►
From YouTube: DevoWorm ML: Week 10 (General discussion)
Description
Tenth DevoWormML meeting, November 6. Attendees: Bradly Alicea, Richard Gordon, Jesse Parent, and Thomas Harbich
A
A
A
Perpendicular
to
the
optic
axis
so
in
facing
the
same
problem
as
a
habit
right
now,
and
the
point
is,
if
you
were
just
bring
it
here
between
two
surfaces
like
glass
plates,
maybe
slider
and
coverslip
the
tendency
to
adhere
to
one
of
these
surfaces
and
it's
time
for
a
movie.
So
just
absolutely
free
water.
A
If
we
bought
us
and
so
it's
a
problem,
so
the
video
we
have
just
considered
now
are
great
in
the
respect
that
then
we
feel
it
with
the
exception
of
pattern
number
one
because
that's
attached
to
the
substrate,
and
so
it's
really
a
great
luck
to
have
such
an
opportunity
to
take
such
a
video.
So
this
makes
it
so
exceptional.
A
C
A
A
Okay,
so
I
just
Saint
supposed
to
you
our
email,
just
a
second
before
just
simple
taneous
ly,
to
your
mates-
and
this
is
just
a
a
picture
showing
that
they
really
behave
like
a
sinus
wife.
So
they
have
every
ethnicity
and
it's
really
striking
that
these
parents
follow
all
the
rules
which
we
are
expecting.
So
it's
really.
B
A
B
C
A
A
B
A
Some
of
them
are
in
rest
position,
so
the
amount
of
colonies
is
increasing,
is
breaking
the
evening
and
in
the
dark
HR
part
is
in
the
rest
position,
but
that
statistics,
so
you
think,
there's
a
wavelength
of
light
which
you
could
observe
them
but
which
for
them
would
appear
yes
and
I,
made
a
make
the
trial
to
observe
them
in
the
infrared
but
astride.
So
it's
not
a
complicated
thing,
but
the
most
microscopes
are
using
are
not
really
usable
from
for
that,
because
the
LED
light
meant
allocate
lamps
and
the
aliens
don't
have
infrared
part.
A
Does
not
compensate
normal
pipe
and
there
can
make
these
experiments
so
by
then
just
married
at
the
very
beginning.
So
it's
it's
too
early
to
give
any
answer,
but
connecting
well
impressions,
ideas
and
observations.
So
just
watching
particles
moving
along
the
diatoms,
the
outer
diatoms
and
now
make
the
speed
of
them,
and
my
trust
right
now
is
that
they
stay
exactly
the
same
speed
as
the
diatoms
move
relatively
to
the
edge
a
particles.
A
D
C
A
C
B
D
Okay,
let
me
assure
the
screen
here
where's
that
yeah
sure
in
the
lower
right,
I.
D
D
E
B
G
C
B
C
Well,
thanks
Richard
and
Thomas
for
that
update
on
the
bass,
hilarious
stuff,
so
yeah
I
was
looking
through.
The
ever
I
had
a
chance
to
look
through
Thomas's
revisions
and
like
his
stuff
that
he
sent
he
sent
a
pretty
long
email
about
it.
So
I'm
gonna
go
through
that
later
this
week
or
by
the
end
of
the
week,
probably
and
no
longer
responds,
but
it.
A
Just
took
me
two
days
to
work
it
out
and
present
the
results
so
just
to
be
a
secret.
What
do
you
think
about
making
an
appendage
to
your
paper
and
just
for
comparison
and
say,
what's
possible
distressing
methods?
I
think
it
works
only
if
you
have
a
perfect
video
where
the
colonies
are
perpendicular
to
the
observation
direction
to
the
optical
axis
and
there's
enough
structure
in
each
tire
tool
to
identify
this
way
in
this
video
I.
A
It's
it's
not
easy
to
model
that,
but
anyway,
so
I've
been
working
out
the
detail,
but
for
the
paper,
which
is
New
York
proposal
to
make
appendix
just
to
say
the
paper,
there
was
a
mess
of
classical
and
methods
for
tracking
parts
of
a
picture
and
that
could
be
used
as
well,
but
it's
not
as
flexible
as
intelligence
and
machine
learning
approach.
So
it's
not
comparable.
So
in
some
cases
maybe
it's
more
effective.
It's
more
successful
to
use
these
classic
methods,
but
in
general
I
would
say
the
powerful
method
paves
the
way
for
for
future.
A
C
Yeah
I
think
that's
a
good
way
to
interpret
it
too,
so
yeah
I'll,
look
it
over
and
I'll
I'll
give
you
some
feedback.
It's
I
mean
it
looks
like
you
know.
Everything
is
good.
So
thanks
for
yeah,
thanks
for
that
and
we'll
just
keep
talk.
Well,
you
know
you
can
provide
I,
don't
know
what
additional
information
you
want
to
provide,
but
I'll
ask
you
if
I
need
anything
specific
on,
you
know
special
that
you
haven't
provided
so.
C
Yeah,
ok,
so
yeah
thanks,
so
Jesse
I
saw
that
you
were
busy
we're
discussing
kind
of
a
little
bit.
Well,
it
actually
isn't.
We
were
discussing
like
you
said
that
you
wanted
to
find
out
like
how
to
sort
of
navigate
the
area
of
machine
learning
and
all
that-
and
he
said
you
didn't
really
have
a
background
in
in
biology
or
machine
learning
too
much.
C
A
H
H
C
Mean
like
and
I
can
speak
for
myself.
I,
don't
know
you
know.
A
lot
of
the
stuff
is
pretty
new,
so
I
didn't
learn
this
in
school.
Necessarily
so
I
mean
like
in
this
everyone.
When,
when
new
things
come
up,
you
know
you
have
to
kind
of
pick
up
pick
them
up
as
you
go
and
like
so
I
guess,
yeah
I,
like
the
idea
of
like
creating,
maybe
not
formal
education
materials
but,
like
you
know,
sort
of.
C
Recently
so
this
is
again,
this
is
one
example,
I
sort
of
started
this
as
a
way
to
do
almost
nothing
about
machine
learning.
You
could
go
to
these
preferences
and
get
some
idea
of
what's
going
on
and
this
needs
to
be
improved,
but
this
is
we
have.
This
is
all
text
really,
so
we
have
section
off
the
shelf
tools,
so
you
have
a
set
of
tools
that
you
can.
You
know
you
know
rival
tools
that
you
could
look
at
and
see
which
one
works
best
for
you,
majors
venues.
C
You
know,
there's
some
peer
review
journals
in
here,
but
a
lot
of
the
action
actually
happens
on
archive
and
in
this
thing
called
distill,
which
is
an
open
journal,
and
so
you
know
that,
like
it's
good
to
have
an
idea
of
where
to
find
this
work.
Where
published,
where
is
it
discussed?
And
then
we
have?
You
know
some
examples
of
like
pre
train
models,
so
we
have
the
blog
post,
but
we
also
have
this
set
of
models
that
you
know
you
can
look
at
examples
of
then
special
topics.
C
So
it's
almost
kind
of
structured,
like
maybe
a
course
you
would
have
at
university,
almost
like
a
syllabus,
so
Neurotech
might
open
syllabi.
You
know
just
having
like
something
assembled
that
you
can
go
through
and
you
know
look
at
special
topics.
You
know.
So
this
is
just
general
topics.
Special
topics
like
naivebayes
evolution
strategies,
and
you
say
well
what
are
those
well,
you
can
link
to
them.
C
This
isn't
like
an
exhaustive
bibliography,
but
I
think
I
know
again
it's
something
that
people
can
go
to
and
like
look
at,
you
know,
look
at
those
references
and
go
to
other
references,
but
that's
why
I'd
actually
like
to
have
maps,
because
you
know
this,
you
know
you
can
give
people
references,
but
it's
hard
to
integrate
all
that
information.
So
having
a
map
is
a
good
tool,
a
good
sort
of
top-down
tool,
okay,
yeah
so
Jessie
says
yes,
the
syllabus
and
key
issues
and
branches
of
development
in
a
field
that
can
help
a
lot
see.
C
I
mean
this
is
again
this
is
you
know
this
is
kind
of
structured
like
a
syllabus,
but
it's
meant
to
get
people
some.
You
know
some
background
on
different
topics,
different
you
get
a
feel
for
the
field
and
then,
of
course,
we
have
MOOCs
and
general
tutorials.
So
those
have
been
invaluable,
I
think
especially
for
like
for
this
field,
because
it
moves
so
fast.
You
know
you
have
different
MOOCs
rival,
MOOCs,
like
geoff
hinton
offer
and
during
offers
one.
C
But
I
mean
for
this.
For
this
purpose
again,
it's
just
kind
of
like
really
basic
stuff,
and
so
yeah
I
mean
that
that's
that's
sort
of
like
for
Jesse.
You
know
like
that
would
be
for
just
for
machine
learning
and
maybe
some
biology
I
have
another
reference
for
an
evil
worm.
So
when
I
do
the
Summer
of
Code
stuff
I
get
people,
you
know
they
come
to
me
and
they
want
to
do
a
project.
C
They
want
to
write
a
proposal
and
then
either
at
that
point
or
maybe,
if
they're
selected,
and
they
want
to
get
in
sort
of
get
sort
of
immersed
in
the
topics
like
basic
C,
elegans
biology,
say:
oh
I
have
a
set
of
references
for
that
too,
and
they
are
able
to
go
through
those
I've
done.
This,
in
a
way,
that's.
C
C
C
You
know
this
is
it's
open
access?
You
have
original
peer-reviewed
chapters
and
all
different
topics
molecular
biology,
developmental
control
and
so
forth,
now
they're
at
as
a
rather
high
level.
But
you
know
this
is
gonna,
be
in
conjunction
with
some
other
references
resources.
So
the
virtual
worm
model
I
like
this,
because
you
can
go
from
like
an
article
like
that
to
like
this
virtual
model
of
C
elegans,
and
so
this
is
the
actual
model
here
and
you
can
play
with
it.
C
You
can
like
you
can
add
in
cells,
it's
not
working
right
now,
but
okay,
here's
an
exhibit.
These
are
screenshots.
Okay,
so
you
know
you
can
look
up
specific
cells
in
this
model,
and
so
this
is
an
adult
C
elegans
and
the
the
C
elegans
is
unique
because
they
have
every
cell
mapped
in
the
adult,
and
so
you
can
actually
play
with
the
different
cells
in
the
adult,
and
this
is
actually
neuron
with
projections
and
you
can
actually
learn.
C
Like
you
know,
in
the
literature
they'll
talk
about
specific
cells,
you
can
go
back
to
this
and
look
at
the
cell
where
it
is,
you
know,
maybe
get
some
information
about
what
it
does,
and
so
I
need
to
maybe
work
on
this
as
well,
because
this
is
very
sort
of
just
links.
You
know
it's
are
textual.
I
was
kind
of
hoping
to
annotate.
This
item
got
around
to
it,
but
this
is
another
reference
and
then
microscopy
were
I've.
C
Had
students
interested
in
microscopy
to
say,
like
I,
read
this
stuff
when
I
want
to
come
into
the
group
or
when
I
want
to
write
a
proposal,
but
I
don't
know
any
really
anything
about
microscopy,
and
so
there
are
some
existing
microscopy
tutorials.
But
these
are
things
that
you
know
again.
You
can
point
them
to
and
again
you
know
people
sometimes
come
at
this
with
almost
no
experience
in
like
biology
or
maybe
no,
you
know,
but
most
of
the
people
have
had
have
no
experience
in
biology.
C
So
I
have
to
give
them
a
crash
course,
but
I'm
not
going
to
sit
there
with
a
slide
deck
for
four
hours
and
go
through
it,
and
you
know
I
think
this
is
more
useful.
You
know
and
then
you
have
some
papers
here
again.
This
is
a
just
an
abbreviated
bibliography,
but
it's
something
I
think
we
can.
They
can
learn,
learn
a
little
bit
more
about
the
topic.
So
I
mean
those
are.
C
Those
are
different
options
and
again
I
think
my
goal
I
think
long
term
for
that
stuff
is
to
make
it
more
visual
to
have
like
you
know,
images
or
like
I,
like
the
virtual
worm
simulation,
because
you
can
do
you
know
you
can
actually
look
and
see
what
the
cells
are.
What
you
know,
the
spatial
relationships
are,
and
you
don't
have
to
you
know-
go
find
a
specimen
or
dissect
it.
You
know
that
it's
all
labeled,
so
that
yeah
I
mean
that's.
C
That's
basically,
I
think
that's
helpful
to
like
in
Jesse's
case,
where
you
know
you,
people
come
into
a
field
and
they
don't
really
know
very
much
about
how
do
they
get
oriented?
It's
a
common
problem,
but
you
know
it's
it's
a
lot
of
times
it's
hard
to
get
the
information
you
need
so
also
yeah
I
saw
Jesse
that
he
made
a
infographic
on
of
your
own
I'm,
like
machine
learning
and
artificial
intelligence.
C
So
I
like
that.
That
was
pretty
good.
You
have
a
lot
of
milestones
on
it.
You
are
you
planning
to
get
onto
that
or
what
are
you.
C
It
makes
them
scratch.
Well,
that's
okay.
I
mean
you
know
that
like
I've
done
a
couple
of
them,
were
you
kind
of
go
to
different
sources
on
the
way
with
you
yeah?
So
you
you
put
you
find
out
when
the
events
took
place,
you
put
them
on
the
on
the
map
and
the
idea
is
they
have
like
a
collection
of
points.
You
know
data
points
in
your
history
and
you,
you
know
you
can
element
it
by
putting
in,
like
you
know,
key
references
for
each
milestone.
That's
well
sometimes
good
yeah.
C
There
are
all
sorts
of
things
you
can
do
there.
I'm,
not
yeah
I
mean
so
it's
it's.
Whatever
is
helpful
to
you.
I
think
is
a
good
idea,
so
I.
Actually
today,
I
was
gonna
present
on
game
theory
and
modeling,
but
I'm
not
going
to
do
that
this
week,
because
when
they
couldn't
make
it
and
yeah
so
I
was
just
gonna.
Wait
on
that
for
next
week,
I
I
hope
everyone
I
know.
Everyone
saw
the
blog
post
by
now,
the
one
that
we
put
up
for.
C
C
C
Came
up
with
this
topic:
pre
train
models
for
developmental
biology,
so
pre-trained
model
is
a
model
that
I
had
mentioned
in
the
paper
where
you
can
create
a
model,
that's
less
specific
to
a
certain
type
of
data.
It
could
be
like
a
feature
like
a
rectangle
or
it
could
be
like
features
like
a
sphere
or
could
be
like
words,
and
things
like
that,
so
you
have
a
benefit
of
having
features
that
you
know
are
similar
to
the
ones
you
want
to
extract
from
your
data,
but
they're.
C
C
I
mean
this
isn't,
like
you
know,
an
AI
where
you
might
teach
it
a
language
but
or
like
phonemes,
or
something
like
that,
but
it's
just
kind
of
a
something
where
you
already
know
kind
of
what
you're
looking
for,
and
so
we
talked
about
some
of
these
things
in
here
we
talked
about
pre
tree
models
and
we
talked
then
we'd
go
in.
In
this
part.
We
talked
about
a
vision
for
a
developmental
biology,
specific
model,
so
in
this
case
we're
thinking
about
developmental
biology
and,
like
you
know,
embryos
and
cells.
C
C
So
we
don't
get
into
a
lot
of
technical
detail
in
this
post,
but
if
we
do
mention
the
stuff
bass,
hilarious,
stuff
and
the
C
elegans
stuff
in
terms
of
the
projects
a
summer-
and
so
you
know
just
kind
of
like
discussing
a
little
bit
about
the
challenges
there
and
then
you
know
really
I
mean
we
don't
really,
then
there's
an
advertisement
for
the
diva
worm
ml
group
which
were
know
in
but
I.
Don't
we
didn't
really
propose
any
specific
type
of
pre
train
model.
That's
maybe
sort
of
for
next
summer.
C
C
You
have
to
put
our
division,
then
you
have
to
actually
implement
it,
which
is
much
harder
but
I
think
that's
a
I
think
it's
a
good
way
to
sort
of
merge
the
two
areas
you
know,
I,
don't
really
think.
There's
a
lot
of
work.
That's
been
done
in
developmental
biology
and
machine
learning.
You
know
other
than
maybe
just
applying
it
to
problems,
but
this
is
kind
of
a
new
thing
that
we're
kind
of
proposing
here.
C
Let
me
give
you
a
link
to
the
blog
post,
so
this
is
the
it's
also
posted
do
another
blog,
but
I
cross
posted
it
here
maximize
the
impact.
So
this
is
the
blog
post
for
Thomas,
for
you
know,
I
know
Richard
and
Jesse
have
seen
it,
but
and
so
yeah
I
mean
there's
a
definitely
it's
just
kind
of
proposing
a
big
picture
view
of
pre-training
models
and
then
how
we
might
implement
one
for
developmental
biology.
C
But
we
haven't
given
much
thought
to
that
other,
like
the
technical
details
of
it
and
then
the
other
thing
is:
is
that
want
to
do
one
more
blog
post
by
the
end
of
the
year
like
by
the
end
of
the
calendar
year
and
so
I?
Don't
know,
probably
in
the
next
couple
weeks
we'll
talk
about.
Maybe
what
that
might
be.
I
wanted
to
talk
a
little
bit
about
game
theory
too,
and
I
kind
of
draw
out
some
issues
in
that
area
and
I
think
it's
actually
in
doing
research
for
that.
C
It's
really
interesting,
some
of
the
stuff,
that's
going
on
so
row
kind
of
a
short
version
of
it.
Last
week
we
were
talking
about
models
competing
with
one
another
so
like
in
in
the
gans,
we
had
a
generative
model
that
was
generating
fake
samples,
and
then
we
had
data
being
classified
by
a
classifier,
and
there
was
a
competition
between
the
two
deep
learning
model.
To
sort
of
you
know
maximize
the
parameters
you
know
is
it
that
your
classifier
is
being
able
to
you
know.
C
The
idea
here
is
that
you're
you're
you
have
data
that
you're,
classifying,
so
I
might
have
a
data
set,
that's
very
clear
and
very
simple,
where
I
have
very
clear
visual
features
and
your
classifier
is
extracting
those
features
and
saying
whether
they're
you
know
one
thing
or
another,
and
then
your
your
generator
is
generating
adversarial
examples.
So
that
means
it's
generating
things
that
are
similar,
but
not
really
the
same
thing,
and
so
the
idea
is
that
that
classifier
then
K
also
takes
a
meta
information
and
has
to
correctly
classify
those
adversarial
examples.
C
Now
sometimes
those
adversarial
examples
are
real
or,
like
you
know,
different
versions
of
the
stimulus,
maybe
at
an
angle
or
rotated
and
sometimes
they're,
just
nonsense.
So
then
you
know
it's
the
work
of
the
classifier
to
get
all
that
right,
and
so
there's
a
competition
going
on
between
those
two
models
and
that's
where
the
game
theory
comes
in,
and
it's
really
interesting
of
people
of
applied
game
theory.
But
at
the
same
time
it's
a
very
open
area.
There
hasn't
been
a
lot
of
work
done
on
like
really
understanding
like.
C
What's
possible,
people
apply
the
concept
of
the
Nash
equilibrium,
which
is
where
you
have.
You
know
you
have
an
equilibrium
where
you
can't,
like
you
know,
come
up
with
a
better
strategy
to
defeat
your
enemy.
So
you
have
to
stick
with
a
certain
set
of
strategies.
Each
player-
and
you
know
they
can't
ever
improve
their
position
in
the
game
by
using
another
strategy.
C
So
they
fall
in
at
this
equilibrium
and
it's
really
interesting
if
people
apply
it
to
performance
and
models,
because
I
mean
that's
something
that
I
wouldn't
have
thought
of
before
I
heard
of
games
like
something
I
want
to
come
up
with,
maybe
in
my
own,
but
eventually
but
I,
don't
think
it
would
have
thought
about
it
and
said:
oh,
that's
an
obvious
thing
that
you
would
have
in
that
area.
So
I'll
give
that
talk
next
to
me
and
now
I,
don't
wanna!
C
I
didn't
want
to
do
it
today
and,
let's
see
I,
don't
know
if
there's
anything
I
want
to
talk
about
this
week,
but
I've
related
to
this
I
think
I'll
save
it
for
next
week.
So
next
week
we'll
have
that
presentation
and
maybe
we'll
consider
what
we
might
do
for
another
blog
post.
First
of
all
should
ask:
does
anyone
have
any
ideas
of
their
own
about
what
they'd
like
to
do
for
a
second
blog
post.
C
We
can
also
do
something
more
biologically
oriented.
I
was
thinking,
you
know,
you
know
we
didn't
do
in
the
pre-training
model.
Blog
post.
We
didn't
really
have
a
lot
of
biology
in
it.
Actually
I
wish
I
would
have
had
some
screenshots,
but
I
didn't
couldn't
really
find
anything
suitable,
but
we
could
you
know
we
could
do
something
more
along
the
lines
of
you
know
comparing
or
classical
methods,
maybe
something
if
we
get
the
paper.
C
You
know
there
are
people
reading
their
article.
If
you
do
something,
that's
a
little
bit
more,
you
know
less
specialized
in
terms
of
the
writing,
so
I
mean
that's
sort
of
why
I
started
blogging
in
the
first
place
was
to
like
it
have
a
outlet
where
it
wasn't
as
technical,
whereas
it
still
ends
up
being
pretty
technical,
but
that's
a
possibility
too.
You
know
and
you
don't
have
to
decide
now,
but
you
know,
if
you
can.
C
Let
me
know
soon
just
contact
me
email
slack,
whatever
we
can
talk
about
it,
some
more
am
I
getting
any
responses
from
readers
of
the
note.
I
haven't
gotten
any
like
responses,
so
we
had
the
the
guy
from
China,
but
he's
you
know,
I
mean
I,
don't
know
where
he
learned
about
that.
It
sounds
like
he
may
have
read
that
blog
post,
but
I,
don't
know
other
than
that.
I
haven't
gotten
much
response
from
people.
It's
you
know,
I,
think
you
know
it's
hard
to
a
lot
of
people
in
developmental
biology.
C
They
don't
really
have
a
lot
of
background
in
it.
So
it's
like
our
you
know.
Maybe
we're
pitching
it
a
too
high
a
level
for
people
for
it
to
be
useful
to
a
lot
of
people.
That's
all.
Maybe
an
intermediate
post
would
be.
Maybe
you
know
more
received
on
in
that
community
and
when
I
say
intermediate
post.
That
means
something
more
basic
like
you
know
how
to
how
to
apply
an
algorithm
to
some
data
but
I.
You
know,
I,
don't
know
how
deep
in
the
weeds
I
mean.
C
C
Like
do
people
really
have
the
background
to
get
into
something?
That's
where
are
they?
Are
they
interested
so
sometimes
a
lot
of
people
if
they're
in
a
certain
area
will
think
that
things
aren't
really
relevant
to
them?
Like
sometimes
you
know,
they're
new
techniques
that
come
out
that
people
say
wasn't
really
relevant
to
me
so
I
mean
there's
a
case
for
making
something
relevant
to
people,
but
I,
don't
know.
I
haven't
had
any
responses
from
the
node
post.
C
So
it's
hard
to
know
really
what
how
people
find
it
useful
or
if
they
maybe
they
do
find
it
useful.
Maybe
it
takes
a
while
for
them
to
figure
that
out
like
you
know
they
send
it
to
their
friends
and
then
they
send
with
their
friends.
And
then
someone
says
oh
yes,
this
is
a
great
article
or
a
great
blog
post
see
have
any
other
chat.
A
C
Get
some
basic
analytics,
probably
between
the
two
posts,
maybe
50
or
60,
which
isn't
doesn't
sound
like
a
lot,
but
it's
not
a
it's
a
specialized
area,
so
really
good
yeah.
It
usually
takes
a
while
for
people
like
I
said
to
get
it
to
start
reading
it
like
you'll
get
some
hits
and
it's
usually
like
you
know,
the
calculator
read
is
like.
If
you
stay
at
the
site
for
more
than
three
or
four
minutes,
people
are
reading
part
of
it,
maybe
but
they're
also
sending
it
on
to
their
friends
or
colleagues.
A
Just
as
morale
of
the
visitors
of
our
more
or
less
scientific
website,
responding
to
the
author
and
F
the
same
problem
with
my
website,
entire
films
I
know
that
a
lot
of
people
coming
to
this
site
and
stay
for
a
while,
but
just
really
a
few
giving
some
response
is
quite
Romans.
Pretty
homogeneous
yeah.
C
Sarah
Richards
that
I've
occasionally
posted
on
node
with
no
response.
Yeah
I,
don't
know
if
people
are
really
again.
It's
really
depends
on
the
platform.
Some
and
I've
gotten
like
responses
on
things
just
by
email.
But
it's
you
know
it's
like
people
will
read
it
and
then
they'll
have
to
go
and
respond.
You
know
like
it's
an
extra
step
for
people,
so
sometimes
they'll
respond
and
sometimes
they're
really
interested,
but
they
don't
okay.
So
richard
has
a
link.
C
C
C
The
portal
well
yeah,
we
could
use
a
portal.
I
would
imagine
an
advanced
methods
of
analysis
in
embryos
yeah,
but
yeah.
That
might
be
a
good
idea
to
I.
Think
you
know,
like
I
was
showing
Jesse.
We
have
like
a
sort
of
basic
reading
list,
but
also
I.
Think
a
portal
would
be
useful
too.
If
people
are
going
to
you
know
if
you
want
to
have
a
like
a
clearinghouse
of
methods,
and
then
you
have
to
decide
like
who
do
you
want
to
appeal
to
you
know
this
is
the
thing
about
open
worm.
C
We
get
people
who
are,
you
know,
kind
of
casual
learners.
You
get
like
people
with,
maybe
an
undergraduate
level
background,
or
maybe
they're.
Even
you
know
high
school
students
or
something,
and
then
you
get
people
who
are
like
hardcore
researchers,
and
so
you
have
these
two
groups
that
you
have
two
better
interested
in
these
resources,
but
of
course,
they're
gonna
have
to
pitch
differently
and
so
I,
like
the
portal
idea,
I
mean
they're,
probably
I,
don't
know.
Who
would
would
that
be
something
that
we
would
just
point?
C
C
Yeah,
it's
a
like
you
know
with
dirt
with
our
group
we
get
a
lot
of
computer
scientists
and
they
don't
really
know
that
much
about
biology.
So
that's
always
the
challenge.
But
then
you
also
get
biologists
who
want
to
know
like
you
know,
because
you
do
have
tools
like
the
like
the
atlases
and
things
like
that.
That
are
digital
and
they
want
to
you
know
kind
of
link
it
to
their
work,
and
so
it's
it's
always
a
challenge
to
get
to
satisfy
both
groups
of
people
but
yeah,
so
I
think
that's
I.
C
F
B
B
B
B
The
only
thing
is
that
this
was
a
model
for
a
machine
cells,
realize
it
and
therefore
measure
the
pressures
and
the
forces
between
cells
epithelium
from
the
shapes
of
the
Sun
from
the
curvatures
of
the
boundaries
between
the
Sun.
So
it's
sort
of
like
it's.
You
have
to
segment
the
images,
but
also
get
the
curvature
at
each
boundary
between
each
curve.
Neighboring
cells,
okay,
so
this
is
not
pretty
done
and
these
guys
is
going
to
be
very
nice
chocolate.
B
So
this
is
an
it's
an
example
of
the
way
sophisticated
analysis,
not
the
best
image
segmentation.
Finite
element
analysis
there
any
two
dimensions:
it
has
to
be
two
three
dimensions
and
then
working
with
the
Susan
Crawford
game,
who
has
developed
a
flipping
microscope
which
allows
one
to
see
the
whole
circus
of
an
excellent
okay,
so
you'd
be
pulling
this
stuff
together.
B
F
B
B
B
C
B
C
B
C
B
With
okay
in
the
French
flag
model,
we
assume
the
30s
one
dimensional,
like
a
word
with
an
extremely
thin
work,
says:
okay,
you
see
that
there's
a
source
of
some
chemical
at
one
end
and
a
sink
of
the
other
end.
So
an
established
gradient
winces
reaches
steady
state
and
then
we
assume
that
the
cells
respond
to
a
register
to
a
range
of
concentrations
and
change
into
different
kinds.
B
So
in
this
case
the
first
third
of
the
cells
would
turn
one
color
I
forgot
the
order
of
the
colors
of
the
French
flag
returned,
let's
say:
blue
the
next
one,
whatever
okay
I
mean
in
this
red
shirt,
so
this
cell.
So
the
model
is
that
there's,
a
gradient
set
up
somehow
usually
not
explained,
and
the
cells
respond
to
differently,
depending
on
the
body,
the
gradient
episode
and
then
they
change
class,
disability,
okay,
okay,
a
differs
from
the
training
level,
which
we
also
discuss
in
that
paper,
the
Turing
level.
B
You
need
at
least
two
chemicals
which
interact
with
each
other
and
set
up
waves,
and
then
the
cells
respond
to
the
concentration
of
one
of
the
chemicals
in
waves,
okay,
so
the
source
and
sink
there's
no
source
and
sink
at
one
end
of
the
other.
In
the
turn,
rod,
okay
and
the
differentiation
won't
modest,
entirely
different
paper
where
you
could
propagate
close
to
cells
and
there's
the
wave
propagates
the
change
back.
B
C
B
D
B
C
There
are
a
lot
of
well
there's
a
lot
of
like
references
like
like
the
worm
base.
I
was
showing
that's
got
a
series
of
papers,
there's
like
worm
book.
It's
a
worm
base
is
like
you,
look
up
a
gene
and
you
find
out
information.
One
book
is
where
you
have
specific
articles,
so
I
mean
there's,
there's
a
lot
of
stuff
for
C
elegans,
but
the
way
it's
set
up
is
usually
like.
Almost
like
an
archival
resource.
C
Where
you
have
a
you
know,
a
bunch
of
information,
you
look
it
up
as
like
a
neophyte,
you
might
say:
what
does
this
gene
do?
What
does
a
cell
do
and
then
it
gives
you
the
information,
so
I
mean
that's
the
way.
It
would
probably
the
best
way
to
set
it
up,
although
you
could
set
it
up
in
other
ways
like
just
like.
You
know,
a
tree
structure
where
you
kind
of
like
follow
things
down.