
your
different
electrodes
that
are
being
used,
you
might
have
the
index
of
your
ATM
and
electrodes.
Is
there
any
way
that
we
can
add
any
of
those
other
parameters
to
the
data
class?
D
A
A
This
is
actually
new,
so
we
didn't
previously
have
sort
of
a
frequency
domain
data
object,
but
now
we
do-
and
so
it's
very
simple
but
there's
a
couple
things
that
are
just
sort
of
handy,
and
so
perhaps
you
want
to
get
you
know
which
frequencies
are
or
if
you
have
named,
transmitters
or
stuff
like
that,
we
can
start
to
like
you,
can
inherit
this
class
and
then
add
what
you
need
for
that
sort
of
thing.
So.
G
Hopefully
good,
so
what
is
it
before
cuz
say
like?
Let's
say
you
have
some
Ireland
being
a
state
like
you
want
to
put
that
in
class
work
or
the
circuit
class
we're
like?
How
are
you
actually
the
person
that
could
set
so
there's
two
things.
Let's
say,
although
they're
all
the
information
that
Mike
said
it's
actually
abandoned
like
a
fire
outside
yeah
you're,
reading
demos,
information,
some
of
them
are
considered
to
be
never
there.
Other
pieces
might
be
useful.
G
A
I
already
in
class
I
know
so,
and
it's
just
something
we
should
talk
about,
but,
like
my
tape
on
that
is
that
the
I/o
it
should
be
separate
and
it's
actually
like
a
utility
or
a
parser
type
package.
Does
each
one
like
every
file
format,
it's
very
custom,
and
so
I
don't
want
to
bog
down
like
a
data
class
with
everybody
everybody's
different
fire
formats,
but
to
have
a
like
a
a
utils
dog
breed
yes
and
then
bring
that
in
and
give
you
a
surveyor
object
to
the
object.
Oh.
E
B
G
A
I
know
I
think
it's
totally
fine
to
think
about
bringing
in
if
there
is
supplementary
data,
that's
useful
or
if
there's
utilities
on
how
you
want
to
look
at
your
data.
So
it
like
you,
know,
sort.
My
transmitter
is
by
packet
length
like
that
I
think
would
make
sense
in
the
data
class,
because
that
is
we're
acting
on
their
cake
object
and
we're
actually
missing
a
data
object,
but
a
lot
of
the
I/o
and
parsing
and
stuff
like
that,
because
it's
so
custom.
G
Well,
goodness,
emergency
everyone
that
much
information,
how
bad
the
data
are
and
that
way,
actually,
if
they're
in
any
way
they
after
you
do
the
inversion
like
okay,
it's
this
portion
of
the
day
for
off
and
that
portion
of
state
is
wrong.
How
I
know
that
that,
through
cell
phone
hundred
information.
G
G
G
A
So
I
think
that
it's
like
actions
that
you're
taking
on
the
data
object
once
it's
already
ancient.
So
it's
it's
instant,
maybe
this
to
make
the
object
and,
if
you're
doing
a
decent
survey-
and
there
is
like
you-
want
a
sort
by
barrel
length
that
should
I
think
that
that's
fine
having
it
across
I
think
that,
like
bringing
in
if
there
is
metadata,
we
can
have
like
a
place
to
put
some
structured
metadata
in
the
data
class.
I.
Think
that's
totally
fine!
I!
D
Or
we
form
the
data
class,
we
just
kind
of
specify
like
what
hitters
we're
going
to
have
basically
for
our
different
data
columns.
So
just
have
like
the
data
object.
Feet
is
pretty
flexible
thing,
so
I
gave
requires
like
a
minimum
number,
the
family.
It
requires
your
observed
data
and
maybe
your
uncertainties
to
set
up.
D
A
I
think
those
one
other
thing
that's
to
just
look
be
a
little
careful
of
is
like
some
pig,
doesn't
think
of
data
as
looking
an
Excel
spreadsheet
like
we
don't
think
sort
of
XY,
it
said
data
so
that
that's
just
one
thing
to
sort
of
keep
in
mind.
So
it's
not
starring.
The
obvious
is
a
vector
the
source.
The
office
is
the
vector
something.
B
A
Well,
we
build
a
dictionary
based
on
the
survey
that
keeps
track
of
with
what
indices
in
the
data
vector
belong
to
web
sources
innocent,
so
that
dictionary
itself
doesn't
actually
contain.
You
know
it
just
contains
I
mean
the
others
yeah
it
just
contains
introducing
that
this
index.
This
is
just
a
short
list
for
supreme,
and
so
you
can
use
the
exact
significant
area
on
your
uncertainties
right
because
it
your
uncertainty,
should
be
the
same
workers
you
dropped
and.
P
A
C
A
L
A
L
A
But
this
implementation
is
going
to
require
that
you're
doing
inheritance
properly
and
what
I
mean
by
that
is
that
or
why
that's
going
to
be
important
is
because
there's
a
lot
of
certain
steps,
these
each
of
these
properties
here
that
we
implement
on
the
middle
class.
That's
actually
something,
then,
when
we
serialize
the
class.
So
when
we
go
to
either
save
it.
If
you
want
to
sort
of
save
that
object
at
this
or
if
you
want
to
pickle
it
so
that
we
can
paralyze
it.
A
These
are
the
pieces
of
information
that
gets
saved,
and
so,
when
you
inherit
from
the
base
box,
we
also
inherit
those
pieces
of
information
that
gets
it,
and
so
here,
when
I'm
doing
an
instantiation,
this
one's
a
couple
steps
because
I'm
trying
to
be
kind
to
people
when
we're
renaming
stuff.
But
this
is
really
the
important
step
is
we're
instantiating
on
the
parent
class,
so
super
just
says:
call
my
preparing
class
with
the
instance
of
self.
A
So,
with
the
the
instance
of
this
guy
and
then
we're
going
to
instantiate
it
with
these
keyword
argument,
and
so
what
that's
gonna
do
is
it's
then
going
to
go?
Okay,
what
what's
in
the
base
FDDI
resource?
So
here
my
base,
FDM
source
isn't
serving
so
it
has
frequency
and
it
has
a
receiver
list,
and
so
it's
going
to
bring
those
objects
with
it.
A
A
But
if
you
overwrite
these
properties
and
don't
actually
do
them
as
properties
and
you
lose
all
of
that
functionality.
So
when
you're
building,
try,
try
and
inherit
and
use
the
functionality
from
the
base
and
then
only
the
implement.
If
you
really
do
need
something
different
to
be
happening,
and
so
we
can
deal
with
that
or
not
a
case-by-case.
N
G
A
You'll
only
be
serializing
what's
set,
and
so
that's
an
important
thing
to
to
keep
in
mind
is
that
whatever
you
need
to
recruit
class
so
here
in
order
to
recreate
a
reasonable
frequency
resource,
I
need
a
frequency,
and
the
receiver
list
can
be
active,
so
I
don't
actually
need
that
so,
but
we
do
need
a
frequency.
This
is
required,
so
it
has
to
be
a
property
that
we
see
realize
because
I
need
it
to
recreate
the
class.
If
I
had
something
else
here
like
cohesion,
for
example.
A
A
A
So
it's
not
it's
nicely
spaced,
but
it
saves
everything
that
I
need
to
recreate
my
mesh
parameters
class
or,
if
we
do
actually,
this
will
work
if
I
do
mashed-up
serialize
tested,
so
it
saves
which
class
it
is.
It
saves
the
number
of
Seles
image
dimensions
that
you
need
to
know
that
to
recreate
the
class,
the
Cartesian
origin,
because
this
is
a
so
much
and
then
my
H
spaceman's,
and
that's
it
because
that's
all
that
I
need
to
recreate
the
class
right
to
build
all
the
operators
and
stuff
like
that.
A
A
B
E
E
B
E
And
it
started
to
work,
but
then
I
was
remembering:
oh
yeah,
there's
something
with
this
negative
depth
value,
so
I'm
trying
to
build
right
now,
just
like
a
deist
class
or
data
class.
That
will
just
send
off
an
observation
object
for
you,
guys
Air
Force
in
peg,
and
is
it
going
to
be
negative
or
not
like,
which
wasn't
decided
we're
just
gonna.
Keep
it
positive
up.
Okay,.
A
N
E
O
E
B
B
G
G
E
P
J
E
B
B
Yeah
I
mean
it's
it's
all
there,
but
it's
it's
missing.
I
mean
it's
only
getting
to
the
kind
of
to
the
Fourier
and
then
it's
missing
all
the
kind
of
specific
things
to
the
MT
theta
I'm,
taking
the
Fourier
coefficients
and
actually
making
them
into
the
empty
theta.
At
least
I
didn't
find
it
in
this
on
the
kit.
But
again,
I
didn't
spend
a
whole
lot
of
time.
Looking.
E
Yeah,
it's
pretty
bare
right
now,
I'm
just
kind
of
put
up
what
I
and
we
had
four
just
stacking
DC
time-series.
So
yeah,
there's
no
40
analysis
yet,
but
yeah
come
in
due
time
work.
A
little
hectic
and
I
was
on
vacation
too,
but
I
got
like
a
basic
structure
here:
yeah
take
it
through
yeah
I'm,
just
I
think
we
were
discussing
than
this.
Originally,
you
know
we'll
just
have
whatever
kind
of
windows
you
want
for
filtering.
You
just
don't
get
this,
make
that
clear.
E
Just
all
methods
or
functions
sd-6,
oh
okay,
so
numb
caps
is
just
I.
Guess
it's
kind
of
a
poison
hat
lingo
that
we've
been
using
like
in
since
University.
It's
just
like
how
many
points
really
when
we
would
plot
them
in
school.
We
bought
them
as
caps,
and
it
would
just
be
like
a
step.
So
you
just
have.
They
would
be
a
line,
a
point,
a
line,
a
point
and
you
just
and
so
you're
using
that
to
say
so
you
put
how
many
tabs
is
you.
B
A
A
E
So
yeah
that's
caps
is
just
referring
to
and
like
I
think
it
was
a
MATLAB
command,
be
like
tap
this
thing
and
it
would
just
plot
it
up.
I
think
that's
where
I
came
from
so
right
now,
yeah
I've
still
needed
a
branch
over
this
report
over
this
code
from
my
C++,
but
yeah
we've
got
Kay,
Kyser,
anding
chebyshev
right
now,
and
anybody
who
has
ideas
and
any
other
ones
just
throw
them
at
me
and
you
throw
them
in
there
another
function,
just
if
you
have
so
after
you've
stacked.
E
You
can
throw
your
stack
in
to
get
primary
voltage
and
you
just
tell
it
wherever
you
want
it
to
calculate
the
voltage
over
and
Allah
returned
your
VP
back
load,
Deus
I
think
this
is
where
I
was
trying
to
make
listens
just
specific
stuff,
so
yeah
I'd
put
up
some
files.
If
you
just
wanted
to
load
the
file
into
a
nice
little
data
structure,
the
data
structure
for
the
DEA's
is
exactly
what's
in
the
header,
so
it'll
be
like
load.
Deus
and
it'll.
Give
you
a
object
like
Diaz
taught
EP
air
in
air
current.
E
And
then
so
I
started
I
was
thinking,
we'll
call
the
just
name,
a
trade
out
at
DC
or
dcpip
time
series.
This
will
be
like
the
general
class
that
I'll
form
for
whatever
kind
of
data
that
you
want
to
load
in
the
BS
GDD.
Wherever
data
it
will
end
up
making.
This
is
in
general
class
and
then
yeah
I'll
populate
this
with
what
we
need.
I'm,
not
you're,
sure
much
more.
E
You
need
from
just
the
time
series
just
really
want
the
time
base,
sample
rate
data,
then
oh
yeah,
and
then
you
can
just
tell
it
how
many
so
I
get
some
statistical
stuff.
How
many?
How
many
period
you
have?
How
many
half
periods
help
you
make
your
windows
in
this
and
then
yeah.
Just
this.
This
is
my
first
crack
at
going
at
brut
stack,
okay.
So
this
is
one
of
my
older
codes
but
yeah
it's
pretty
efficient.
It
was
written
breaking
region
numpy,
but
this
one
just
uses
the
it
doesn't
use
a
window.
E
It's
just
like
your
know.
Your
know
whole
stack
unfiltered.
It's
just
kind
of
like
a
boxcar,
except
for
the
two
ends
a
little
bit
doubled,
but
yeah
the
throws
of
a
stack
if
you're
just
looking
PI
series
and
just
want
to
briefly
look
at
the
amplitudes,
and
then
this
is
the
one
I'm
working
on
right
now.
This
is
my
fast
one.
This
is
the
one
that
I
use
for
our
code.
E
What
halfway
done
I
found
some
refactoring
I
want
I,
maybe
I'll
change
this,
but
I
should
have
that
ready
by
next
week,
yeah
and
then
I've
started
in
with
ensemble
stacking.
So
once
I
get
the
normal
stack
going,
this
one
won't
take
much
longer
because
I'll
just
have
to
break
it
into
its
ombos.
The
staff
rejection
will
take
a
little
bit
of
time
but
yeah.
So
there
will
be
those
three
methods
available
at
this
time.
For
you
guys,
and
then
this
is
just
yeah,
stuff
old
stuff
that
I
had
for
loading
ideas
file.
G
G
E
G
E
E
C
E
E
E
G
E
Actually,
yeah
I
guess
that's
a
good
point
yeah.
So
what
this
is
doing
right
now
is
it'll
just
give
you
that
vector
which
is
the
stack
and
then
you
would
have
to
feed
that
stack
into
something
else.
To
get
say,
your
I
see
what
it
was
per
se
yeah
so
yeah.
Maybe
that
should
be
rethought
so
that
you
not
always
just
passing.
B
G
G
E
So
that
would
be
like
kind
of
like
your
next
step,
because
you're
talking
about
extracting
yeah
your
decay
first,
for
example,
so
right
now,
that's
not
really
like
what
I
was
going
to
do
is
just
have
him
again,
a
method
that
you
feed
at
this
stack.
And
then
you
feed
it
your
channel
time
with
those
right
and.
B
E
A
Thing
that
might
look
at-
and
we
talked
about
this
mirror
originally
thinking
about
it
processing-
is
to
use
the
ideas
that
we
have
in
the
mapping
class,
okay
and
so
what's
kind
of
interesting
about
that
is
each
of
these
steps
in
a
sense
you
can
think
of
it
as
a
matrix
operation.
It
isn't
the
way
you
can
think
of
like
okay,
so
we've
got
the
vector,
so
there
is
like
a
defined
shape
the
process,
and
so
what
we've
done
with
mappings
is
you
can
basically
compose
all
the
chef's
just
wait
with
the
multiplication.
A
So
here,
if
you
wanted
to
first
filter
and
then
Stan
I
mean
it's
supposed
to
be
sex,
so
maybe
you
want
to
play
with
the
fiddle
with
your
parameters,
choose
the
correct
way.
So
we
wanted
to
go
in
that
one
data
set.
We
do
those
things
independently
and
then
you
can
create
any
mapping.
It's
just
a
composite
of
those
two
and
so.
A
B
A
A
R
E
G
G
A
A
E
S
A
E
A
A
It's
an
identity
map,
so
you
get
the
same
thing
back
and
then
potentially
a
inverse
that
derivative.
But
then
what's
interesting
is
this
guy
here
and
so
this
happens
when
you
then
propose
max,
and
so
in
same
thing,
we
do
that
with
the
multiplication.
So,
for
example,
if
you
are
inverting
and
frequency
domain
data-
and
you
just
want
to
invert
for
one
thing-
use
a
power
exponent
or
law
conductivity
below
the
subjects
yeah.
In
order
to
model
it,
we
need
real
conductivity.
A
Everyone
are
doing
so
you'll
develop
an
X
which
takes
log,
it
translates
it
and
then
we
have
inject
yourself
back.
So
you
have
those
two
steps
and
we're
gonna
multiply
them
together
to
go
from
lot
of
comics
of
e-books
or
this
to
comic
in
there,
and
so
all
this
guy
does
like
when
we
do
the
multiplication.
B
A
E
A
I
A
A
Above
and
then
here
it
just
creates
a
list
and
then
basically
says
like
these
operations,
but
what's
potentially
interesting,
processing
flow,
and
especially
if
we
start
looking
at
things
like
x-ray
or
some
of
these
more
optimized
packages
that
have
really
thought
about
like
let's
fill
the
pipeline
and
then
efficiently
compute
on
it,
we
could
potentially
play
with
some
of
that
sort
of
stuff.
At
this
level,
Thanks.
N
A
E
E
E
E
E
C
E
E
E
C
G
E
Had
a
job
I
remember,
but
it
is
rather
large
I
think
there
is.
There
should
be
at
least
at
least
5,000
injections.
I
would
say
the
very
least
on
this
data
set
and
each
injection
probably
has
upwards
of
2,000
eyeballs
for
yen
per
injection.
I.
Think
this
way
around
its
kind
of
to
offset
three
by
six
grids
like
sorry,
three
kilometers
by
six
kilometer
grids,
they're
just
kind
of
just
offset
a
little
bit
so
yeah,
it's
it's
pretty
decent
area,
maybe's
inside
I,
think
it's
four
hundred
and
ninety
thousand
data
points.
J
J
E
I,
don't
remember
the
cell
size
that
she
was
using
on
the
immersion
right
now,
but
the
range
of
dipoles
go
anywhere
from
thirty
to
think
it's
around
four
hundred
we
did
kind
of
us
or
I.
Did
it
together.
I
made
an
algorithm
that
does
like
a
smart,
dipole
sizer.
So
if
it's
far,
if
the
bills
are
farther
away
from
the
injection,
it
makes
a
bigger
dipole,
so
yeah,
it's
gonna
vary
from
being
work
from
there
and
I
think
in
increments
it
was
thirty
north-south
and
sixty
east-west
for
the
spacing
of
the
the
nose.
P
E
Was
just
the
first,
the
first
one
was
where
they
found
South
arrow
and
because
of
that,
that's
when
we
did
the
second
third
and
yeah
there's
some
more
nice
empty
stuff
there,
but
yeah
I
can
ask
Charlotte
to
see
if
she
just
has
that
subset.
You
just
use
arrow
I
think
that
one
is
still
approximately
like
for
three
hundred
and
fifty
four
hundred.
Fifty
thousand
eight
points
as
well.
E
Well,
actually,
no
one
would
be
less
yeah,
probably
looking
every
what
probably
maybe
have
as
much
200
to
250,
because
if
I
do
got
it,
we
just
we
have
so
many
digits
heads
right
now
we
have
the
smart
sizer,
the
one
that
was
used
for
the
South
arrow
deposit
was
a
just
collective,
but
it
wasn't
sized.
It
was
just
between
30
and
400
every
dipole,
so
it's
gonna
be
a
little
bit.
It'll,
be
a
bigger
guy,
it'll,
be
the
smaller
smaller
area
but
and
weird
angle,
but
isn't
any
dipoles
or
did.
E
P
P
M
E
Computer
I
think
we
have
just
250
some
256,
yeah
I,
think
Flynn
upgraded
and
then
yeah
the
drives.
A
special
drive
that
works.
I
can't
remember
what
he
connected
it
through,
but
it's
a
fast-acting
like
hard
drive
or
it's
kind
of
like
a
mix
between
the
hard
drive
and
RAM.
It's
just
it
stores
like
hard
drive,
but
it's
fast
accessing
like
a
ram
and
then
I
think
we
have
like
three
or
four
gigs
or
terabytes
on
that.
E
A
I
know
I'm
Joe
at
Clarksville,
mines
he's
been
working
on
auction
and
so
he's
got
a
fork
on
his
personal.
They
have
repo
and
he's
got
the
quadtree.
Some
of
the
basic
operators
forming
and
think
he's
got
made
some
progress
on
the
octree
and
he's
a
very
solid
C
programmer
and
Python.
So
he
knows
how
to
layer
that
so
it's
easy
for
us
to
access,
so
we're
gonna
connect
with
him.