►
From YouTube: DevoWorm (2021, Meeting 29): GSoC #10, Developmental Game Theory, Landscapes, and Liquid Crystal Bio
Description
Update on Summer of Code activities (Mainak Deb, Improving DevoLearn), Review of submissions document. Semantic Versioning for Open-source Software. Game Theory of Developmental Processes preview. Developmental Landscapes for C. elegans, Liquid Crystal Biology (LCB), and Biological Symmetry-breaking. Attendees: Bradly Alicea, Krishna Katyal, Susan Crawford-Young, Akshay Nair, and Mainak Deb.
B
A
D
A
E
A
C
Okay,
okay,
so
I'll
get
started
with
weak
lines,
update
all
right,
I'll,
just
go
ahead,
and
I
just
square
when
I'm
opening
the
weekly
blog
post.
Okay.
So
so,
can
you
see
the
blog
post?
Obviously,
yes,
okay!
So
the
first
thing
that
I
did
in
this
week
is
that
I
actually
had
to
replace
the
static
images
on
the
web
app
with
the
interactive
images.
So
what
these
interactive
images
would
do
is
that
it
will.
It
will
actually
give
the
freedom
to
the
user.
C
C
D
C
G
C
D
D
C
F
C
E
C
C
D
D
C
C
C
D
C
F
C
C
A
C
A
You
can
get
maybe
a
bit
more
detail,
but
you
also
get
more
noise
and
it
depends
on
the
image
and
then
a
higher
threshold.
You
exclude
a
lot
of
maybe
signal,
but
you
definitely
exclude
noise.
So
that's
that's
something
people
can
play
around
with,
and
it's
good
that
that
that
we
have
that
that's
numeric,
so
that
people
can
report
what
the
actual
threshold
value
is
when
when
they
find
the
thing
that
they
want
and
then
they
can
download
the
images
yeah
yeah.
A
Yeah
thanks
yeah
yeah,
and
so
that
should
that
should
give
people
if
they
want
to
do
further
analysis,
they
can
do
it
in
image
j,
which
would
allow
you
to
do
some
things
with
counting
cells
or
whatever.
So
that's
that's
something
we
can
do
like
that.
Yeah
I
like
the
interface
and
then
so.
The
new
version
is
coming
out
next
week,
yeah,
okay,
so
that's
that'll
be
what
what
version
number
do
you
think
that'll
be.
A
C
A
A
C
A
So
yeah,
that's
okay,
yeah,
that's
good
and
it's
usually
like.
If
you
have
like
a
s,
you
have
to
have
a
static
location
on
the
web.
So
what
they'll
do
is
when
they
get
the
submission
they'll
put
it
they'll
open
up
a
link
that
you
give
them
and
then
that
link
should
have
everything
listed
like
a
like
a
description
of
the
project
and
what
you
did
and
and
then
it
has
a
like
a
link.
A
It
basically
they're
going
to
download
it
they're
going
to
try
to
run
it,
and
then
you
know
they're
going
to
see
if
it
runs.
So
if
everything
runs
and
you
pass
and
you
know
I'll,
do
the
evaluation
which
is
not
like
you
know,
that's
fine,
and
then
you
know,
that's
so
just
there's
some
format
issues
you
might
want
to
be
aware
of.
A
I
maybe
I'll
try
to
send
you
some
examples
from
years
past
of
how
people
have
done
this,
but
basically,
if
you
just
use
a
readme
file
like
in
a
static
repository,
that's
fine
people
have
used
just
in
the
past,
but
you
don't
have
to
do
that.
You
can
just
use
like
something.
That's
stable
and
then
you
know
just
like
a
quick
write-up.
A
You
know
just
basically
describing
your
your
experience
a
little
bit
of
your
experience,
highlighting
it
your
process,
maybe
some
of
the
things
you
didn't
get
to
do
or
some
of
the
things
you
really
are
proud
that
you
did
and
then
the
link,
and
so
it
doesn't
have
to
be
that
long,
but
it
has
to
be
like
something
that
they
can
just
kind
of
quickly
go
through
and
then
make
sure
you
know
the
deadlines.
You
know
they're
pretty
tight,
they're
pretty
hard
on
the
deadlines.
A
So
I
think
that's
next
week,
I'm
not
sure
the
exact
date.
I
got
an
email
about
it
this
weekend.
That's
why
I
bring
it
up:
yeah,
the
16th
so
yeah.
So
it
looks
like
yeah.
He
did
a
pretty
good
job
on
that
and
it
looks
like
we've
got
something
so
we'll
have
to
once.
We
you
know
once
the
summer
code
officially
ends.
A
A
G
A
A
C
F
A
So
there's
a
semantic
versioning.
This
is
their
version
of
it,
and
so
then
you
get
the
major
minor
patch
system
here.
So
you
increment
the
major
version,
which
is
the
first
number
when
you
make
incompatible
api
changes.
So
that's
like
zero
one
to
minor
version.
When
you
add
functionality
in
a
backwards
compatible
manner,
that's
the
second
number
and
then
the
patch
is
when
you
make
backwards,
compatible
bug,
fixes
and
then
that's
basically
it,
and
that
just
makes
it
everything
like
consistent,
so
a
release.
A
A
So
that's
what
the
the
one
two
and
three
are
the
three
they're
not
really
bug.
Well,
I
guess
they
might
be
bug
fixes,
but
they're
minor
fixes
that
aren't
really
adding
a
lot
of
functionality.
It's
you
know.
In
any
case,
I
think
I
think
yeah
we
can.
We
can
adopt
something.
I
mean
we're
kind
of
doing
something
similar
to
that.
We
can
kind
of
adopt
this,
keep
this
in
mind
when
we're
making
the
versions,
so
you
know
adding
functionality.
Of
course
we
have
our.
A
A
A
A
D
D
D
D
D
B
D
A
Well,
good!
Luck
with
that
yeah!
It's!
It
sounds
like
it's
a
an
exciting
area.
I
didn't
know
you
know
I
don't
know
like.
Are
you
just
trying?
Are
you
trying
to
come
up
with
examples?
Is
that
the
best
way
forward,
or
is
it
like
really
kind
of
making
the
the
process
simple
like
what?
What
does
it
do
and.
D
One
professor
by
dr
sharif
he's
egyptian.
He
wants
to
change
his
equipment
around,
so
it
is
new
and
novel,
so
he
wants
new
and
novel
optical
coherence
tomography
that
he
wants
to
to
work
with
this.
It
has
to
be
different
from
everybody
else's
in
the
world.
This
is
what
he
wants,
and
the
other
guy
just
wants
to
check
out
the
mechanical
properties
of
things,
and
I
suggest
we
use
insects
as
a
subject.
D
D
A
D
A
D
E
B
B
B
B
A
A
For
the
deep
learning-
and
you
know
it's
sponsored
by
a
bunch
of
groups
so
last
year
they
started
the
computational
neuroscience
course,
and
it
was
kind
of
by
the
seat
of
the
pants
effort.
You
know
during
the
pandemic
and
this
year
they've
got
a
lot
of
sponsors,
so
this
is
nice
that
they've
been
able
to
pull
together
this
many
sponsors
in
this
period
of
time.
It's,
but
this
is,
you
know,
like
the
idea.
Is
you
want
to
have
a
summer
school?
That's
sort
of
anyone.
A
All
right
thanks,
so
this
is
yeah.
They
have
a
bunch
of
sponsors
here
this
and
the
idea
is,
you
know
this
is
a
summer
school
that
anyone
can
join.
You
know
and
they
want
to
make
it
accessible
to
people
as
many
people
as
possible.
So
I
think
this
year
there
are
a
lot
of
people
who
don't
really
know
a
lot
about
deep
learning
who
are
doing
this,
like
you
know,
people
in
in
neuroscience
and
maybe
in
bio
other
areas
where
they
want
to
get
this
introductory
level
education.
A
So
this
is
the
welcome
video
check
that
out
concept
map,
which
is
something
that
jesse
from
the
group
is
really
into
into,
and
so
this
is
just
basically
describing
you
know
the
different
things
you're
going
to
learn
here.
So
you
have
this
deep
learning.
You
have
the
basics,
which
are
the
technical
details.
A
You
have
the
mathematics
that
includes
data
and
you
have
advanced
methods.
So
these
are
things
that
are
more
advanced,
like
you
know,
reinforcement,
learning
and
continual
learning,
and
things
like
that,
and
then
this
doing
more
with
fewer
parameters
is
kind
of
like
the
optimization
part.
So
this
is
where
you
have
attention
and
transformers
and
honoring
covenants,
and
you
know
once
you
learn
the
basics.
A
You
move
to
some
of
the
advanced
methods
and
then
you
say:
well,
okay,
I
have
the
basics.
How
do
I
make
this
better?
How
do
I
make
these
models
better?
And
so
you
just
learn
all
these
and
then
you
have
to
have
also
a
foundation
of
mathematics
which
also
helps
you
understand
like
these
more
advanced
areas.
A
So
that's
I
like
that
that
sort
of
layout
of
the
course
and
then
they
have
the
links
here-
the
schedule,
the
technical
help.
They
have
a
lot
of
volunteer
labor
on
this.
So
there
are
a
lot
of
people
from
like
different
programs
and
in
deep
learning,
different
pro
computer
science
departments
and
so
forth,
who
are
interacting
with
the
with
the
group.
A
So
you
know
there's
this
whole
syllabus
here
of
resources.
So
you
know
you
have
like
optimization
regularization
the
basics
and
then
you
like,
okay,
so
there's
like
the
basement,
reinforcement
learning,
for
example,
you
know
you
have.
I
don't
think
they
have
anything
right
now
in
well.
They
have
the
source
file
here.
So
you
can
download
it
this
basically
links
to
their
github
repository,
so
you
can
just
get
things.
I
don't
know
if
I
don't
think
this
has
been
made.
Yet,
let's
go
to
the
first
week.
A
Okay,
yeah
that's
great
yeah,
and
this
is
for
every
lesson,
so
you
get,
you
can
go
walk
through
it
at
your
own
pace
and
we've
talked
about
you
know
with
the
summer
with
the
computational
neuroscience
school.
It's
like
this
is
a
very
condensed
course,
as
you
can
see
like
it's
just
a
couple
weeks
and
you
could
spend
like
a
year
on
this,
be
honest
with
you.
A
So
one
of
the
things
that
they
did
last
year
for
the
computational
neuroscience
is
that
what
they
call
slow
pod,
which
is
where
they
organize
the
the
groups
into
pods,
and
they
do
like
a
you
know,
a
slow
pod
which
is
like
slowing
down
this
whole
curriculum
and
stretching
it
out
over.
Maybe
about
three
months
where
you
know
people
meet
in
small
groups
and
they
kind
of
work
through
the
problems
and
they
kind
of
think
about
it.
A
And
you
know
they
do
it
at
their
own
pace
and
we've
talked
about
maybe
doing
one
maybe
like
jesse
had
talked
about
it,
but
I
don't
I
they
never
did
they
didn't
do
one
last
year
you
know
it's
something
that
you
know
it's
just
up
to
the
people
doing
it
if
they
want
to
how
they
want
to
go
through
it.
But
the
point
is
you
can
do
these?
A
A
You
know
maybe
in
the
fall
or
something
to
go
through
one
of
these
and
talk
about
like
how
the
method
works
so
anyways,
the
the
link
surety,
put
the
link
in
the
chat
here
and,
if
you're
interested
check
it
out.
Maybe
if
you're
interested
in
doing
like
a
slower
version
of
some
of
these
tutorials,
let
you
know
let
us
know,
maybe
we
can
go
through
some
of
them.
One
week,
maybe
you
know
something,
you
know
something
that
gives
us
a
little
bit
more
insight
into.
What's
going
on
so
and
it's
you.
B
D
A
A
A
We
have
these
basic
models
here.
It's
basically
the
basic
mathematical
models
and
data
structures
of
what
we
use
in
the
group
of
development
or
trying
to
boil
it
down
to
some
of
these
basic
models,
and
then
something
like
this,
where
you
have
a
little
bit
more
advanced
models,
things
that
we've
put
papers
out
on
in
the
group
and
this
category
theory
thing:
I
need
to
go
back
and
look
at
one
of
the
papers
and
see,
look
and
put
it
down
a
very
simple
diagram.
D
A
You
know
how
do
those
intersect
and
how
do
we
use
them
in
the
group
and
then
this
one
here
with
game
theory
and
differentiation
trees
using
a
computational
agent,
those
things
can
be
described,
and
so
that
would
be
the
idea
there.
So
that's
coming
along.
We
have
these.
I
like
the
test
of
williamson
symbiosis,
that's
a
project
that
involves
some
bioinformatics
of
looking
at
different
gene
sequences.
A
A
A
A
You
know
they're,
usually
pretty
short
and
they're
not
published
in
the
proceedings,
but
they
give
you
some
exposure.
Your
work,
some
exposure,
so
there
you
know
a
number
of
different
workshops.
You
can
look
at
here
and
maybe
check
out
if
there's
something
that
catches
your
attention,
maybe
they
have
a
link
to
the
website
here.
Maybe
you
have
to
go.
Look
it
up
on
google,
but
you
know
you
can.
D
A
Things
that
we
have
outstanding,
we
have
a
couple
other
things
like
the
kindle
book
that
are
kind
of
you
know
in
waiting
in
the
wings.
So
those
are
all
things
you
know
we
just
want
to
keep
in
the
front
of
our
minds
and
then
I'm
going
to
talk
about
this
talk,
I'm
giving
in
a
couple
weeks
I'm
going
to
walk
through
it
really
quickly.
A
D
A
Is
a
theme
we've
been
kind
of
working
on
in
the
group
for
years,
and
it's
been
like,
you
know
it's
kind
of
like
here
and
there
we
have
maybe
published
a
couple
of
papers
and
you'll
see
in
the
slides
what
those
papers
are
and
we've
talked
about
it
a
little
bit,
but
never
really
gotten
anywhere
with
it
in
a
unified
way.
So
this
isn't
something
that
people
this
isn't
like
a
field.
This
is
just
some
idea
that
we
have
about
like
you
know
how
to
apply
game
theory
to
developmental
problems
and
development.
A
So
you
know
so
this
could
be
improved
upon
and
it's
in
a
session,
that's
kind
of
interesting.
It's
like
a
bunch
of
dynamical
systems,
people
and
a
bunch.
You
know
a
bunch
of
different.
It's
a
grab
bag
of
different
of
different
presentations,
so
it'll
be
an
interesting
session.
It
weirdly
fits
together
too.
It's
like
this
session,
where
you
have
all
these
talks
that
kind
of
fit
together.
I
don't
think
they
were
meant
to
be
together,
but
they
seem
to
work
at
least,
hopefully
we'll
see
what
the
session
is
like
in
reality.
A
But
this
is
the
game.
Theory.
Developmental
processes,
so
the
idea
here
is
the
three
things
that
we
want
to
talk
about
here:
we're
talking
about
variation,
morphogenesis
and
learning
and
adaptation.
So
this
is
a
picture
of
claude
shannon
and
his
thesis,
which
is
an
artificial
agent
that
goes
around
a
maze
and
the
agent
you
know
kind
of
thinks
about
where
it's
going
or
there's,
maybe
some
intentionality
there,
but
it's
a
machine,
so
it
has
to
do
its
artificial
learning
of
the
maze.
A
A
Finally,
this
is
a
computational
model
of
a
cellular
automata
of
a
pattern
and
again
that
model
has
to
learn
and
it
has
the
same
constraints
as
the
egg
and
it
shows
a
lot
of
the
same
variation
as
the
egg,
but
it
also
has
features
of
this
kind
of
learning.
So
how
do
we
kind
of
move
stitch
all
this
together?
A
So
the
agents
are
like
the
units
of
analysis,
units
of
like
the
player
or
the
agent
and
they're
emitting
these
states,
and
the
states
are
drawn
from
this
distribution
of
strategies
that
they
can
play,
and
so
it's
you
know
it
seems
a
little
bit
anthropomorphic
because
you
know
you
say
well,
surely
humans
can
do
this,
but
how
about
you
know
things
that
don't
have
a
consciousness
like
a
you
know
a
a
gene
or
a
cell.
You
know
so.
A
We've
talked
about
like
maybe
some
of
the
options
in
terms
of
non-neuronal
cognition,
but
that's
not
really.
What
we're
going
to
get
into
in
this
talk.
The
developmental
game
is
the
interaction
of
these
states
and
actually
maybe
I
should
change
this
a
bit
to
talk
about
like
the
emission
of
of
strategies
and
things
like
that.
But
I
don't
know
I'm
going
to
put
a
note
here.
A
I
just
you
know
this
gives
me
an
idea
of
what
I
need
to
change
in
the
slides.
So
I
talk
about
developmental
agents.
This
is
an
example.
You
have
this
agent,
which
is
a
single
like
sphere
and
then
there's
differentiation
of
the
sphere
and
then
there's
further
geometric
differentiation,
not
a
function,
which
is
this
case,
but
like
things
that
form
and
move
around
to
different
parts.
A
So
in
the
adult
you
have
this
thing
that
looks
almost
like
a
hockey
puck
or
a
something
like
that,
and
it
has
different,
you
know,
has
sensors
on
the
edge
and
it
has
a
brain
in
the
middle.
So
this
is
how
these
developmental
agents
behave.
All
these
cells
in
the
developmental
agent
are
interacting
and
taking
shape
and
moving
around
then
there's
this
idea
of
the
antigenetic
agent,
which
is
a
little
bit
different.
A
A
So
if
we
think
about
like
how
you
know,
you
have
a
process
that
branches
that
you
know
you
could
there's
a
sort
of
you
know
you
can
observe
like
an
egg
that
have
these
different
shapes
here.
Each
of
these
go
through
a
process
of
branching.
Where
there's
you
know,
there's
either
there's
one
state
or
another
that
it's
going
to
go
into
the
next
step.
A
So,
like
you
know,
the
eggs
will
form
based
on
their
interactions
of
the
environment,
and
so
given
that
that
trajectory
of
decisions
or
constraints
you
end
up
with
a
different
shaped
egg,
and
so
the
the
anti-genetic
agent,
then,
is
this
idea
that
you
know
you
can
take
these
things
and
they
form
sort
of
this
intentionality.
But
it's
not
like
intentionality
like
we,
you
know,
decide
to
go
to
the
to
the
park.
It's
this
intentionality
that
results
from
these
constraints.
A
A
This
is
biological,
intentionality
kind
of
going
through
this
a
little
bit
more
and
then
we
go
through
the
different
examples
of
zero
player
games,
which
are
these
like
game
of
life,
where
you
have
a
single
cell
that
can
either
be
dead
or
alive.
So
there's
a
cell
operating
a
point
process
and
it
can
either
be
dead
or
alive,
and
it's
playing
this
game
where
you
know
maybe
its
neighbors
are
trying
to
predict
the
state
or
maybe
it's
trying
to
regulate
its
own
state
and
so
that's
an
example
of
a
zero
player
game.
A
Then
this
is
payoff
matrix,
which
you
can
derive.
That's
a
good
way
to
show
how
these
different
strategies
give
a
payoff.
So
the
idea
behind
game
theory
is
that
the
agent
will
try
to
maximize
its
payoff
to
a
certain
value
so
like,
for
example,
if
the
payoff
for
being
alive
is
high,
then
the
agent
will
gravitate
towards
being
alive
just
simply
because
the
payoff
is
higher
so
depending
on
this
payoff
structure,
that
will
determine
the
behavior.
A
So
it's
actually
interesting
because
we
talk
about
human
choice
in
game
in
game
theory
applied
to
humans,
but
it's
really
about
the
payoff,
and
so
if
the
payoff
is
low,
you
tend
not
to
do
it,
and
so
it's
an
interesting
kind
of
approach
to
looking
at
that.
But
in
any
case,
in
the
biological
case,
we
don't
care
about
the
intention,
but
we
care
about
the
intentionality
in
the
sense
that
it's
sort
of
more
aligned
with
this
idea
of
being
determined
by
payoffs
and
the
payoffs
are
determined
by
energy
or
whatever,
and
so
asymmetric.
A
Payoffs
like
this,
as
opposed
to
the
payoff
being
like
even
up.
You
know
whether
you
know
say
like
the
payoff
for
being
alive
is
the
same
as
being
dead,
then
there's
no
real
preference
for
either
and
you
switch
between
the
two.
But
if
there's
a
high
payoff
for
being
alive,
you
stay
alive
and
this
asymmetric
payoff
is
determined
by
information,
so
that
could
be
biochemistry
or
global
fitness
and
then
so.
A
This
is
then,
the
one
player
games
which
are
the
games
against
nature,
which
are
these
things
where
you
know
you're
an
agent
that
has
to
predict
the
weather
and
you,
if
you
predict
it
correctly,
there's
a
higher
payoff
than
if
you
don't
so
that's
a
game
against
nature,
you're
playing
against
a
stochastic
process
and
trying
to
beat
it
and
it's
hard
to
do.
But
one
example.
A
So
you
have
this
sort
of
two-dimensional
genome
in
this
case,
where
you're
choosing
locations
and
you're
opening
up
like
a
possibility
space.
But
if
you
hit
the
wrong
place,
if
you
get
a
lethal
mutation,
then
that's
you're
dead.
So
this
is
the
kind
of
thing.
This
is
kind
of
a
contrived
model
that
I'm
proposing
here,
but
this
is
basically
the
idea
of
being
able
to
play
this
game
against
nature,
where
you're
trying
to
you
know,
get
as
much.
You
know
good
stuff
in
your
genome
without
dying
or
hitting
on
a
lethal
mutation.
A
D
A
You
keep
that
you
keep
doing
that
defensive
move
and
no
one
can
win
the
game.
So
there's
this
equilibrium
where
there's
always
a
tie,
and
so
this
is
the
idea,
though,
of
like
imagine
this,
instead
of
being
x's
and
o's,
this
being
like
spatial
locations
of
cells
in
their
position
in
the
embryo,
and
so
this
is
something
that
we
published
on
in
biosystems
a
couple
years
ago,
and
this
is
what
it
looks
like
so
you're
when
you
have
a
cell
division.
A
You
know,
there's
a
first
move,
there's
a
second
move,
there's
a
third
move
and
this
kind
of
partitions,
the
space
within
the
egg
or
the
embryo,
and
you
know
the
more
moves
you
make
if
they're
optimal
between
sort
of
these
these
sub
lineages,
then
you
end
up
with
this
sort
of
optimized
spacing
and
things
like
that.
So
that's
the
idea
here
and
so
the
idea
would
be.
You
could
actually
model
that
as
a
game
that
I've
just
described-
and
you
know,
come
up
with
different
arrangements
for
these
things.
A
A
Is
in
equilibrium,
if
the
we
just
did
it
with
the
two
two,
the
two
sub
lineage
case,
but
it
could
be
extended,
there's
also
this
this
first
player
dynamics
and
connectome
formations.
Another
paper
was
published
last
year
by
myself,
where
we
analyzed
synaptic
connectivity
in
c
elegans,
using
a
first
mover
model,
the
same
kind
of
example
as
I
showed
before,
but
this
one
actually
involves
connecting
cells
together
based
on
their
time
of
of
terminal
differentiation.
A
So
you
know
you
have
these
different
groups
of
cells
that
are
connected
together.
They
all
have
different
times
of
terminal
differentiation
and
the
ideas.
You
know
whether
the
cell
that
was
terminally
differentiated
first
drives
the
process
or
the
cell.
That's
terminally
differentiated,
and
you
know
in
a
second
drives
the
process
or
whatever,
and
so
you
know
I
didn't
put
every
it.
You
know
it's
kind
of
hard
to
describe
in
one
slide,
but
basically
the
idea
is
you
have
these
different
strategies
that
are
sort
of
these
different
coupling
strategies
that
enable
you
know.
A
A
first
mover
is
the
first
cell
that
emerges,
and
then
they
connect
to
another
cell,
and
then
they
connect
to
another
cell
and
so
on,
and
so
it's
either
driven
by
the
first
cell
that
emerges
or
it
could
be
driven
by
ladder
cells
that
emerge
and
so
there
you
know
we
can
analyze
the
data
in
different
ways
and
see
these
strategies
and
then
calculate
their
payoffs.
So
two
player
games
really
quickly.
This
is
the
prisoner's
dilemma.
A
A
These
are
iterative
payoffs
that
you
can
calculate
by
one
agent
following
another
and
tracking
them
and
reinforcement
learning
algorithms,
so
I
might
bring
that
up
in
the
talk
and
then
finally,
there
are
these
developmental,
stable
states
where
you
have
these,
what
they
call
epigenetic
landscapes,
that
kind
of
mimic
cannabization
and
you
can
model
those
as
a
game,
and
so
then
that's
the
whole
talk,
and
so
that's
yeah.
So
that's
the
whole
talk
and
I'm
gonna
work
on
that,
some
more
and
present
it
at
the
dynamics.
A
Days
conference-
and
I
don't
know-
I
think,
they're
gonna-
be
some
interesting
connections,
they're
still
kind
of
in
progress,
though,
and
then
the
session
will
I
hopefully
be
pretty
interesting.
I
don't
know,
but
this
is
a
conference
that
it's
you
know
a
lot
of
physicists,
a
lot
of
people
and
complexity.
Some
people
buy
out
biological
physics,
but
it's
not
it's
not
machine
learning.
It's
not
not
even
game
theory.
So
I
don't
know
people
will
probably
understand
it
though
hopefully
or
get
something
out
of
it
there.
A
A
Let's
see
there
we
go
so
there
are
some
interesting
ideas
in
here.
Some
interesting
papers
there's
this
one
that
I
found
yesterday.
So
one
of
the
last
slides
in
that
presentation.
We're
talking
about
epigenetic
landscapes
and
those
are
these
things
are
interesting
because
they
kind
of
describe
development
as
this
state
of
possibilities.
This
landscape,
where
you
know
you
start
at
the
top
and
you
roll
downhill
and
as
you
roll
downhill,
you
tend
to
go
down
to
certain
paths
and
those
paths
are
described
as
channels.
A
So
this
is
very,
you
know
it's
a
very
analogical
model
where
you
have
this
analogy
of
a
landscape,
but
that's
you
know
it's
really
kind
of
a
tree
where
you
know
there
are
different
decisions
being
made
they're
different
switches
that
exist,
so
you
switch
from
one
state
to
another
as
your
cells
differentiate
and
you
end
up
at
the
bottom
and
some
bin,
and
this
tends
to
happen
over
and
over
again
in
development.
So
this
paper
talks
about
the
waddington
epigenetic
landscape
in
the
c
elegans
embryo.
A
This
is
from
the
bio
archive,
so
the
abstract
says:
waddington's
epigenetic
landscape
provides
a
visual
model
for
both
robust
and
adaptable
development,
generating
and
exploring
a
waddington,
epigenetic
landscape
for
the
early
c.
Elegans
embryo
suggests
that
the
key
shapers
of
the
landscape,
which
is
the
things
that
form
these
channels
and
decide
you
know,
determine
how
exactly
the
one
cell
embryo
differentiates
and
goes
down.
This
landscape
are
genes
that
lie
at
the
nexus
between
stress
response
and
behavior
and
include
genes
that
are
regulated
by
transgenerational,
neuronal,
smaller
rnas
and
so
this.
A
A
So
you
have
these
signals
that
are
sent.
You
know
these
small
rnas
and
they
just
in
reproduction.
They
end
up
in
the
germline
and
then
they
end
up
affecting
these
generations
and
eventually
they
just
kind
of
wear
off
and
the
you
know
the
environmental
stimulus
has
to
be
experienced
again.
So
that's
what
they
mean
by
that.
Curiously,
several
genes
shape
the
early
landscape
of
one
lineage
and
then
pattern
differentiate
or
enriched
in
another
lineage.
A
So
that's
interesting.
I
don't
know
what
that
means.
Several
genes
shape
the
early
landscape
in
one
linea
of
one
lineage,
and
then
pattern
different
gene
enrich
in
another
lineage.
So
that's
interesting,
an
interesting
finding.
I'm
not
really
sure
what
the
consequence
of
that
is.
Additionally,
paralogs
of
similar
expression
profiles
contribute
differentially
to
shaping
the
model
landscape.
This
work
suggests
that
robust
embryonic
development
is
initialized
by
different
deployment
of
redundant
genes
by
transgenerational
cues,
and
so
this
is,
I
wanted
to
show
the
figure
what
this
looks
like
this
kind
of
goes
down
here.
D
A
A
figure
in
here-
okay,
this
is
it
here.
So
this
is
this
epigenetic
landscape,
where
you
have
the
ball
up
here
at
the
top,
and
it's
going
to
go
down
this
landscape
and
it's
going
to
end
up
down
one
of
these
channels,
which
is
the
common
mode
of
development.
So
it
survives
to
an
adult
successfully
and
then
these
things
down
here
are
supposed
to
be
genes
with
differential
expressions.
So
each
of
these
black
bars
and
then
these
lines
are
the
differential
expression.
A
This
all
underpins
the
landscape,
so
you
know
there
are
different
points,
switch
points
at
which
you
end
up
shifting
your
trajectory,
and
so
this
is
how
this
is
supposed
to
work.
This
is
a
diagram
from
like
the
50s,
so
you
could
probably
draw
this
better
with
like
modern
technology,
but
we'll
just
go
with
this
for
now.
So
this
shows
you
how
the
you
know
the
cell
lineage
goes.
You
know
in
c
elegans.
A
It
goes
from
the
single
cell
down
to
these
different
cells,
and
you
can
see
that
this
lineage
tree
is
organized
in
an
interior
posterior
fashion.
So
you
start
with
a
b,
and
it
goes
down
to
the
top.
P1
goes
down
to
the
bottom,
to
the
back,
and
then
the
germline
is
in
here
where
it
differentiates
and
it
forms
where
the
eggs
are
going
to
be
so.
This
is
this
is
an
example,
I
think,
of
the
landscape
they're
forming
so
they're
forming
this
landscape
out
of
the
different
lineages,
and
this
is
g.
A
Sub
g
is
where
you
have
the
epigenetic
landscape
of
the
early
embryogenesis
computed
from
the
tentoriadel
dataset.
So
they
take
a
data
set
that
exists.
They
compute
this
landscape
and
they're
using
wd
as
a
as
a
parameter
here,
and
so
this
is
like
so
there's
a
height
to
this
landscape.
There's
this
width.
This
is
the
anterior
posterior
axis,
and
then
this
is
developmental
time
down
here.
So,
as
you
know,
time
goes
on.
You
get
these
smaller
blocks
which
represent
sublimages
of
sublineages.
D
A
A
A
So
where
is
the
wadding?
I
don't
know
what
that
means
to
go
back
here,
not
really
sure,
but
they
anyways
they
have
a
pretty.
They
have
a.
They
have
a
quantitative
value
for
this.
So
usually
the
criticism
of
these
landscapes
is
that
they're,
not
quantitative,
and
they
don't
really
give
you
a
good
sense
of
what
the
what
the
quantitative
value
should
be,
and
so
this
actually
they're
coming
up
with
a
metric
for
this.
A
So
I'm
not
going
to
get
into
what
I'm
not
going
to
go
and
look
anymore.
What
a
wadding
is,
but
if
you
want
to
look
at
the
paper,
you
can
look
at
it
so,
okay,
so
I
guess
they
kind
of
describe
it
here.
I
computed
a
value
that
corresponds
to
rope
tension
from
each
gene
to
each
region
in
the
matrix,
the
tensioner,
based
on
gene
expression,
measured
in
transcripts
per
million,
then
each
gene
would
pull
on
each
region
in
the
matrix,
with
a
tension
equal
to
expression
level.
A
So
this
is
like,
where
you're
balancing
expression
between
the
different
categories,
however,
a
gene
expression
in
each
cell
normalized
to
1
million
transcripts,
so
they
can
count
these
number
of
rnas
in
a
sample
and
it's
actually
easy
to
do
with
the
modern
technology
that
we
have.
But
you
can
do
this,
and
I've
done
this
with
quantitative
pcr.
A
This
is
the
contribution
to
f
sub
et,
which
is
this
other
metric
that
they
use
so
they're
using
yeah
they're,
really
putting
a
quantitative
cast
on
this
whole
thing.
This
is
a
nice
paper
for
understanding,
c
elegans
development
specifically,
but
also
how
you
might
you
know,
quantify
these
landscapes
and
give
us
an
idea
of
how
these
things
work
in
development.
A
So
I
mean
you
know
again.
This
is
something
that
may
or
may
not
may
or
may
not
be
right.
I
don't
know,
I
don't
think,
there's
really
an
answer,
whether
it's
definitively
right
or
not.
But
it's
it's
a
nice
approach
to
this
so
they're
using
epigenetic
tension.
That's
basically
the
approach,
so
I
think
I'm
gonna
go
to
this
one
here
and
I
was
gonna
talk
about
another
thing.
A
But
oh
maybe
I
will
that's
a
few
papers
in
here
on
this
so
last
week
I
think
susan-
and
I
talked
about
this
idea
that
you
have
you
know
you
can
have
like
we're
talking
about
gene
expression,
we're
talking
about
pattern
formation
and
she
says
well
yeah.
You
can
get
the
same
thing
just
with
physical
phenomena,
and
I
said
it's
true.
You
can't
and
so
there's
this
area
of
science,
where
they
look
at
different
types
of
liquid
crystal
physics
and
there's
a
book
by
dejanae
on
liquid
crystals.
A
And
if
you
go
to
that,
I
know
it
sounds
kind
of
removed
from
biology,
but
liquid
crystals
are
actually
a
physic
area,
physics,
physics,
research,
where
they
study
these
liquid
crystals
that
are
partially
ordered
materials,
intermediate
between
conventional,
solid
and
liquid
phases.
And
so
the
reason
why
that's
important
is
because
you
can
take
some
of
the
principles
you
learn
from
liquid
crystals
and
you
can
actually
look
at
things
that
are
biological.
A
There
are
some
unique
physical,
optical
and
rheological
properties
of
liquid
crystals,
and
we
know
that
they're
actually
ubiquitous
in
daily
life,
so
some
clay,
soaps,
human
dna
cell
membranes,
polymers,
elastomers
and
other
things
behave
like
liquid
crystals.
We
actually
know
that,
like
some
bacteria,
when
they
behave
collectively
also
kind
of
behave
like
liquid
crystals.
Although
that's
one
of
these
things,
where
you
know
you
put
a
model
onto
a
system
and
you
kind
of
make
some
assumptions
about
it,
but
liquid
crystals
can
be
classified
in
a
number
of
ways.
A
You
can
have
pneumatic
liquid
crystals,
which
are
thicker
like
where
they're
aligned
in
a
certain
way,
cholesteric
liquid,
crystals
and
symmetic
liquid
crystals.
So
they
have
these
different
types.
The
pneumatic
liquid,
crystal
phase,
for
which
the
constituent
rod
like
molecules,
exhibit
long-range
orientational
order
with
no
positional
order.
Is
this
this
one?
That's,
I
think,
probably
most
relevant
to
a
lot
of,
like
you
know,
single
cell
biology,
where
single
cells
are
behaving
in
these
groups.
A
So
this
is
where
you
might
have
like
rob
like
molecules
like
bacteria,
where
it
could
be
even
like
diatoms,
going
back
to
diatoms,
where
they
have
this
sort
of
long-range
orientation
orders
they
all
line
up
like
in
the
same
direction
and
they
line
up
in
a
layer,
but
they
don't
really
have
a
positional
order.
So
it's
just
kind
of
like
random
with
respect
to
position,
so
they
kind
of
mimic
this
sort
of
pattern
formation.
A
But
it's
you
know
there.
It's
it's
not
driven
by
signals,
it's
just
driven
by
the
physical
constraints,
there's
some
other
different.
So
in
the
cholesterol
phase
you
get
these
naturally
twist
natural
twists
of
molecules,
they
exhibit
helical
patterns,
and
you
know
so
that's
kind
of
like
dna
and
in
the
symmetric
phase
the
molecules
arrange
themselves
in
layers
and
there
is
an
orientational
order
within
the
layers
and
the
layers
can
slide
past
each
other.
So
they
talk
in
this
review.
A
They
talk
about
a
lot
of
these
different
liquid,
crystal
systems
pneumatic
liquid
crystals
in
particular,
and
so
you
know
they
kind
of
go
through
all
of
this
information
about
like
the
mathematics
of
it
and
there's
a
book
of
course
on
the
physics
of
liquid
crystals
by
dejan.
That's
a
classic
book,
and
I'm
gonna
show
that
book
here,
but
this
is
kind
of
following
up
on
that
work,
and
so
there's
a
lot
of
mathematics
that
you
have
to
apply
to
these
models.
A
So
again,
this
is
something
that
you
know
it
doesn't
require
any
chemical
signaling,
it's
just
the
physical
constraints,
and
so
let's
see
that's
not
what
I
want
to
talk
about
here.
I
I
was
going
to
talk
a
little
about
symmetry
breaking,
but
I
think
it's
getting
rather
late
to
get
into
that.
I'm
just
going
to
talk
about.
D
A
D
D
A
A
No
can't
remember
anyways
there.
There
is
this
review
that
I
had,
and
I
don't
know
where
I
put
it.
It
was
on
active
matter
and
on
you
know
some
of
the
it
was
a
review
article
and
some
of
the
details
on
that
and
how
it
links
to
this
field
so
yeah.
This
is
definitely
a
thing
you
know.
Maybe
we
should
follow
up
on
because
it's
a
really
interesting
area.
D
A
I
think
that
would
be
great
yeah
it'd
be
great,
so
let's
put
a
pin
in
that
and
see
yeah,
it's
definitely
something
that
is
to
follow
up
on
and
then
the
other
thing
I
wanted
to
mention
here
is
that
that
there
are
these
papers
that
I
had
for
that
talk
about
symmetry
breaking
and
their
involvement
in
biology,
so
symmetry
breaking
is
something
from
physics
that
involves
where
you
have
this
thing.
That's
symmetrical.
A
A
So
I
think
dick
talks
a
lot
about
that
in
some
of
his
work
on
on
differentiation,
trees,
where
you
know
the
the
symmetry
breaking
is
an
important
part
of
where
you
have
these
changes
in
shape
between
like
the
front
and
the
back
or
the
sides,
and
so
this
is
something
else
we'll
come
back
to,
but
I
just
wanted
to
kind
of
highlight
it
now,
because
I
have
this
folder
open,
but
there
are
a
number
of
different
papers.
I
have
on
symmetry
breaking.
A
This
is
the
original
paper
on
this
from
rashevsky
in
the
bulletin
of
mathematical
biophysics.
This
is
1940,
and
this
is
called
the
physio
physical
mathematical
aspects
of
some
problems
of
organic
form.
So
it
talks
about
a
little
bit
about
the
you
know
how
you
get
these
differential
embryos
and
differential
embryonic
phenomena.
A
So
one
of
the
things
about
they
started
looking
at
cell
polarity,
which
is
where
you
have
a
polarity
in
the
embryo,
and
so
you
can
go
from
that.
You
end
up
with
this
symmetry
breaking,
which
is
where
the
symmetry
no
longer
holds.
So
you
don't
have
a
mirror
image
on
one
side
or
another,
and
so
this
kind
of
describes
some
of
the
mathematics
of
this.
A
It
talks
about
in
terms
of
axial
gradients
of
metabolite
concentration
across
the
the
embryos.
So
there's
this
very
even
distribution
and
you
go
from
an
even
distribution
to
something.
That's
you
know
patterned
or
non-normal.
So
there's
this
breaking
that
happens
so
and
then
this
is
a
it's
from
quantum
magazine.
So
this
is
a
more
maybe
understandable
version
of
this
where
they
talk
about.
A
You
know
rotational
invariance,
in
physical
systems,
so
this
is
a
type
of
symmetry
and
then
they
talk
about
the
symmetry
of
phase
transitions
here,
and
so
this
is
and
it
goes
into
the
same
sort
of
area.
But
it's
a
little
bit.
You
know
it's
a
little
bit
different
topic
when
applied
to
the
embryo,
but
so
this
is
a
conformal
invariance
in
the
izing
model.
So
this
is
where
you
have
a
phase
change
and
you
can
transform
different
shapes.