►
From YouTube: DevoWorm (2021, Meeting 35): Hacktoberfest, 2021 annual update, Complex Systems and Cell Projections
Description
Hacktoberfest plans, 2021 update to the OpenWorm Foundation. Upskilling and conference participation. Wolfram's views on rulology and metamodeling. Papers on graph processing systems, developmental hypergraphs, and topological folding in Drosophila embryo/how to capture this computationally. Attendees: Bradly Alicea, Jesse Parent, Sanjay, Susan Crawford-Young, Akshay Nair, and Madhuri Srinivasan.
B
A
C
B
C
D
B
A
F
A
A
A
A
A
So
yeah
this
group,
we
kind
of
mix
the
two
we
have
like
the
sort
of
the
biological
side
and
then
we
have
the
computational
side.
So
that's
what
we
try
to
bridge
here,
and
so,
if
you
have
any
things
you
want
to
follow
up
on
in
a
meeting,
let
me
know
or
let
any
you
know
someone
else
is
presenting.
Let
them
know
what
you
find
interesting
because
there's
usually
usually
like
the
saturday
meetings.
There
are
a
lot
of
topics
that
we
kind
of
gloss
over
and
then
there's.
A
I
find
it
useful
to
go
through
a
lot
of
these
things
at
a
high
level
because
it
introduces
people
to
a
lot
of
things,
and
then
you
can
do
those
you
know
follow
up
on
those
things
later
right
yeah,
so
we
also
have
jessie
and
sanjay
welcome
all
right,
okay.
So
the
first
thing
I
want
to
talk
about
was,
let's
see.
A
A
I'm
going
to
share
my
screen,
so
one
of
the
areas
that
we're
going
to
have
as
a
repository
for
hacktoberfest
as
the
digital
basel
area
and
I've
been
working
with
ujol
singh
and
asmitzing
and
thirun,
and
you
know
those
guys
from
the
monday
meetings
and
we're
meeting
kind
of
separately
to
talk
about
these
sort
of
doing
it.
A
second
pass
at
the
digital
vessel
area
stuff.
A
So
this
is
the
project
we
have
this
diatom
colony
that
we're
you
know,
looking
at
the
morphology,
we're
looking
at
the
behavior
and
so
we're
using
movies
to
segment
out
or
to
analyze
those
images
to
define
the
cells
themselves
and
then
define
the
colonies
and
the
movement,
and
all
that.
So
we
have
a
bunch
of
different
repositories
and,
of
course,
we've
put
out
a
paper
on
this,
which
is
a
recent
book
chapter
and
it
you
know,
involves
deep
learning
and
bio
biomechanics
and
all
this
sort.
A
A
These
movies-
and
so
we
had
some
interesting
talk
about
some
cutting
edge,
deep
learning
technology
and
hopefully,
if
they
can
get
it
working,
they
can
present
on
it.
In
this
group
they
used
a
pre-trained
model
for
this
version
of
the
analysis,
which
is
deep,
lev
version
three
and
that
that's
something
you
can
implement
in
tensorflow,
and
so
we
have
a
lot
of.
We
have
a
lot
of
stuff
that
we
did
for
this
and
we're
trying
to
move
to
a
second
phase
and
I'm
still
trying
to
figure
out
what
the
issues
are
for
this.
A
I'm
gonna
maybe
go
through
these
rep.
This
repository
and
kind
of
pull
out
some
issues.
We
also
have
the
divo
learn
repository
and
that's
here.
So
this
is
the
diva
learn
repository
and,
of
course,
this
has
been
worked
on
by
my
oak
and
my
knock
dub
and
they've.
You
know
worked
out
so
this
has
a
more
active
set
of
this
was
active
last
october
hectoberfest
and
it's
been
active
all
year
in
terms
of
accepting
contributor
requests
so
or
contributor
contributions.
A
B
A
We
put
your
name
on
a
list
and
we
advertise
it.
You
know
in
various
places,
so
you
know
and
then
of
course,
if
you
make
a
number
of
contributions,
of
course
you
know
that
might
lead
to
other
things.
So
you
know
we
don't
we
don't
go
the
swag
route
on
this,
but
nevertheless
you
know
we're
gonna
see
how
many
people
we
can
get
in
during
hacktoberfest.
A
A
So
this
is.
This.
Is
still
outstanding,
I
need
to
revisit
this
diva
learned
paper,
which
is
been
put
on
hold
for
a
long
time,
because
we've
had
other
things
come
up,
but
this
is,
of
course
the
thing
I
just
showed
you
the
diva
learn
repository
now
we
just
finished
google,
some
of
our
code.
So
that
means
that
we're
pretty
much
at
a
stopping
point
for
modifying
the
platform
and
then
soon
there
will
be
a
new
release
on
that.
So
this
paper
will
be
active.
Once
again,
we
have
some
other
things.
A
Some
other
ideas,
like
williamson
symbiosis,
molecular
level
simulations
of
the
diatoms
that
we
showed
before
this
williamson
symbiosis.
Is
this
idea
that
that
some
organisms
have
taken
like
basically
the
genomes
of
two
different
animals
in
like
two
different
animals,
developmental
programs
into
their
genomes,
and
they
express
them.
You
know
this
is
for
organisms
like
ones
that
undergo
metamorphosis
and
other
forms
of
transformation
and
development,
and
so
you
know
there
are
ways
to
look
at
this.
A
It
involves
a
lot
of
genomics
and
you
know
similarity
analysis,
so
this
is
something
that's
you'd
have
to
be
pretty
advanced
in
bioinformatics.
To
do
this.
This
molecular
level
simulations
of
diatoms-
that's
maybe
a
little
bit
different.
This
is
where
you're
looking
at
motion
you're
looking
at
higher
order
motion,
so
this
is
something
that
might
require
us
to
finish
up
this
analysis
of
the
movies,
but
it's
open.
A
We
have
another
diatom
movement
item
here,
looking
at
the
smoothness
or
jerkiness.
This
is
kind
of
the
same
thing.
Jerkiness
is
like
where
you
get
really
hot.
You
know
really
fast
movements.
So
you
have
these
derivatives
of
position.
You
have
motion
derivatives
of
motion.
Actually
you
have
acceleration,
you
have
other,
and
then
you
have
higher
order.
A
Derivatives
and
those
are
basically
what
you
would
be
looking
at
in
this,
this
quantitative
comparison
of
our
key
and
shape
droplets.
I
think
this
is
something
that
my
knock
was
looking
into.
He
did
a
demo
a
couple
weeks
ago
and
I
don't
know
what
the
state
of
that
is.
So
I
can
actually
probably
make
a
note
of
that.
If
I
haven't
already
oh.
B
A
Those
are,
I
think,
closed
now
for
submissions,
but
you
know
if
you're
interested
in
attending
your
ips,
if
you're
interested
in
machine
learning
or
deep
learning,
that's
a
great
place
to
go
for
that
and
just
participates.
You
can
participate
virtually.
I
think
this
year
and
it's
a
great
place
to
go,
especially
the
workshops
where
they
have
workshops
and
a
number
of
topics.
A
Not
all
of
them
are,
you
know
just
kind
of
like
self-contained
machine
learning.
Some
are
sort
of
like
metacognition
and
things
like
that,
so
that
this
link
to
the
workshop.
Let
me
let
me
make
a
copy
of
this
link
here
and
I'll
put
it
in
the
chat
for
people,
so
they
can
look
those
over
and
I
think
the
registration
for
rips
is
rather
inexpensive.
A
And
then
so
that's
that's
near
ips
and
then,
if
you're
also
interested
in
even
more
conference
participation,
there
is
neuromatch
4.0,
which
is
october
25th.
Is
this
submission
deadline
for
abstracts,
so
we're
going
to
be
participating
in
this
conference?
This
is
a
virtual
conference
that
occurs
in
december
at
right
around
the
same
time
as
an
r.I.p
house,
but
it
doesn't
overlap
with
their
ips
because
I
think
the
a
lot
of
the
same
people
go
to
both
so
that
that
won't
overlap
too
much.
A
But
you
know
it's
going
to
be
a
pretty
good
time.
It's
usually
a
pretty
good
conference
and
speaking
of
which
and
I've
created
an
abstract
submissions
document
here
for
neuromatch.
So,
like
I
said
I
well,
I
have
a
group
on
saturdays
that
also
is
going
to
be
participating
in
neuromatch.
So
what
I'm
doing
here
is
I'm
trying
to
assemble
a
list
of
submissions
that
we
can?
A
A
A
A
A
You
know
in
subsequent
conferences
of
the
same
series,
because
I
always
think
we
have
something
new
to
say.
I'd
always
like
to
think
that
we
can
add
some
value
to
it
as
we
resubmit
them.
So
I
would
like
to
you
know
if
we
do
submit
something
like
this.
I
would
like
to
update
it
a
bit
this.
The
same
goes
for
this.
It's
a
poster.
A
We,
you
know,
probably
submitted
as
a
talk,
so
it
would
change
and
it's
the
way
it's
presented,
but
then
we
could
add
some
new
things
in
as
well,
so
these
are
and
then
there's
this
embryo
connectome
networks,
which
is
also
a
diva
worm
thing
and
that's
again,
that's
something.
That's
been
submitted
to
other
conferences,
but
in
different
forms,
and
so
you
know
we
can
recycle
most
of
the
slides,
but
we
do
want
to
add
something
new.
A
Please
you
know,
take
a
look.
I
think
the
sharing
is
on
on
that
document
you
might
want
to.
If
it's
not,
you
might
want
to
request
access,
but
if
you
want
to
take
a
look-
and
maybe
you
know
put
some
ideas
in
if
you
have
something
and
we
can
follow
up
on
it
as
we
go
through,
that
would
be
good.
I
think
that's
something
that
is
of
benefit
to
most
people.
A
Some
of
you
haven't
been
in
the
group
that
much,
but
that's,
okay,
you
can,
you
know
you
can
still
maybe
propose
something
or
work
on
something
with
someone
else.
If
you
want
to
work
on
a
presentation
with
someone
else,
you
could
make
a
note
and
say
this
person
is
interested
and
then
you
know
we
can
see
it
through.
The
important
part
here
is
seeing
something
through.
So
you
know
you
put
an
idea
down
and
you
kind
of
try
to
flush
it
out
in
over
well.
In
this
case,
it
would
be
a
couple
weeks.
A
We
have
like
maybe
three
three
and
a
half
weeks
by
the
you
know
until
the
abstract
deadline,
and
then
you
know
we'd
basically
work
out
an
abstract
and
then
we'd
have
to
work
out
the
slides
and
then
once
that's
done,
you
have.
You
know
something
to
present
at
a
conference,
and
you
know
it's
it's
it's
pretty
simple.
It's
only
a
15
to
20
minute
session.
A
A
A
There's
some
develop
not
really
a
developmental
biology
track,
but
there's
some
like
developmental
there's
some
interest
in
development.
So
that's
something
we
can
sort
of
address
and
then
there's
also
like
the
deep
learning
machine
learning
side
of
things.
So
if
people
are
interested
in
like
deep
learning
and
machine
learning,
submissions
nurips
is
actually
quite
competitive.
So
this
might
be
an
alternative
to
that.
A
But
I
don't
know
that
might
actually
not
be
a
bad
idea.
Actually,
some
people
might
prefer
that
so
well,
more
more
on
that
later,
let
me
know
if
you
have
ideas
about
that,
you
can
share
them
in
the
slack
okay.
So
now
that
we've
done
that,
I
want
to
talk
about
this.
Actually,
I
don't
want
to
go
over
major
tasks.
A
I
do
want
to
talk
about
the
open
world
general
meeting
so
tomorrow
morning,
we're
going
to
have
at
least
my
time
we're
going
to
have
an
annual
update
to
the
open
worm
foundation
general
meeting,
and
so
this
is
the
it's
like
a
five
to
ten
minute
presentation
on
what's
been
going
on
in
the
diva
one
group.
So
this
is
the
diva
worm
group.
This
is
these
are
our
this
is
our
web
presence.
A
We
have
some
other
things
that
we're
working
on
model
organisms,
data,
science,
demos,
theory,
building
some
other
educational
initiatives-
things
that
you
know
are
kind
of
in
different
states,
they're,
not
as
developed
as
diva
learned.
But
that's
the
idea
there
is
to
have
like
a
an
educational
portal
for
people
to.
A
To
go
through
so
this
is
the
talk
here,
this
update
for
2021
and
then
going
through
this.
I
noticed
that
there
were
a
lot
of
things
that
we
did
and
I
just
didn't
know
like
you
know,
the
pace
of
things
is
like
you
know
you
kind
of
go
week
after
week,
maybe
monthly,
you
take
an
assessment
of
where
you
are
and
you
just
work
on
things,
but
it's
you
know
going
back
over
the
course
of
a
year.
A
You
get
to
see
like
the
full
extent
of
what
has
been
going
on,
and
it
turns
out
that
actually,
we've
done
a
lot
of
stuff
in
this
group
over
the
course
of
the
year.
So
I'm
proud
of
that
and
I'm
going
to
talk
about
this
to
the
group
tomorrow
morning.
So
this
is
the
title
slide
here.
This
is
the
logo
for
the
group,
the
very
arty
kind
of
embryo
type
logo.
A
So
one
of
the
things
we
do
in
the
group-
and
I'm
going
to
stress-
is
deep
learning
and
quantitative
morphology.
So
this
includes
things
like
axolotl
brain
development.
You
can
see
that
here
you
also
do
c
elegans
embryogenesis,
which
you
can
see
here,
there's
also
this
theory
building
to
explain
developmental
processes,
which
is
this
gif
of
a
sphere
turning
into
a
cow.
A
So
there's
a
thing
that
people
use
in
modeling
and
there's
also
a
joke
about
it,
about
spherical
cows
in
a
vacuum,
and
so
that's
what
this
rep
refers
to
often
times
when
you
have
an
underspecified
model
and
you
use
it
to
analyze
some
really
complex
system.
People
often
say
that
you're
looking
at
a
spherical
cone,
a
vacuum,
and
so
that's
where
this
comes
from,
and
it's
just
kind
of
an
idea
that
you
know
in
modeling
we're
not
doing
like
we're,
not
we're.
You
know
we.
A
A
This
work
here
just
shows
an
example
of
you
know:
segmenting
a
bunch
of
cells
as
you
go
through
a
stack
of
microscopy
images
from
the
top
to
the
bottom
of
the
embryo.
So
you
can
see
how
this
algorithm
is
able
to
pick
out
the
cells
as
it
moves
through
the
stack,
and
so
you
know
can
pick
out
different
variations
in
the
morphology.
A
But
then
you
also
have
this
situation
where
the
cells
migrate
in
development
and
they
differentiate
and
they
divide,
and
so
all
those
things
have
to
be
captured,
and
so
that's
that's
where
it
becomes.
You
know
more
and
more
dif,
increasingly
difficult,
as
you
add
those
things
in
and
then
finally,
we
have
a
bacillary
cell
comedy
morphology,
which
I
talked
about
earlier
in
the
meeting.
A
So
this
is
telling
the
group
the
larger
group
about
divalern.
This
has
been
a
success.
It's
just
hit
this
was
from
when
was
this.
This
wasn't
too
long
ago,
it's
like
a
month
ago
or
something
in
any
case,
so
our
pipey
instance
of
this
just
hit
10
10
000
downloads.
I
think
that
was
about
a
month
ago,
so
it
was
up.
B
A
A
In
the
last
30
days,
which
is
a
good
trend
to
have,
if
you
can
get
more
downloads
in
a
shorter
period
of
time
like
that,
that
means
people
are
starting
to
check
it
out,
and
certainly
we've
made
progress
on
it.
You
know
when
it
when
it
was
first
released,
there
were
some
buzz
and
then
it
was
like
you
know
we
had
to
wait
a
couple
months
for
people
to
start
contributing,
and
then
you
know
now
we
have
a
second
summer
of
code
project
that
focused
on
this.
A
A
Looking
at
things
like
warm
movement,
they're
looking
at
things
like
modeling,
the
nervous
system,
modeling
different,
neurons
they're,
looking
at
things
like
the
biophysics
of
movement,
so
they're
looking
at
things
that
are
not
only
in
the
adult
but
also
have
a
lot
of
complexity
to
them,
and
you
know
this
kind
of
model
might
help.
I
know
there's
been
a
lot
of
interest
in
machine
learning
and
deep
learning
more
generally
in
in
open
world,
but
we
don't
really
have
a
lot
of
people
doing
anything
outside
of
diva
worm
in
that
area.
A
A
You
know
gain
new
skills
that
maybe
they
wouldn't
be
able
to
get
in
a
in
a
degree
setting
or
maybe
even
like
you
know,
in
a
linkedin
course
or
a
you
know,
some
sort
of
mooc,
where
they
just
learn
the
sort
of
the
theory
of
it
where
they
kind
of
learn
like
from
some
examples.
In
this
case,
you
can
actually
work
with
the
data,
so
this
is
our
evil
learn
platform
over
the
20
and
21
seasons.
So
we
have
this
umbrella,
this
devil
and
umbrella.
A
A
So
we
have
those,
and
then
we
have
the
devozu,
which
has
a
focus
on
model
organisms
and
digital
organisms
and
sort
of
you
know
introducing
those,
and
I
think,
that's
something
that
we
can
revisit,
because
I
know
that
I
think
I
said
this
last
week.
Maybe
he
was
at
my
other
meeting
where
there's
this
interest
in
model
organisms,
not
as
like
what
they
can
do,
but
like
kind
of
bridging
between
different
model
organisms.
A
So,
for
example,
the
devozu
features
a
lot
of
different
model
organisms
like
zebrafish,
fruit
flies,
c,
elegans,
of
course,
spiders
octopi.
And
then
you
have
these
digital
organisms
that
you
can
create
that
you
know,
allow
you
to
model
different
aspects
of
development,
so
those
are
all
in
the
devo
zoo,
and
I
think
people
have
an
interest
in
that.
A
Like
there's
some
sort
of
synthesis
there,
I
think
as
well,
where
we
can
understand
a
lot
about,
like
you,
know,
general
phenomena
by
looking
across
model
organisms,
and
you
know
having
a
resource
where
people
can
go
and
learn
more
about
like.
Why
do
we
use
zebrafish
or
what's
the
the
utility
of
using
like
xenopus,
which
is
the
frog
model?
A
What
you
know
what's
the
utility
of
that,
so
there's
lots
lots
to
develop
on
that
end.
I
don't
know
exactly
how
to
drive
this
forward
in
terms
of
funding
your
projects,
you
know,
but
I
mean
I
guess
the
best
way
to
do
it
is
just
to
you
know,
maybe
that's
something
else
we
could
raise
during
oktoberfest,
not
as
a
whole.
You
know
repository,
but
as
an
issue
like
look
into
devozu
and
suggest
changes,
so
this
is
something
that
can
be
done
as
well.
B
A
A
A
They
work
on
this
thing
called
fair
neuroscience,
which
is
it's
like
this
acronym,
for
you
know
what
we
should
expect
of
open
data,
so
you
know
it's
like
transparency
and
reusability
and
and
all
these
other
things
I
I
don't
remember
exactly
what
the
acronym
means
right
now,
but
so
we
we
were.
We
were
participated
in
the
incf
assembly
with
the
rest
of
the
open
room
foundation.
We
did
this
poster
at
the
international
c
elegans
conference
on
the
open
arm.
A
A
We
also
participated
in
the
neuromatch
conference
and
this
was
neuro
match
3,
which
was
back
in
I
think,
october
or
november
of
last
year,
and
we
did
a
couple
of
talks
there.
We
did
this
thing,
the
psychophysics
of
non-neuronal
cognition,
which
we
just
finished
a
paper
on
this
in
the
group
and
so
there's
a
whole
paper
on
this
now
and
it's
coming
out
as
a
pre-print
soon,
but
you
can
read
about
it
or
you
know.
This
talk
is
online
and
this
is
kind
of
weighing
out
sort
of
this
idea
of
that.
A
You
can
use
this
metaphor
of
cognition
as
a
way
to
look
at.
In
this
case,
we
use
the
basilar
example,
but
I
think
this
talk
was
a
bit
broader
and
then
we
have
this
contrast
between
biological
and
artificial
neural
networks,
which
is
something
that
krishna
katyal
and
myself
worked
on.
So
you
know
we
participated
in
that.
A
We
also
participated
in
networks
2021.
We
did
this
embryo
networks.
Presentation,
which
is
this
idea
that
embryo
cells
are
connected,
can
be
conformed
net.
What
they
call
networks,
complex
networks,
so
they
have
linkages,
they
have.
You
know
based
on
their
spatial
position
and
their
distance,
and
you
can
it's
it's
informative
in
a
number
of
ways,
and
in
this
particular
talk,
I
talked
about
how
you
can
use
this
type
of
model
to
look
at
what
happens
as
the
connectome
starts
to
form,
and
that
starts
to
diverge
from
the
other
cells
in
the
embryo.
A
A
There's
also
this
talk
of
dynamics
days
on
game
theory
of
developmental
processes,
that's
also
on
our
youtube
channel,
and
then
this
advanced
computational
neuroscience
network
talk
is
also
on
our
youtube
channel.
I
believe
this
is
another
version
of
this
network's
talk.
It
actually
features
a
little
bit
more
work
on
the
on
the
network
side
of
it,
and
so
that's
something
you
also
check
out
on
youtube.
A
There
are
a
couple
also
a
couple
papers.
This
year
we
had
the
neuro
informatics
contribution.
This
was
in
the
special
issue:
building
the
neuro
commons.
This
is
a
paper
called
data,
theoretical
synthesis
of
the
early
developmental
process.
This
is
something
I
think
I
showed
to
the
group
at
the
beginning
of
this
calendar
year.
A
This
was
a
paper
on
you
know,
looking
at,
like
you
know,
using
secondary
data
to
build
models
and
hypotheses
and
then
also
building
theory
and
there's
this
whole
special
issue
this
this
building,
the
neural
commons
special
issue
is
put
out
by
people
in
the
incf
working
group
on
open
data,
so
they
have.
If
you
go
to
the
special
issue,
they
talk
a
lot
about
some
of
the
the
fair
standards
and
things
like
that.
A
So
that's
something
to
check
out-
and
I
know
it'll
be
a
great
interest
to
the
divor
or
the
open
worm
foundation,
because
one
of
the
founders
of
the
open
room
foundation
actually
has
pretty
strong
ties
to
that
that
world
of
neuroinformatics
so
and
then
there's
this
biosystems
paper
with
oshawa
and
jesse
as
co-authors
of
me,
periodicity
in
the
embryo,
and
this
is
a
sort
of
a
very
low
level.
So
I
know
susan
is
talking
at
the
beginning
of
the
meeting
about
signal
processing.
A
There
wasn't
a
lot
of
signal
processing
in
here
per
se,
but
we
did
look
at
with
different
aspects
of
the
time
series
in
a
periodicity
of
cell
division
and
cell
differentiation,
and
what
that
you
know
what
that
says
about
how
order
is
formed
in
the
embryo.
So
so
the
idea
here
is
that
you
know
I
kind
of
want
to
highlight
our
process
for
opens
sort
of
open
involvement,
and
so
this
is
preparation
for
google
summer
of
code.
But
this
kind
of
applies
year-round
in
terms
of
people
coming
into
the
group.
A
A
A
I
think
I
pick
people
who
have
put
in
like
some
contribution
to
that
we're
just
that
repository.
But
of
course
we
have
other
people
make
meetings,
and
I
didn't
put
them
in
this
list.
This
was,
I
think,
for
hacktoberfest
of
last
year,
so
this
year,
I
think
if
you
contribute
in
some
way,
you'll
make
a
new
list,
they're
also
cross-disciplinary
pathways,
so
we
have-
and
this
is
something
that
midduri
is
experiencing
currently.
A
This
is
where
we
have
this
orthogonal
research
and
education
lab,
which
I
operate
as
head
scientists,
and
we
have
jesse
and
madurai,
are
part
of
that,
and
then
I
think
christian
is
also
part
of
it,
and
people
like
shruti
who
have
contributed
to
this.
This
is
a
broader
set
of
opportunities
in
an
area
called
neuroai,
there's
also
ethics
and
computational
modeling
work
that
goes
on
here
and
then
you
know
you
can
interact
with
evorum,
which
is
computational
developmental
biology.
A
We
also
were
interested
in.
We
talked
on
saturday
about
neural
embryoids
or
neural
organoids,
and
so
those
things
are
possible
as
well.
We
kind
of
have
this
crossover
where
we
have
spillover
research
or
spillover
opportunities,
so
we
want
to
kind
of
make
that
more
explicit.
I
want
to
bring
that
to
the
attention
of
the
foundation
as
well
and
then
finally,
thanks
to
our
contributors.
So
this
is
not
a
like,
I
said,
not
an
exhaustive
list.
This
is
sort
of
people
who
are
some.
You
know
oftentimes
regulars
in
our
meetings.
A
So
if
I've
missed
you
here,
you
know,
I
could
probably
add
you,
but
this
is
just
a
sampling
here
and
then
we
have
our
google
summer
of
code
people
and
you
know
thank
you
for
that.
So
that's
that's
what
I
wanted
to
cover
in
that
talk.
You
know
I
don't
want
to
I'm
not
going
to
be
that
far-ranging
in
the
when
I
presented
the
foundation.
I
just
want
to
go
over
some
of
those
points.
A
A
A
and
imagine
if
you
started
the
workforce
back
in
like
the
early
90s
and
you
you
know,
had
an
undergraduate
degree
and
then
you
didn't
go
back
to
school
and
you
don't
really
have
time
to
go
back
to
school,
and
so
you
need
to
learn
this
new
skill.
So
you
know
that's
one
thing
you
can
learn.
Is
you
know?
That's
that's
upskilling
in
terms
of
new
technology
that
you
can
learn,
and
you
know
there
have
to
be
pathways
for
that.
A
A
You
know
you
can
take
a
course
at
a
local
college,
but
it's
takes
time
to
do
that,
like
you
have
to
commit
to
an
entire
semester,
pay
a
lot
of
money
for
it
or
you
know
so,
and
then
you
still
may
not
really
get
the
full
experience.
You
know
you
might
have
to
take
a
course
in
microbiology
which
isn't
really
c
elegans.
You
know,
but
the
point
is:
is
that
there
you
know?
Maybe
there
are
opportunities
for
this
upskilling
and
that's
something
that
I
don't
know.
A
A
B
A
Like
sometimes
very
practical
considerations
get
in
the
way
like
can
I
take
a
course
at
night
or
can
I
do?
I
really
need
to
take
like
an
entire
course
to
learn
exactly
what
I
want
to
know
for
my
thing.
So
yeah,
I
think
that's,
that's.
That's
always
a
thing
I
mean
that's
where
a
lot
of
the
you
know
that
they
have
these
micro
credentials
online,
where
you
can
take
that
this
is
common
in
computer
programming
or
you
can
take
like
a
micro
credential
for
a
language
like
some.
A
A
C
A
A
Automata
we've
talked
about
this
a
lot
in
the
group
and
he
used
it
to
model.
You
know
just
to
run
these
models
and
he
extracted
what
he
called
these
rules
of
order.
So,
like
the
idea
is
you
have
like
these
grids
of
cellular
automata
and
they
generate
rules
if
you
run
them
under
certain
conditions
and
those
rules
then
can
be
classified
so
you
can
like
give
them
numbers
and
those
rules
map
to
things.
A
You
know
in
complexity,
classes
in
computer
science,
or
sometimes
they
map
to
systems
like
biological
systems
or
physical
systems
in
the
real
world.
So
you
can
build
these
models
that
are
pretty
abstract,
that
just
have
a
bunch
of
cells
interacting
and
changing
their
state,
and
they
can
actually
approximate
real
behaviors.
A
A
There's
a
you
know,
type
of
morphogenesis
you
can
achieve
with
these
rules
with
some
of
these
rules,
but
also
you
can
map
them
to
more
complex
systems
and
not
just
developmental
systems
where
there's
pattern
formation,
but
emergent
phenomena
such
as
like
urban
settlement
patterns
or
traffic
patterns
or
other
things.
So.
A
Of
the
print
computational
equivalence
principle,
which
is
that
you
know
when
I
said
before,
the
models
are
pretty
abstract
and
you
know
they.
You
know
you
can't
necessarily
say
that
you're
modeling,
something
just
because
you're
applying
a
model
that
looks
like
it's
similar.
You
know
you
really
have
to
work
at
getting
it.
You
know
specified
in
the
right
ways.
There's
also
this
rule
called
the
you
know:
computational
equivalence
principle,
which
is
the
idea
that
you
can
map.
A
You
can
indeed
map
from
models
to
real
world
phenomena
that
there
is
this
sort
of
continuity,
that
what
you're
modeling
isn't
just
this
heuristic,
but
it
actually
does
have
you
know
some
credibility
in
terms
of
like
modeling
real
phenomena.
So
you
know
you
have
this
these
computational
rules
of
assembly.
A
Those
things
you
know
are
actually
what
you
see
in
nature,
it's
just
much
more
abstract
than
like.
If
you're
you
know
watching
a
bunch
of
birds
flocking,
you
know
at
dusk.
You
know
that
that
has
some
computational
principles
underlying
it.
So
that's
the
idea
here
and
he's
really
into
rules.
So
he
has
this
thing.
He
says
charting
a
course
for
complexity,
meta,
modeling,
ruleology
and
more
so
in
the
world
is
metamodeling
and
reallyology.
A
So
this
is
actually
a
pretty
good
article
he's
talking
about
in
the
1980s
he
started.
Working
on
this
he's
studied
the
approaches.
People
tried
to
use
non-equilibrium
non-equilibrium,
thermodynamics
synergetics,
which
is
a
type
of
complexity.
Theory,
it's
a
type
of
dynamical
systems;
theory,
non-linear
dynamics,
cybernetics
general
systems
theory.
So
we've
talked
about
cybernetics
a
lot
general
systems.
Theory
is
like
a
lot
of
boxes
and
arrows
where
you're
looking
at
the
flows
in
a
network
or
in
a
complex
system
and
you're
trying
to
characterize
the
different
boxes
as
entities
and
the
arrows
as
flows.
A
I
imagine
that
the
key
question
was
starting
from
disorder
and
randomness.
How
could
spontaneous
self-organization
occur
to
produce
the
complexity
we
see,
so
we
assume
that
complexity
must
be
created
as
a
kind
of
filtering
of
ubiquitous
thermodynamic
like
randomness
in
the
world.
So
you
have
this
randomness
going
on
as
as
sort
of
the
initial
state,
and
then
you
build
from
there
there's
this
selection
mechanism
that
selects
out
a
lot
of
this
extraneous
disorder,
and
you
end
up
with
order
and
so
you're
looking
at
things
like
snowflakes
galaxies
life
forms
turbulence.
A
A
A
Is
this
the
first
thing
that
he
talks
about,
and
that
is
the
meta
science
of
finding
minimal
models
for
models-
and
I
know
jesse
is
interested
in
this,
so
this
is
but
the
idea
here
is
this.
I
was
at
a
complexity
conference
once
talking
to
someone
who
was
modeling
fish
in
their
behavior.
Probably
the
person
showed
me
a
simulated
fish
tank.
How
many
parameters
does
this
involve?
I
asked
about
90.
He
said
my
gosh.
I
said
with
that
many
parameters.
You
could
put
an
elephant
in
your
fish
tank
too.
A
So
this
is
you
know
just
this
is
an
older
context,
but
if
you
want
one
to
make
a
simulated
fish
tank
display
just
for
people
to
watch
and
having
all
those
parameters
might
be
just
fine,
but
it's
not
so
helpful.
If
one
wants
to
understand
the
science
of
fish,
the
fish
have
different
shapes.
They
swim
around
different
configuration.
A
A
We're
interested
in
a
lot
of
like
detail
in
the
model.
Maybe
we
want
to
know
like
a
lot
of
the
variations
so
depending
on
that,
we
need
to
have
you
know
more,
a
few
greater
or
fewer
parameters,
different
parameters
and
so
forth,
but
this
is
kind
of
getting
into
this
idea
of
meta
modeling,
finding
a
minimal
model
for
models.
So
you
want
to
find
to
understand
how
many
parameters
you
need.
You
need
a
minimal
model
of
your
model
and
then
you
can
find
that
answer.
A
So
it's
the
pure
basic
science
of
what
simple
rules
do
so
you
know
that
that
seems
kind
of
like
another
tautology,
but
it's
actually
like
you
know.
We
have
these
rules
that
we
can
extract
from
these
programs.
Now
what
happens
when
we
put
them
together,
or
what
do
they
actually
do?
They're
generating
a
pattern?
So
sometimes,
if
you
look
at
some
of
these
computational
rules
that
wolfram
has
produced,
they
look
like
things
in
the
world.
For
example,
I
think
rule
30
looks
like
like
a
seashell
right,
the
seashell
pattern.
A
A
So
we
can
understand
the
pattern
how
this
program
generated
the
rule,
but
we
can't
necessarily
understand-
and
we
can
understand
how
the
pattern
is
generated
in
the
biological
system,
but
we
can't
necessarily
understand
how
those
things
connect.
You
know
they're,
probably
not
the
same
process,
but
they
must
have
some
similarity
because
they
look
very
similar.
So
what
is
that
connection
there?
That's
that's.
A
rulology,
at
least
to
me
seems
like
it's
useful
for
and
so
this
I
don't
want
to
give
you
the
link
to
this
article.
A
It's
it's
fairly
long,
but
it's
off
of
his
personal
blog.
I
think
that
somebody
will
find
it
quite
interesting,
but
I
would
recommend
reading
it
if
you
want
to
unders.
You
know
just
kind
of
ruminate
about
some
of
these
ideas
of
complexity,
theory
and
biological
systems,
and
you
know
that
sort
of
thing.
So,
okay,
now
I'm
going
to
get
on
with
our
papers
and
if
you
need
to
leave
at
the
top
of
the
hour,
that's
fine.
A
A
A
A
A
In
like
social
networks
and
and
things
like
that,
and
then
graph
learning,
I
think,
is
an
offshoot
of
like
just
you
know,
using
graphs
in
computer
science,
so
like
graph
theory
and
computer
science
and
you
you're
just
applying
algorithms
to
it
now
this
is
so
those
are
graphs.
You
know
and
they're
different
rules.
You
can
have
directed
graphs
bi-directional
graphs.
A
A
A
A
You
can
put
that
graph
data
existing
graph
data
into
this
pipeline,
where
you
can
then
analyze
them
and
use
them
for
a
number
of
different
or
render
graphs
in
different
ways
for
scientific
computing
for
augmented
reality
and
visualization
for
machine
learning.
You
get
this
processed
output,
and
so
this
is
just
kind
of
goes
through
sort
of
the
history
of
graphs
and
graph
processing
and
then
the
next
decade,
which
I
think
is
where
they're
going
with
this
with
a
vision
for
you
know
applying
these
types
of
things.
A
So
the
idea
is,
you
know
it
comes
from
graph
theory.
It
comes
up
through
a
number
of
different
fields
and
then
the
there
you
know.
The
idea
is
that
you
have
these
diverse
sources
of
data.
These
are
diverse
types
of
graphs
coming
together
and
that
you
can
process
all
these
different
types
of
graphs
in
one
place
in
one
pipeline.
B
A
Like
graph
processing
like
you
might
do
it
in
applying
it
to
some
specific
you
know,
data
set.
This
is
where
you're
really
kind
of
pulling
together.
Everything
in
your
you
know
so,
like
we
have
different
types
of
data
in
developmental
biology.
Even
just
pick
one
domain
you
have,
like
you,
know,
gene
gene
regulatory
networks.
You
have
morphological
data,
you
have
metabolic
data,
all
of
those
can
be
represented
with
graphs,
but
they're
all
look
very
different.
They
all
have
different
types
of
nodes
and
specifications.
A
This
is
a
way
to
like
bring
that
together
into
one
single
sort
of
representation
and
process
it.
I
think
it's
interesting.
So
there's
that
and
then
there's
this
other
article
on
large-scale
graph
processing
systems
a
survey.
So
this
is
about
a
graph.
Is
a
significant
data
structure
describes
a
relationship
between
entities,
it
kind
of
goes
through
a
lot
of
the.
A
A
A
So
at
the
top
is
the
anterior
and
the
bottom
is
the
posterior
end,
and
you
can
see
this
process
of
cells
are
moving
kind
of
they're
migrating,
but
they're
migrating
in
concert.
So
you
can
see
that
there's
they're
labeled
by
colors
here
and
you
can
see
the
movement
if
it
plays
again
and
then
you
have
the
cells
that
are
sort
of
black
and
white
and
you
can
see
the
or
you
know
transparent,
and
you
can
see
the
movement
of
different
parts
of
the
embryo
and
they're
moving
in
concert
and
there's
this
process
called
gastrulation.
A
Where
there's
this
it's
you
know,
they're,
not
only
migrating
sort
of
folding
in
different.
You
know,
layers
of
cells
that
are
going
moving
in
concert.
So
that's
the
idea.
That's
what
we're
trying
to
understand
here
in
this
paper,
so
this
is
deconstructing
gastrulation
at
the
single
cell
level,
and
so
the
summary
of
this
paper
is
a
gastrulation
movements
in
all
animal
embryos,
start
with
regulated
deformations
of
patterned
epithelial
sheets.
A
Current
studies
of
gastrulation
use
a
wide
range
of
model
organisms
and
emphasize
either
large-scale
tissue
processes
or
dynamics
of
individual
cells
and
cell
groups.
Here
we
take
a
step
towards
bridging
these
complementary
strategies
and
deconstruct
early
stages
of
gastrulation
in
the
entire
drosophila
embryo,
where
transcriptional
patterns
in
the
blastoderm
give
rise
to
region-specific
cell
behaviors.
A
So
this
is
where
you
have
these
in
the
blastoderm,
which
is
this
stage
of
the
of
development.
You
have
these
transcriptional
patterns
and
individual
cells,
where
genes
are
being
expressed
and
they're
being
expressed
in
a
certain
way,
and
this
gives
rise
to
region-specific
cell
behaviors,
so
in
cells
in
different
parts
of
the
embryo
start
getting
these
differential
transcriptional
patterns
or
they
start
expressing
differential
transcriptional
patterns,
and
this
could
be
like
any
one,
gene
being
regulated
differentially
in
different
parts
of
the
embryo,
and
this
is
when
this
starts
to
happen.
A
Our
approach
relies
on
an
integrative
computational
framework
for
cell
segmentation
and
tracking
and
on
efficient
algorithms
for
event
detection.
So
what
they're
doing
is
they?
First,
they
segment
the
cells,
they
track,
the
cells
to
see
where
they
go
in
space
and
then
they're
able
to
detect
events,
and
these
events
could
be
different
genes
that
are
being
expressed.
So
they
use
these
fluorescent
markers
and
that's
why
they
have
the
color
coding
in
in
that
figure
and
they
look
at
the
intensity
of
it.
A
A
So
this
is
a
pivot
that
really
kind
of
goes
through.
All
of
this,
so
they're
doing
that
they
have
this
method
for
kind
of
looking
at
this
process,
they're
able
to
start-
and
actually
this
is
interesting
because
they
actually
start
by
forming
a
two-dimensional
projection
of
the
apical
side
of
the
epithelium
sheet.
For
each
time
point,
then
they,
the
cells
in
the
projection
of
the
first
time,
point,
are
segmented
and
assigned
a
unique
track.
Id
cell
boundaries
and
track
ids
are
then
propagated,
iteratively
and
automatically
over
time.
A
They
deform
each
projected
image
to
accurately
match
the
position
of
each
cell
within
the
same
cell
with
the
next
time
point.
So
these
algorithm
to
align
with
the
different
views
of
this
embryo
as
the
cells
are
moving
around
and
it's
not
just
the
cells
that
are
migrating.
It's
that
she
the
sheet
that
these
cells
form
is
folding.
So
you
have
to
understand.
A
B
A
All
image
processing
techniques.
Lastly,
the
segmented
and
track
2d
projection.
Each
time
point
is
used
to
generate
a
3d
polygonal
mesh
that
reveals
the
apical
surfaces
of
the
cells
and
an
n-total
map
of
cell
cell
adjacencies,
and
so
this
is
their
technique
for
sort
of
visualizing.
All
of
these
data
bringing
them
together,
and
so
so
they
focus
on
the
germ
band,
a
domain
of
the
embryo
that
converges
and
extends
as
a
consequence
of
multiple
cell
intercalations.
A
Most
of
these
events
involve
cell
quartets,
so
cells
in
four
be
behaving
together
and
proceed
through
so-called
t1
transition,
in
which
two
non-adjacent
cells
come
together
and
split.
The
orthogonal
cell
pair,
another
intercalary
event
involves
more
than
four
cells
and
proceeds
through
a
configuration
where
several
contracting
interfaces
generate
a
so-called
rosette
state.
So
these
are
like
these
clusters
that
they
form,
so
they
are
able
to
look
at
all
these
different
types
of
geometric
transformations
that
the
cells
are
making
as
they're
moving
around
and
they're
engaging
in
these
interpolations.
A
A
Okay,
so
this
is
exactly
what
it
looks
like.
This
is
the
drosophila
embryo,
it's
different
than
the
c
elegans
embryo,
because
there
are
a
lot
more
cells,
but
also
they
go
through
the
state
of
cellularization.
So
they
have
this.
They
have
these
different
c
elegans
embryo
the
cells
are
pretty
much
divide
and
they
sort
of
form
this.
A
What
it's
going
to
look
like
in
the
adult
with
some
you
know,
there's
some
differentiation
in
the
embryo,
but
then
you
know
it's
not
that
the
developmental
states,
you
know
you,
don't
have
these
transformative
states
like
you
do
in
in
drosophila
and
drosophila.
You
have
the
state
where
their
cellularization
and
then
there's
a
lot
of
folding
that
goes
on
here.
So
you
have
these
different
parts
of
the
embryo
that
fold
together
they
fold
over
and
they
turn
and
they
do
a
lot
of
coordinated
movements.
A
A
They
can
also
look
at
some
of
these
things
in
this
ventral
view
and
this
mercury
projection,
which
is
taking
this
curved
surface
and
flattening
it
out,
and
this
is
of
course,
we've
talked
about
this
with
the
with
the
axolotl
data,
where
we're
trying
to
take
flat
images
and
we're
trying
to
wrap
them
around
a
sphere.
This
is
the
opposite
process
or
you're.
Taking
this,
that's
I
don't
know
if
it's
a
sphere,
it's
like
an
elongated
beam
and
flattening
it
out.
A
So
this
is
a
similar
problem,
so
if
this
is
something
that
might
be
interesting
in
that
respect
and
then
the
this
dynamics
of
cell
intercalations,
so
you
have
these
what
they
call
t1
transition.
That
kind
of
show
up
that
looks
like
in
this
six
cell
rosette
over
time
and
they
show
what
that
looks
like.
A
B
A
So
this
is
an
example
of
something
that
looks
like
a
c
elegans
embryo,
but
it's
actually
like
this
idea
of
a
divergent
generative
integration
hypergraph
that
you
would
see
in
like
a
c
elegans
like
embryo,
so
I
drew
this
out
one
day
and
I
was
sitting
there
and
I
did
it
in
pencil
first
and
then
I
kind
of
worked
out
some
of
the
details
on
this
later,
so
this
is
kind
of
an
early
draft.
What
this
looks
like,
but
this
is
good
enough.
A
I
can
explain
what
this
looks
like,
so
you
start
with
a
single
cell
and
you
go
to
like
a
four
cell
embryo
which
doesn't
really
have
much
differentiation
when
you
get
to
the
eighth
cell
embryo,
you
get
a
germ
cell,
so
that
splits
you
into
two
networks,
the
germline
can
form
a
network,
but
it's
not
really
there's
not
much
there.
It's
just
kind
of
like
germ
cells,
but
they're,
segregated
out
from
this
other
network,
which
is
the
somatic.
B
A
So
the
point
is
that
there's
a
division
point
here
in
terms
of
your
networks.
Now
each
one
of
these
nodes
represents
a
number
of
cells
that
are
that
form
a
network.
So
this
is
what
they
call
a
hypergraph.
So
this
is
seven
cells
here.
This
is
actually
a
network
within
this
node.
This
is
14
cells.
This
is
28
cells
and
the
germ
line
continues
to
increase,
but
in
a
different
network.
So.
A
Interact
at
all,
then,
when
you
get
to
28
cells
here
in
this
putative
organism,
you
split
this
splits
off.
So
now
you
start
to
get
neural
cells
that
start
to
differentiate,
but
you
still
have
your
embryo
network
so
that
that
network
expands
in
size
as
well,
because
you
have
cell
divisions.
So
now
you
have
a
network
of
55
developmental
cells
or
somatic
cells
and
three
neural
cells
and
they
form
a
very
small
connectome.
A
But
then
you
get
begin
to
start.
You
know
you
start
getting
more
and
more
neurons
and
you
continue
to
see
more
somatic
cells
and
you
get
this
interesting
phenomena
where
you
can
get
cells
that
at
one
stage
were
part
of
the
somatic
network
or
this
embryo
network
that
then
differentiate
into
neurons
and
join
this
network,
so
they
actually
switched
to
a
different
sub
network,
and
so
you
can
see
there
isn't
any
interchange
between
the
connectome
and
the
embryo
network.
A
It's
just
in
this
direction
from
the
mro
network
to
the
neural
connectome,
and
so
you
can
see
that
over
time
you
get
and
then,
when
you
go
to
194
cells
in
the
summary
network,
you
get
a
further
bifurcation,
which
is
basically
where
you're
starting
to
get
like.
You
know
different
parts
of
the
nervous
system.
So
in
this
case
it
would
be
in
the
head
and
maybe
in
the
tail
of
c
elegans
or
whatever
this
organism
looks
like
and
then
you
could
interchange
between
those
networks
perhaps
and
you
get
defections
to
the
connectom.
A
The
idea
is
that
it's
like
a
lineage
tree,
but
it's
actually
a
set
of
networks
that
differentiate
over
time,
and
so
you
can
characterize
the
entire
embryo
with
a
series
of
sub
networks,
and
you
can
say
things
about
the
interactions
between
them
as
those
networks.
As
those
cells
diverge,
they
form
divergent
networks,
but
they're
also.