►
From YouTube: DevoWorm #19: Community Period for GSoC, four GNN papers, Plasmodium collectives and soft materials.
Description
Beginning of the GSoC Community Period (week 1). Open Source and Open Science reading repository and a tour of OpenWorm Foundation (its history and projects on the Slack team). Source datasets for the D-GNN project and a discussion of the GNN literature. Vortices, sorting, and soft material principles in Plasmodium collectives. Attendees: Susan Crawford-Young, Karan Lohaan, Harikrishna Pillai, Jiahang Li, and Bradly Alicea
A
C
A
So
it's
a
holiday,
I
know
in
the
u.s,
so
I
don't
know
how
many
people
are
going
to
show
up.
I.
E
C
Right
now,
what
I'm
trying
to
do
is
found
an
algorithm
on
github
which
works
on
neural
networking,
so
it
what
it
does
is
it
converts
the
2d
images
of
an
object
which
is
taken
which
are
taken
from
different
views
into
a
3d
object.
So
I'm
trying
using
the
that
algorithm
I'll,
be
training
the
algorithm
producing
the
result.
Then
I'll
see
what
happens.
D
A
A
All
right
all
right,
yeah
and
share
the
link
when
you're
ready.
I
guess
we
can
get
started.
This
is
waiting
for
a
few
more
people,
but
I
guess
we
can
do
this
now,
so
I
wanted
to
go
over
some
things:
the
community
period,
so
first
of
all
welcome
to
the
community
period,
if
you're
watching
this
later
or,
if
you're,
watching
it
now
for
google
summer
of
code,
so
our
community
period
actually
started
last
week
and
I've
been
in
the
slack
with
some
resources
for
people.
A
I
want
to
make
sure
that
both
projects
are
covered,
but
I
also
have
some
general
materials
and
I'll
go
over
what
those
are
and
then
okay
here
we
go
and
then
we'll
go
through
some
of
these
other.
I
have
some
other
topics
to
cover
today
and
we'll
see
who
shows
up.
So
this
is
the
nerf
link,
so
this
is
okay.
So
this
is
a
link
to
github.
A
Oh,
this
is
for
like
a
mixed
reality
or
a
virtual
reality,
volume
rendering
sort
of
thing,
and
then
you
can
also
use
other
things.
Yeah.
A
Okay,
yeah
yeah.
This
looks
good,
so
this
is
oh.
This
is
an
nvidia
project
or
product
yeah.
Okay,
I
think
I've
heard
of
this
before
just
trying
to
like
kind
of
remember
where
I'd
heard
the
name
yeah
nerf.
A
So
this
is
okay,
sparse
input,
photos
from
different
angles,
so
you're
taking
the
same
image
from
different
angles:
you're
building
a
three-dimensional
mesh
and
you're
getting
360
degree
views.
Okay,.
B
C
C
So
the
thing
is
we'll
be,
you
know,
dividing
like
at
least
I
think
I'll
be
doing
the
modeling
part
and
hari
krishna.
C
How
to
finally
display
the
3d
object?
You
know
on
the
website,
so
for
that
you
know
we
discussed
potential
options
so
so
far
hero
will
be.
You
know
most
probably
going
forward
with
the
aeropool
script
only,
and
I
think
we
still
if
we
can
get
a
better
fit
for
the
algorithm
that
we
have
to
use
in
this.
We'll
probably
try
that,
but
so
far
you
know
we'll
still
continue
with
the
you
know,
methods
that
I
had
outlined.
C
So
that
is
the
any
anything
else
you
want
to
add
krishna.
No,
that's
it.
I
guess
those
are
all
the
questions
yeah.
So
for
you
like
we
keep
on.
You
know,
checking
out
new
things
as
well,
like
any
rf
was
kind
of
an
interesting.
You
know
idea
that
we
wanted
to
investigate
as
to
how
it
would
fit
our
model
here.
E
Yeah,
I'm
good,
I'm
sorry
that
I
could
not
attend
the
meeting
the
last
week,
yeah,
I'm
quite
busy
this
month,
yeah
yeah
we've.
I
mean
we've
already
finished
a
paper
and
we've
submitted
to
another
conference
which
is
similar
to
the
learning
on
graph
yeah.
Okay,
the
new
ribs
2022.
Maybe
you've
heard
that
yeah
yeah
yeah.
D
A
Good
all
right!
Well,
that's
great!
So
why
don't
we
get
into
the
I
wanted
to
go
through
some
community
period,
things
for
gsoc,
so
all
of
you
got
the
messages
in
the
slack
okay.
So
let
me
go
over
some
of
this,
so
this
is
our
slack.
This
is
well
the
open
worm
slack.
So,
like
I
said
last
week,
you
know
the
community
period
is
about
like
investigate,
looking
through
the
open
arm
foundation
more
generally
and
seeing
what's
there.
B
A
There's
a
lot
of
stuff
going
on
in
open
worm.
Besides
us,
you
know
you
have
like
c302,
which
is
a
group
that
models
a
connectome
and
they
do
a
lot
of
stuff
with
like
they
have
a
software
where
they
model
the
302
neurons
of
the
c
elegans
nervous
system.
A
So
you
know
they're
doing
a
lot
of
like
pull
requests
in
this
channel.
This
is
a
c302,
beginners
and
all
of
the
history
here
and
that
there's
also
a
place
to
discuss
things
about
c302.
A
A
This
is
a
good
summary
of
it
here
in
the
github
readme,
so
there's
a
github
for
openworm
and
you
don't
have
to
join
that,
but
just
to
let
you
know
that
they
it
exists
and
that
there's
a
there
are
a
lot
of
contributions
being
made
there
and
that
they're
doing
a
lot
of
things
with
you
know
different
aspects
of
modeling,
c
elegans
they're
doing
a
lot
of
things
in
python.
So
you
know
we're
probably
going
to
do.
A
You
know
a
lot
of
things
here
and
we're
going
to
be
not
contributing
directly
to
the
open
worm,
github
repo.
But
it's
you
know
going
to
be
something
that
maybe
people
will
be
interested
in.
A
Then
there
are
other
groups
like
channel
worm
where
they're
modeling
ion
channels
in
the
cells
themselves.
So
all
cells
in
c
elegans
have
ion
channels.
The
neurons
have
ion
channels
that
produce
electrical
activity
that
is
similar
to
it's,
not
exactly
a
action
potential,
but
it's
similar
to
an
action
potential
and
you
know
the
so
the
nervous
system.
A
It
is
very
ion
channel
dependent,
but
you
also
have
other
cells
that
have
ion
channels.
It's
very
important
to
know
what
those
ion
channels
are
doing
to
really
model
what
the
cell
is
doing
and
the
behavior
of
the
cells.
A
So
that's
that
channel
we
have
an
education
channel
which
is
somewhat
dormant.
It's
been
but
we've
in
the
past.
We've
had
a
lot
of
educational
opportunities.
We
have
educational
materials
for
different
things.
This
is
the
hermaphrodite
nervous
system
in
c
elegans.
So
that's
like
the
the
you
know,
99
of
all
worms
and
c
elegans
are
hermaphrodites,
so
this
is
an
overview
of
the
nervous
system.
You'll
find
a
lot
of
interesting
things
in
this
channel.
Incf
has
some
training
space
materials
that
get
repurposed
in
open
worm
and
things.
B
C
C
A
Like
our
community
and
how
people
are
learning
about,
you
know
c
elegans,
there's
also,
you
know,
like
engineering
which
deals
with
some.
You
know
software
engineering
issues
openworm
really
started
as
sort
of
a
you
know.
It
had
a
heavy
emphasis
on
software
engineering,
because
people
were
trying
to
engineer
all
these
different
pieces
of
software
that
are
put
together
to
form
open
worm
and
form
this.
You
know
where
you're
simulating
different
parts
of
the
worm,
so
there's
that
aspect
of
it,
but
there
are
also
these
different
systems
level
aspects.
A
So
there's
movement
analysis,
which
is
where
you
have
people
talk
about.
I
think
I
talked
about
this
last
week,
where
they're
doing
movement
tracking
of
the
worms
so
they're,
looking
at
a
worm
through
a
microscope,
they're
looking
at
it
moving
along
this
agar
surface
and
they're
kind
of
characterizing
the
movements,
because
c
elegans
has
a
number
of
stereotype
movements
that
you
can
record.
A
And
you
know
if
you
have
the
right
technology,
they
use
what
they
call
a
worm
trap:
a
movement
tracker
which
is
like
the
cell
tracking
technology
that
we've
talked
about
here,
but
it's
just
tracking
the
actual
worm
and
its
movements.
So
it's
it's.
They
actually
have
data
sets
that
are
available
and
those
data
sets
might
be
of
use
to
you
as
you
work
on
some
of
your
models
and
validating
them,
I'm
not
so
sure
about
digital
microspheres,
but
definitely
the
graph
neural
networks
project.
A
You
might
look
into
some
of
these
like
the
tipsy
tracker
software
and
they
also
have
something
at
movement.openworm.org.
A
B
A
We
also
have
have,
in
the
past,
had
a
lot
of
focus
on
building
robots,
so
there
there
was
an
early
version
of
an
open
worm
robot
that
was
like
a
hard
robot.
It
looked
like
a
like
a
roomba
vacuum
cleaner.
A
It
was
a
round
robot
that
would
move
around
and
you
know
was
programmed
with
the
algorithms
that
were
developed
from
the
open
worm,
connectome
and
other
types
of
tool,
software
tools
that
were
developed
in
open
worm
and
they
were
able
to
drive
this
robot
to
do
different
behaviors,
and
so
that
got
a
bit
of
a
media
attention.
And
if
you
look
it
up
on
online,
you
look
up.
Openworm,
plus
robot
you'll
find
a
number
of
articles
on
this,
and
a
lot
of
them
are
pretty
spectacular
in
the
sense
that
they're,
not
really.
A
You
know
it's
just
a
robot
that
you
put
software
into
and
it
behaves
in
a
certain
way.
It's
not
like
going
to
become
conscious.
So,
but
in
any
case,
that's
that's
where
the
project's
been
more
recently
they've
worked
on
building
soft
robots
that
have
you
know,
and
I
know
tom
portages
has
been
in
this
group.
A
number
of
times
where
he's
been
in
this
group
historically,
he
hasn't
really
attended
meetings
much
anymore.
B
A
He's
done
work
with
robotics
as
well,
so
you
know
he's
really
good
at
like
building
robots
and
building
software
for
like
artificial
intelligence,
adaptive
systems
type
applications.
So
you
know
this
is
something
that
you
know
you
have
a
worm.
A
B
A
Another
channel
you
might
want
to
check
out.
You
have
to
go
back
quite
a
ways
to
find
some
of
the
meteor
work
here,
but
you
know
that's
the
thing
with
slack
everything
is
archived.
A
Of
course
the
problem
is
it's.
You
know
you're
trying
to
find
interesting
things,
so
I
mean
there's
just
a
lot
of
if
you
really
want
to
get
into
the
guts
of
like
how
people
build
robots.
You
know
this
has
been
chronicled
in
this
robot's
channel.
It's
really
kind
of
interesting.
A
So
that's
that's
our
slack.
I
just
wanted
to
bring
that
up,
because
I
think
you
know
I
want
to
encourage
you
to
go
outside
the
bounds
of
diva
worm
for
a
couple
weeks
and
see
what's
out
there
in
our
organization,
and
you
know
if
you
find
someone
who's,
you
might
find
interesting.
You
want
to
talk
to
them.
You
know
drop
them
a
line,
you
know,
send
them
an
email
or
send
them
a
slack
message,
or
I
can
get
you
in
touch
with
someone
specific.
A
If
you
want
to
talk
to
them,
you
know
it
might
be
better
for
me
to
contact
them
because
I
can
they.
They
know
me
more
than
they
know
you,
so
they
might
be
more
likely
to
get
in
touch
with
me,
but
there
are
a
lot
of
different
things
going
on
in
the
open
world
foundation.
A
They're,
you
know
they.
They
vary
because
it's
an
open
source
organization
there's
a
lot
of
variability
in
when
people
are
working
on
things.
So
you
know,
diva
worm
has
been
pretty
constant,
but
like
robotics,
you
know
you'll
get
people
who
are
interested
in
robotics,
they
might
start
working
on
a
little
project
and
then
they
might
produce
something
really
nice
and
then
they
might
drop
it
for
a
while
and
then
come
back
to
it.
A
A
Nice,
nice,
that
you
could
make
it
today.
F
Yeah,
it's
it's
a
bit
busy
here,
but
yeah
I'm
taking
a
break
so.
A
Okay,
good
to
have
you
so
yeah,
that's
and
again,
this
is
you
know
this
is
the
community
period,
so
you're
you're,
encouraged
to
you,
know,
explore
the
bounds
of
your
organization
and
your
organization
is
open.
Worm
diva
worm
is
within
open
norm.
It's
a
project.
You
have
other
projects
doing
other
things.
Our
focus
is.
A
And
the
development
of
the
worm,
so
we
have.
A
Areas
as
well,
so
the
next
part
I'll
go
to
is
I'll,
go
to
the
divorm
channel,
specifically
and
I'll
talk
about
some
of
these
links
that
I
put
in
in
the
chat.
So
this
is
a
drive
of
things
that
their
resources
for
open
source
and
open
source
project
management
readings.
A
So
I
have
this
library
of
things
that
are
based
around
learning
about
open
source
software
and
how
it's
made
and
how
to
manage
it
and
then
some
other
resources
in
open
source
project
management,
and
so
I
encourage
you
to
read
through
these
I'll
go
through
some
of
them
in
a
minute,
and
there
are
other.
You
know
these
other
channels,
like
movement,
analysis
and
robotics
might
be
useful
for
you
as
well.
A
Now
this
person
here
in
these
two
videos
that
I
posted
his
name
is
stephen
larson
and
about
ten
it's
been
more
than
ten
years
now
him
and
a
couple.
Other
people
started
open
worm.
They
got
together
online
and
they
said
this
is
great.
You
know
they
have.
This
idea
about,
like
stephen,
for
example,
got
his
degree
in
artificial
intelligence.
A
He
got
a
master's
degree
from
mit
and
artificial
intelligence
and
then
he
got
a
degree
in
neuroinformatics
or
neuroscience
at
uc,
san,
diego
or
the
us,
and
he
was
finishing
that
up
and
there
are
a
couple
other
people
in
europe
working
on
the
neuron
package,
which
is
a
package
to
simulate
neurons
and
they
were
working
on
these
different
projects
and
stephen,
you
know,
got
in
touch
with
them
and
said
ray.
I
think
he
made
a
call
and
asked
you
know.
People
are
interested
in
simulating
c
elegans.
A
A
It
has
a
very
distinct
and
sort
of
deterministic
developmental
tree
lineage
tree
where
all
the
cells,
you
know
what
they're
gonna
become,
and
it's
very
well
characterized,
and
so
he
said.
Wouldn't
it
be
nice
if
we
could
build
a
digital
model
of
c
elegans,
and
so
everyone
was
excited
and
we
started
off,
and
I
didn't
really
actually
come
into
this
community
until
about
2014
or
2015..
A
So
this
is
this
first
one
is
an
incf
talk,
so
this
was
at
incf
and
he
was
talking
to
that
group
about
openworm,
and
this
is
the
subtitle-
was
a
community
developed
in
silicon
model
of
c
elegans,
so
this
was
back
in
2014,
so
this
is
going
to
be
a
bit
dated,
but
he
talks
about
some
of
these
different
things
that
they're
trying
they're
trying
to
do
at
the
time.
So
this
is
their
more
of
their
vision
for
the
future.
A
This
is
good
because
it
lays
out
a
road
map
for
what
they
want
wanted
to
achieve,
and
some
of
those
things
have
been
achieved,
and
some
have
not
this
talk
here.
This
ted
talk
is
a
little
bit
different.
This
is
kind
of
a
focus
on
how
the
c
elegans
or
the
is
you
know
it's
a
common
name
is
roundworm
is
our
best
bet
to
unlock
the
secrets
of
the
brain.
A
A
But
it's
also,
you
know
you
can
tell
some
things
about
the
human
brain
or
mammalian
brains
by
you
know.
Looking
especially,
if
you
do
the
molecular
biology
you
look
at
the
genes
being
expressed
or
you
look
at
some
of
the
genes
that
exist
in
you
know
mutated
genes
that
exist
in
some
strains
of
c
elegans.
A
One
other
thing
c
elegans
is
known
for
is
this
aspect
of
having
these
defined
newtons,
so
you
can
have
a
c
elegans
that
has
a
specific
mutation
that
it
always
carries
you
just
let
it
reproduce
and
it
always
carries
this
mutation
and
you
can
compare
it
to
a
wild
type,
which
means
that
it
has
like
a
normal.
You
know
in
the
wild
sort
of
genotype,
and
so
it
carries
a
lot
of
mutations,
but
they're
not
what
they
call
the
fine
mutations.
So
there's
just
genetic
diversity.
A
So
if
I
have
a
mutation
in
some
gene,
it
might
mean
that
I
don't
have
a
certain
neuron
or
that
I'm
moved
very
slowly
or
that
I
don't
form
a
full
cuticle
things
like
that,
and
that's
useful
for
just
kind
of
interesting
to
see
how
mutations
work
on
the
phenotype.
But
it's
also
very
interesting
for
disease.
A
You
can
create
define
mutants
for
different
disease,
neural
diseases
that
you
might
see
in
humans
or
in
other
mammals,
and
you
can
understand
things
like
aging.
You
can
understand
other
types
of
processes
that
go
on,
so
this
is
something
that's
very
useful
in
modeling
the
worm
in
this
way
as
a
digital,
almost
like
a
digital
twin,
they
didn't
use
that
terminology
back
then,
but
it's
something
that
would
be
sort
of
analogous
to
this
is
very
useful,
and
that
being
said,
the
digital
models
that
exist
now
are
you
know,
they're
not
complete.
A
They
don't
really
have
a
genome,
they
don't
have
a
genotype,
but
they're,
actually
quite
good
at
the
cellular
level,
in
terms
of
understanding.
What's
going
on
there
having
virtual
models
of
the
cells,
and
things
like
that,
then
this
this
link
here
is
a
link
to
some
materials
that
we
generated
during
an
open
house
back
in
2016.,
so
we
had
an
open
house
and
it
kind
of
went
through
a
lot
of
the
different
projects.
So
there
are
materials
for
each
project.
A
There's
some
paper
background
readings,
there's
a
talk,
maybe
about
five
or
seven
five
to
seven
minute
talk
on
these
different
projects
and
it
kind
of
recaps
sort
of
the
origins
of
each
project.
A
Now
it's
a
bit
dated
it's
from
2016,
so
some
of
these
projects
have
moved
on
since
this
time
you
know
in
terms
of
its
pr
progress,
but
you
know
this
is
a
good.
I
think,
guide
post
to
some
of
the
things
have
been
that
have
been
done
in
open
world.
D
A
I
don't
I
don't
know
why
I
didn't
do
it
before,
but
the
first
is
that
there
was
a
special
issue
of
the
royal
society
b,
which
is
the
proceedings
of
the
royal
society
b,
and
that
was
in
2018
and
the
open
worm
foundation
was
invited
by
the
royal
society
to
give
like
a
seminar
at
in
london
on
openworm,
and
so
there
was
a
lot
of.
There
was
a
lot
of
excitement
around
that,
and
there
were
a
number
of
talks.
A
Some
the
people
were
inside
open
worm
somewhere
outside,
they
were
experts
in
the
field
and
they
talked
about
a
lot
of
things
about
the
connectome
and
about
computational
modeling
of
the
worm,
and
that
ended
up
becoming
a
special
issue
of
the
royal
society
b.
So
I'm
going
to
send
a
link
to
the
journal,
the
the
special
issue.
I
can
get
access
to
the
papers
if
you
want
them.
A
Actually,
I
think
I
have
a
library
of
them
I'll,
send
you
the
library
of
the
papers
as
well,
and
then
there
was
some
other.
You
know
there
was
some
like
live
activity
at
the
workshop
that
I
have
a
summary
of
I'll.
Send
you
that
as
well,
so
that's
that
the
one
thing,
the
one
link
that
I
wanted
to
share
with
you
and
that's
a
little
bit
more
recent
update
on
on
some
of
this
work.
A
The
second
link
is
that
there's
a
docker
container-
and
this
is
something
that's
updated
up
to
this
day-
that
features
all
of
the
programs
and
open
one.
So
I
don't.
This
is
a
little
bit.
I
don't
know
if
you're
gonna
be
able
to
run
this
or
not
on
your
computers,
but
there's
a
docker
container.
A
You
can
download
from
our
github
from
the
open
arm
github
and
that
openworm
github
link
has
this
docker
container
and
you
open
it
up
on
your
machine
and
I
think
the
the
best
way
to
do
it
is
to
use
a
linux
machine.
But
I
think
it
can
also
run
on
windows,
I'm
not
sure,
but
it
basically
opens
up
and
it
runs
all
the
simulations.
A
A
So
it's
you
know
it's
really
supposed
to
be
instructive
of
the
whole
project
and,
like
kind
of
the
whole
scope
of
the
worm
now
diva
worm
has
never
been
able
to
put
anything
in
this
docker
container.
Unfortunately,
so
we're
not
represented
in
that.
But
I've
tried
to
get
in
the
past
people
to
contribute
something.
I've
always
wanted
to
build
like
a
three-dimensional
model
of
the
embryo,
but
we've
never.
A
There
something
that
was
suitable
for
that.
So
that's
you
know
it's
probably
not
what
you're
doing
in
your
projects,
but
you
know
we
can
give
it
some
thought
as
to
maybe
how
to
at
maybe
later
in
the
project
period.
When
we're
you
know
working
on
the
projects,
you
know
how
we
might
use
some
of
these
things
later,
and
so
it
may
be
that
we
make
something
that
will
apply
to
that
docker
container.
C
Hosting
I
think
like
we
could
explore
it
as
an
option
like
as
an
additional
option,
because
then
things
will
be
in
our
hand,
and
you
know
giving
heroku
the
using
as
a
platform.
You
know
it
has
its
own
limitations
with
the
file
sizes
and
the
extent
or
the
number
of
requests
you
can
have
per
month.
So.
A
A
Yeah
so
yeah
any
other
questions.
A
Okay,
yeah,
so
that
that's
I'm
going
to
talk
about
with
that,
then
I
want
to
go
over
to
some
of
the
other
resources
that
we
have
here.
So
we
have.
A
A
This
is
something
that
we
wrote
for
is
sort
of
a
technical
paper
for
the
devolearn
release.
So
this
is
where
we're
using
deep
learning
to
analyze
microscopy
data.
So
here's
some
microscopy
data
that
was
analyzed
and
you
know
basically
the
embryo
is
segmented
and
you
get
these
different
outputs
of
of
defined
cells
and
their
morphology
and
it's
a
pre-trained
model.
So
it's
you
know
something
that
we've
had
to
use.
A
Data
sets
to
benchmark,
and
one
of
the
things
I
wanted
to
point
out
here
is
that
there's
a
table
in
this
paper,
it's
table
one
and
this
table
has
the
different
data
sets.
So
this
is
these
are
different
data
sets
that
are
available
online.
A
A
This
is
a
good
data
set
for
looking
at
nuclei.
In
their
position
in
the
embryo,
so
this
is
something
that
has
data
that
basically
features
a
fluorescent
nucleus
or
a
fluorescent
centroid
for
each
cell
over
time,
and
that
gives
you
a
data
set
of
points
that
you
can
then
generate.
You
know
distances
and
you
can
generate
a
graph
from
that.
A
A
You
just
download
them
yeah
and
there's
another
source
that
I'll
send
you
it's
the
systems
biology
database,
which
has
some
other
embryo
data,
sets
in
it
as
well
I'll
have
to
down
I'll,
have
to
send
that
to
you
guys
after
the
meeting.
A
But
this
is
another
database
that
we
use.
It's
run
out
of
writing
in
japan
and
they
have
a
number
of
different
data
sets
that
are
usually
images
like
this
that
you
can,
then
you
know
take
in
and
you
know
find
features
on
it
and
then
use
those
features
as
input
data.
So
you
know
these.
Are
there
a
lot?
There's
a
lot
of
data
online.
For
this
we
have
something
called
the
devo
zoo,
where
we
have
short
microscopy
movies.
A
A
A
We've
had
people
take,
you
know,
videos,
you
extract
the
images
frame
by
frame,
and
then
you
put
those
images
in
an
algorithm
and
you
do
some
segmentation,
some
basic
segmentation
to
get
a
general
idea
of
where
the
different
features
are
in
the
embryo
and
then,
if
you're
doing
something
like
graph
neural
networks,
then
you
need
to
get
the
data
sort
of
you
need
to
build
a
reference
frame,
and
then
you
can
build
these
networks
because
you
need
to
know
kind
of
where
they
are
in
space,
but
also
where
they
are
in
time,
because
your
your
embryo
is
the
number
of
cells
in
the
embryo
is
growing
over
time,
so
that
number
of
cells
is
going
to
increase.
A
The
reference
frame
is
going
to
change.
You
need
to
normalize
that
spatial
reference
frame.
You
need
to
keep
track
of
the
time.
You
know
the
time.
So
each
frame
in
the
movie
is
is
a
time
point
in
development,
and
then
you
know
you
can
build
a
model
from
those
two
pieces
of
information,
the
the
actually
it's.
I
guess
it's
two
dimensions
of
space
in
a
lot
of
these
movies
because
you
don't
have
a
microscopy
stack.
This
is
all
just
that
one
frame
focal
frame,
so
it's
just
two
dimensions.
A
But
then
you
have
this
third
dimension
of
time
and
then
from
that
you
can
build
actually
a
three
dimension:
three
dimensions
of
space,
because
we
know
that
the
cells
have
a
certain
z
axis
location,
but
that's
something
that
we
have.
You
know
we
can
label
that
after
the
fact,
because
we
have
that
data
in
some
of
these
other
data
sets.
So
there's
a
cell
position.
Data
set
here
that
has
xyz
normalized
position
for
these
cells.
So
you
know
this
is
something
we
can
work
out
over.
A
Maybe
we
should
start
doing
this
now
in
fact,
figuring
out
these
coordinate
systems
and
how
to
get
the
cells,
you
know
arranged
or
organized
it
shouldn't
be
too
hard,
but
it's
going
to
take
some
playing
around
with
the
data.
I
just
wanted
to
point
out,
though,
that
we
have,
I
think
all
the
data
sets
you'll
need
to
do
this,
so
just
we
just
need
to
like
kind
of
go
back
and
forth
on
how
this
is
going
to
look.
A
Then
I
wanted
to
bring
your
attention
to
the
github
repos,
so
I've
had
all
of
you
join
our
get
a
diva
worm,
github
organization,
so
this
is
diva
worm.
It's
not
actually
an
organization,
it's
an
account,
but
it's
fine
because
it
basically
gives
you
permissions
for
this
repository.
A
So
this
is
the
gsoc
2022
repository
here.
The
the
directories
that
I
want
you
to
use
for
pushing
your
work,
maybe
weekly.
If
you
have
work
on
your
own
repository,
you
know
if
you
could
clone
it
to
this.
That
would
be
good
because
you
have
permissions
now
you
don't
need
to
issue
a
pull
request.
A
You
can
just
push
it
up
to
this
place.
However,
you
know
it's
nice
to
be
able
to
review
pull
requests
if
there's
something
that
you
knew
me
to
review.
So
keep
that
in
mind.
I
I
want
you
to
update
these,
maybe
once
a
week,
basically
anytime,
you
do
work
and
I
know
you
all
were
use
github.
So
that's
not
a
problem,
but
I
want
you
know.
A
So
these
are
public
repositories
and
I've,
given
each
project
its
own
directory,
so
devo
graph,
which
is
jia,
hong
you'd,
push
your
things
to
divo
graph
gnn's
developmental
networks
would
be
wataru
and
then
the
digital
microsphere.
A
A
I
just
want
to
make
sure
that
your
work
is
in
separate
directories.
That's
all
so,
whoever
makes
the
first
push
to
whatever
directory
you
make.
Your
push
to
the
other
person
will
pick
the
other
one
and
then
in
the
readme.
Maybe
we
can
make
that
clear
as
to
who's
pushing
to
what
directory
just
so
we
can
keep
everyone's
work
separate
and
I
can
evaluate
it
and
we
can
go
through
it,
but
yeah.
A
So
then
we'll
have
this
as
a
historical
repository
and
then,
after
you
know,
towards
the
end
of
the
coding
period,
we'll
do
like
a
review
of
work
and
then
we'll
you
know,
try
to
work
on
getting
it.
We
have.
Google
wants
you
to
have
something
that
you
you
submit
to
them.
That
has
some
sort
of
executable
in
it.
It
could
be
an
executable
file.
A
I
think
it
could
be
a
docker
container,
it
could
be
a
set
of
digital
notebooks,
but
it's
got
to
be
something
that
they
can
run
and
see
that
it
works.
A
So,
okay,
so
gia
hong
has
a
question
hi
bradley.
Could
I
use
some
time
in
the
meeting
to
introduce
some
resources
related
to
gnn
and
biological
analysis?
Yes,
so
go
ahead.
Jia
hung.
E
Oh
probably
I
need
I,
I
have
to
quit,
create
a
meeting
and
I
will
come
back
again:
okay,
yeah
all.
D
A
After
jiahan's
done,
I'm
going
to
go
through
some
of
these
readings
on
well,
I'm
going
to
just
go
really
quickly
over
the
open
source
readings
and
I
I.
F
A
A
A
Email
each
other
on
that
that'd
be
good.
So
do
you
hung,
were
you
ready
to
go.
A
A
If,
sometimes,
if
you
share
the,
if
you
share
the
tab
instead
of
the
whole
screen,
it'll
work.
E
Yeah
and
I've
served
some
resources
this
week.
Yes,
it's
about
the
graphic
network
and
biological
analysis
yeah
and
first
of
all,
I
want
to
introduce
some
conference,
because
I
I
remember
that
you
mentioned
the
learning
on
graph
right
several
weeks
ago,
yeah,
yeah
and
yeah
yeah.
E
It
will
be
probably
the
october
and
november,
I
don't
remember
exactly,
but
yes,
I
think
we
still
have
time
to
prepare
for
these
three
conferences,
yes
and
actually
such
as
the
this
one
and
this
one-
and
there
are
some
other
more
conferences
for
our
work
in
the
next
year.
Yes
and
yeah.
This
is
the
first
part
I
mean
we
have
a
lot
of
conferences
we
can
consider
if
we
want
to
submit
some
works
to
these
conferences,
yes,
yeah
and
also
there
are
yeah.
There
are
some
other
things
such
I
want
to
introduce.
E
The
second
part
is
about
graph,
your
natural
can
its
application
on
biological
analysis.
Yes,
and
the
second
part
is
I've
surveyed
some
papers,
and
I
will
put
these
names
in
the
chat.
E
The
first
one
is
graphql
natural
for
cell
tracking
in
microscopy
videos-
I,
it
is
probably
related
to
our
work.
Yes,
because
we
are
focusing
on
the
developmental
analysis
of
embryos,
yes
and
actually-
and
this
paper
is
trying
to
introduce
how
to
use
dynamical
temporal
gcn
to
analyze
or
track
the
sales
such
as
they
use.
E
Gcn
to
I
mean
the
group
convolutional
network
to
aggregate
information
for
each
I
mean
the
each
snapshots
of
a
video
and
they
use
such
as
the
recurrence
in
your
networks
or
lstm
to
aggregate
information
across
different
times
or
different
snapshots.
E
E
And
this
paper
is
utilized.
This
is
what
is
utilized
to
solve
some
issues
regarding
the
traffic
prediction,
but
actually,
I
think
the
traffic
prediction
is
very
similar
to
the
steel
tracking.
Actually,
if
you,
if
you
have
a
dive
delve
into
this
paper-
and
it
is
very
similar
to
the
graphene
network
facilitation
tracking-
I
mean
the
tgcn
and
I
think
it
is-
they
are
very
similar.
So
I
would
I
would
not
introduce
the
tgcn,
but
I
think
maybe
you
would
like
to
take
a
view
of
this
paper.
E
A
E
Yes,
I'm
not
sure
this
paper
is
related
to
our
work,
yes,
because
it
is
work
to
try
to
how
to
say
this
project
the
discrete
space
of
a
graph
into
the
latin
continuous
space
such
as
the
hyperbolic
space,
such
as
other
spaces,
and
these
paper
things
we
can
find.
E
A
Yeah,
I
think
it
is
actually
yeah
yeah
go
ahead.
You,
oh
yeah,
I
mean
that's,
that's
one
possibility
you
know
once
you
build
a
network,
it's
not!
You
know
it's
kind
of
like
you're,
creating
this
normalized
space,
because
the
data
that
you're
going
to
get
is
over.
Maybe
several
embryos
or
you're
going
to
have
a
lot
of
displacement
in
the
in
the
images,
because
you
know
it's,
it's
kind
of
things
are
moving
around.
A
Another
thing
is
to
map
it
to
some
higher
dimensional
space,
and
that
would
be
like
you
know.
Hyperbolic
space
is
often
used
in
network
theory.
Actually
to
do
this,
especially
with
something
that
has
a
curvature
like
an
embryo,
where
you
need
to
sort
of
approximate
that
curvature.
But
you
don't
want
to
necessarily
put
it
like
on
a
sphere
or
something
I
mean
you
can
put
it
in
a
space
where
you
have
that
accounted
for.
A
Yeah,
I
think
definitely
because
you
do
have
some
inherent
geometry
here
that,
like
you
know
with
quran
and
our
krishnas
project,
they're,
taking
an
embryo
and
they're,
creating
a
sphere
they're,
taking
flat
images
and
they're
putting
it
on
the
surface
of
a
sphere
in
microscopy
images,
they're
two-dimensional
but,
like
I
said
you
know,
when
you
collect
microscopy
images,
the
stack
of
images
and
every
cell
you
observe
in
that
focal
plane
that
you
have
in
your
movie,
which
is
two-dimensional,
has
some
depth
so
that
image
has
depth
it's
just
that
you're
not
seeing
it
because
you
know
you're
at
different
focal
lengths
from
that
cell,
so
you
go
down
or
you're
looking
you're
looking
down
at
it.
A
A
So
you
know
having
that
sort
of
geometry
you
can
actually,
you
know,
have
a
model
of,
or
you
can
project
some
of
those
that
z
information
into
your
representation.
So
definitely
it
is
useful.
E
Yeah,
okay,
I
got
your
points.
Yes,
I
I
think
I
will
survey
more
papers
regarding
the
geometric
graph
neural
networks.
Yes,
you
analyze
this
yes
yeah
and
yeah.
I
read
another
paper
yeah.
You
have
introduced
this
one
to
me.
Remember
that
this
one
yeah
there
differentiation
process
is
a
spatial
network.
Yes,
I've
read
this
paper,
yes
and
I
think,
and
I
think
you're
trying
to
utilize
the
tools
of
natural
to
analyze.
E
Such
as
the
diffusion
of
cells
such
as
the
glee
consumption
such
as
the
cluster,
yes,
so
I
think,
because
I'm
trying
to
because
in
my
proposal
I
comes
to
a
build
title
that
can,
I
think
they
can
fill
the
graph
based
on
the
videos
of
embryos.
E
B
E
B
E
Yeah
yeah,
okay,
yes,
I
mean
yes,
I
I
remember
that
there
are
some
other
papers,
please,
for
the
moment,.
E
A
E
E
Think
it
is
important
to
build
a
network
so
that
you
can
use
this
network
analyze
tours
to
analyze
the
natural
itself
right,
yeah,
yeah,
so
yeah
so
yeah.
I
think
there's
a
lot
of
targets
of
the
gm
projects
in
the
gsong
yeah
yeah
and
yeah
yeah.
This
yeah
I'm
done.
This
is
all
the
parts
I
want
to
share
today.
A
Okay
well
yeah.
Thank
you.
That's
very
good,
very
thoughtful
going
through
all
these
different
resources
and
I
definitely
think
yeah.
So
I
mean
one
of
the
things.
Of
course
we
want
to
have.
Is
that
geometric
aspect
then
also
the
temporal
aspect,
which
is
you
know,
because
you
have
things
in
time
that
are
unfolding,
so
I
mean
that
there's
a
lot.
I
guess
this
is
a
very
hot
area.
A
There's
a
lot
of
there
are
a
lot
of
papers
coming
out
every
day
every
week,
and
certainly
you
know
it's
something
that
now
with
the
conferences
you
know,
oftentimes
you'll
have
people
doing
like
technical
work
on
some
technique,
and
so,
if
we're
doing
stuff
with
more
applicable
to
biology,
you
know
it
depends
on
how
how
innovative
the
actual
gnn
is
to
that
depends
on
whether
we
get
it
accepted
or
not
that
that
being
said,
that
doesn't
mean
that
it
won't
be
accepted
to
some
of
these
conferences,
because
you
know
it
is,
but
it
is
competitive,
so
you
know
get
it
like
if
we're
and
we're
jumping
ahead
here,
but
if
we're
thinking
of
submitting
it
somewhere,
you
know
that
submission
has
got
to
be
got
to
figure
out
how
to
make
it
so
that
they
really
see
the
value.
A
It's
not
that
it
doesn't
have
value
in
it.
It's
that
these
different
venues
have
different
sort
of
they
have
different,
like
you
know,
priorities
or
criterium,
for
what
is
what
they
want
to
see
in
their
pay
in
their
top
papers.
So
I
mean
you
know
a
lot.
We
can
borrow
a
lot
from
these
papers
and
then
apply
it
to
biology,
and
it
may
be
that
you
know
we
even
target
like
maybe
something
like
a
computational
biology
conference
or
journal
as
well.
You
know
that
might
actually
even
be
a
better
fit.
A
It
depends
on
how
you
know
how
much
progress
we
make
in
the
coding
period,
and
maybe
you
know
if
you
would
like
to.
I
know
that
quran
and
hari
krishna
are
working
together.
You
know,
maybe
you
could
get
in
touch
with
wataru
in
the
slack
and
see
where
he
is
because
it's
there's
a
lot
of
you
know.
I
want
to
make
sure
that
you
guys
aren't
doing
the
same
thing
and
doing
the
same
work.
Yeah
yeah.
A
Yeah
yeah,
but
I
think
with
something
like
this:
it's
like
you
know
we
want
to
have
like
complimentary
work.
Maybe
is
the
most
powerful.
So
if
there's
something
complementary,
you
two
can
do,
and
you
know
that's
that
I
think
then
you
know
when
it
comes
time
for
a
paper.
We
really
have
like
you
know
a
lot
of
ground
that's
been
covered
and
you
know
it'll
make
for
a
stronger
set
of
things
that
we
can
produce
from
this.
A
Okay:
okay,
now
I'm
going
to
go
over
a
couple
of
the
readings
that
jiahan
couldn't
share
on
his
screen
and
I'm
going
to
go
over
one
more
thing
in
the
reading
queue.
So
let
me
go
to
the
share
my
screen
here
and
I
will
talk
about
some
of
these
papers
a
little
bit
in
brief,
so
the
first
one
he
mentioned
was
graph
neural
network
for
cell
tracking
and
microscopy
videos.
This
is
on
the
archive,
so
this
is
where
they
present
a
novel
graph,
neural
network
approach
to
use
for
high
throughput
microscopy
videos.
A
A
So
the
next
one
is
tgcn,
which
is
a
temporal
graph
convolution
network
for
traffic
prediction,
and
this
is
for
traffic
forecasting
in
real
time.
This
involves
integrating
a
number
of
systems
such
as
urban
traffic
planning
traffic
management
traffic
control.
A
So
this
is
something
that
shows
sort
of
the
power
of
the
productive
power
of
these
networks.
So
a
tgcn
is
a
temporal
graph,
convolutional
network
model,
so
this
combines
graph
convolutional
networks
and
gated
or
current
units.
So
this
has
a
specific
type
of
topology
and
we'll
see
if
this
is
something
that
might
be
of
use.
A
A
The
next
one
is
gm
gcn,
which
is
a
geometric
graph,
convolutional
networks.
This
one
is
based
on.
You
know
it's
sort
of
specialized
for
geometric
features.
The
basic
idea
is
that
the
aggregation
on
a
graph
can
benefit
from
a
continuous
base
underlying
the
graph.
So
again
we
need
our
coordinate
system
for
our
embryo
data,
and
this
is
a
continuous
space,
that's
built
under
the
graph
that
gives
you
a
coordinate.
B
A
And
then
spatial
relationships
or
geometric
relationships
and,
of
course,
as
we
mentioned,
you
can
have
different
spatial
frameworks
and
other
things
that
will
you
know
the
sort
of
latent
space
that
allows
you
to
reconfigure
your
representation.
A
Finally,
there's
a
seductive
representation
learning
in
temporal
graphs.
This
deals
specifically
with
temp
the
temporal
aspect
of
what
we
want
to
do
and
temporal
graphs.
More
generally,
our
our
lineage
trees
are
actually
examples
of
temporal
graphs,
although
they're
not
really
what
they're
talking
about
here.
A
Can
you
be
used
to
efficiently
aggregate
temporal
topological
neighborhood
features
so
as
well
as
to
learn
the
time
feature
interactions,
so
temporal
topological
neighborhoods
are
basically
areas
of
like
geometric
areas,
topological
areas
of
the
images
or
of
the
data.
B
B
A
Is
something
that's
very
relevant
to
what
we're
trying
to
do
so?
This
is
just
you
know.
They
use
this
attention
mechanism
to
sort
of
develop
novel
approaches,
functional
time,
encoding,
approaches
to
this
problem,
and
so
this
might
be
very
useful
as
well.
A
Okay,
now
we're
going
to
go
to
one
set
of
readings
that
I
wanted
to
to
talk
about,
and
this
is
a
very
interesting
set
of
research
from.
I
think
it's
from
a
single
group.
I
can't
remember-
but
let's
look
at
this,
so
this
is
this
is
about
parasite,
collectives
and
you're
like
why
are
parasite
collectives
interesting
to
development,
and
the
answer
is
is
because
they
form
very
interesting
patterns
and
we'll
see
some
examples
of
those
patterns
in
a
minute.
A
So
this
is
one
of
these
sort
of
profile
articles
that
you
find
in
journals.
This
is
out
of.
I
can't
oh,
I
will
nature
physics,
and
this
is
called
the
power
of
parasite
collectives
and
so
they're
talking
about
these
organisms
called
plasmodiums
or
sporozoites,
and
so
these
are
different.
These
are
plasmodium
that
can
move
in
rotating
vortices
owing
to
their
chiral
shape
and
mechanical
flexibility.
A
So
chiral
shape
is
like
a
spiral
so
like,
if
you
think
of
corkscrew
pasta,
that's
a
sort
of
a
chiral
shape.
It
has
a
direction
that
turns
and
it
turns
upward
or
downward,
depending
on
how
you're
moving.
B
A
The
chirality
of
the
shape,
of
course
dna
is
also
chiral,
and
we
know
that
there's
a
handedness.
B
A
Dna,
so
it
means
that
there's
a
right-handedness
to
dna
and
and
also
amino
acids
have
this
same
sort
of
chorality.
There's
a
curvature
and
the
curvature
has
to
be
in
a
certain
direction
in
order
for
it
to
be
biologically
viable.
So
this
these
chiral
shapes,
are
important
in
biology.
A
A
So
we've
talked
a
lot
about
collective
motion
in
the
group
about
swarms
and
other
sort
of
collective
behaviors.
In
this
case,
they're
talking
about
parasites
that
behave
in
these
collectives
and
the
properties
of
the
parasites
themselves
actually
lead
to
these
sort
of
biophysics
so
to
perpetuate
itself.
A
Humans
pay
an
enormous
price.
As
we
know
from
human
history,
the
transmission
of
plasmodium
sporocytes
rely
on
their
ability
to
transition
from
their
mosquito
to
their
vertebrate
hosts.
So
they're
in
the
mosquito
mosquitoes
eats
vertebrate
blood
they're
inflict
affecting
the
mosquito,
and
then
they
get
transmitted
to.
A
To
a
vertebrate,
and
so
they
move
between
these
two
different
types
of
organisms,
these
two
different
groups-
quite
quite
effortlessly.
So
remarkably
only
a
few.
Tens
of
sporozites
are
inoculated
individually,
along
with
the
saliva
fluid
of
the
mosquito,
so
they
basically
have
a
very
small
founder
population
that
moves
into
the
host
or
into
the
into
the
infected
individual,
and
then
they
proliferate
from
there,
and
so
they
enter
the
blood
capillaries
and
continue
to
deliver.
B
A
The
authors
have
designed
a
clever
ex
vivo
setting
to
address
this
unsolved
issue,
so
they
can
actually
observe
these
dynamics.
You
know
you
put
like
on
your
covers
on
your
microscopy:
coverslip
you'll
put
some
buffer
and
then
you'll
you'll
drop
these
in
and
you'll
be
able
to
observe
them
moving
around
in
the
liquid.
A
Under
these
conditions,
the
authors
find
that
aggregated
sporozoites
from
rotating
vortices,
in
line
with
the
circling
patterns
of
plasmodium
previously
imagined
in
salivary
imaged
in
salivary
glands.
So
they
can
image
these
things
in
vivo,
but
you
can
see
these
ex
vivo
as
well
or
on
on
under
a
microscope.
A
Looking
at
several
vortices
at
once,
one
can
see
that
the
tightly
aligned,
crescent-shaped
sporozoites,
always
rotate
the
same
front
first
direction
around
the
common
center,
an
example
of
chiral
collective
motion.
So
they
have
this
chorality
to
the
motion.
It's
going.
You
know
in
in
sort
of
this
swarm,
but
the
swarm
has
a
very
distinct
structure
around
a
central
point.
A
These
vertices
are
reminiscent
of
other
spectacular
collective
motions
observed
in
certain
animals,
and
bacteria
rotating
vortices
often
arise
when
these
move
towards
an
attractant,
such
as
a
food
supply
or
pheromone
trails,
while
avoiding
collisions
with
neighbors.
So
there
are
a
couple
of
very
simple
rules
here.
The
first
is
that
you
have
you
know
this
sort
of
toxic
behavior,
so
it's
being
attracted
to
something
repulsed
from
something
some
some
environmental
stimulus.
A
Then
you
can't
be
too
close
to
your
neighbor.
You
can't
be
too
far
away
from
your
neighbor,
so
you
have
that
if
you're
familiar
with
the
boyd's
model,
this
is
exactly
the
type
of
thing
that
you
see
there.
You
see
these
very
simple
rules,
and
then
you
get
this
sort
of
complex,
collective
phenotype
that
emerges
from
it.
A
A
variety
of
mechanisms
and
fitness
strategies
have
been
associated
with
rotating
vortices
and
reflect
the
huge
diversity
of
vortex
scales
and
samples.
So
this
is
an
example
here
of
what
we're
talking
about.
So
this
is
the
key
ingredient
of
vortex
formation.
So
you
see
these
plasmodium
sporozytes
they're
bent
and
they
move
in
a
certain
direction.
A
A
This
is
a
vortex
of
fish
and
then
the
vortex
sporozites
is
sort
of
in
between
there,
and
so
that's
that's
an
example
from
this
one
parasite,
individual
sporozoites
glide
on
a
substrate
in
a
circle
owing
to
their
curved
shape.
They
display
a
strong
preference
for
the
anti-clockwise
direction,
because
the
chiral
shape
of
the
sporozoites
microtubule
cage
dictates
the
orientation
of
the
secretory
organelles
and
proteins
necessary
for
adhesion
and
motility.
B
A
They
do
image
analysis,
they
quantitate
that
image
analysis
and
they
build
the
simulations
from
the
data
and
they
basically
make
the
same
observations
that
the
mechanical
flexibility
allows
single
sporozites
to
be
sorted
according
to
their
curvatures
and
speeds,
and
that
these
effects
increase
with
vortex
size.
So
the
vortex
size
is
limited
by
the
phenotype
and
the
way
that
the
phenotype
behaves
in
the
collective.
A
We
also
find
that
the
vortices
undergo
oscillatory
breathing,
which
is
an
interesting
point,
which
means
that
it's
pulsating
in
some
way,
because
the
thrust
from
the
motility
force
of
the
single
sporozoites
can
be
stored
in
their
elastic
energy.
So
this
is
interesting.
There's
motility
going
on
there's
each
sporozoite
has
an
elastic
potential
energy
potential
and
this
results
in
this
sort
of
breathing
of
the
of
the
collective.
A
If
you
think
about
like,
if
you've
ever
watched,
schools
of
fish
in
coral
reefs,
if
you've
seen
video
of
that
or
if
you've
even
seen
it
up
up
close,
you
see
that
there's
a
sort
of
pulsating
and
the
collectives
aren't
just
these
static
clouds
that
move
they
move,
they
they
kind
of
pulsate
and
they
have
these
internal
dynamics.
So
that's
what
they're
getting
in
here.
A
In
general,
our
work
demonstrates
how
single
particle
shape
and
mechanics
can
determine
the
dynamics
of
large
active
collectives.
So
this
is
an
example
here,
a
little
bit
more
about
some
of
these
collectives
and
how
they're
rotating.
So
this
is
an
example
of
how
you
might
get
an
infection
of
these
parasites.
A
And
so
you
have
this
medium
of
elastic
material,
which
is
the
skin
viscoelastic
material.
And
then
you
have
the
bloodstream,
which
is
also
physical,
elastic
but
less
dense
and
so
you're
moving
they're
moving
through
that
and
they
so
to
do
that.
They
have
to
have
this
sort
of
adaptability
at
the
collective
behavior
level,
and
so
here
are
some
examples
of
different
types
of
observations
that
they've
made.
You
can
see
the
single
phenotype
of
a
single
parasite
here,
asporzite
and
then
the
collective
sporezites
here,
where
you
have
a
bunch
of
them.
A
So
they
kind
of
go
through
some
of
the
stuff.
They
run
their
agent-based
simulations
on
this,
and
so
they
show
some
of
the
examples
of
the
physics
and
how
they
affect
this
collective
and
they
don't
have
any
good
pictures
of
it.
But
I
think
this
is
basically.
This
is
the
model
sporozoite
which
is
where
they
model
each.
A
B
A
A
You
can
measure
things
like
the
force
distributions
and
the
elasticity
using
this
type
of
model.
You
also
have
these
different
radial
distances
and
you
can
measure
the
behavior
of
the
collective
itself.
You
can
look
at
the
radius
of
curvature
and
how
it
affects
these
different
collectives,
and
you
can
also
look
at
the
speed
of
the
rotation.
You
can
look
at
the
collectives,
so
this
is
all
done
in
simulation.
This
isn't
really.
This
is
taken
for
measurements,
but
the
the
actual
biophysics
and
these
graphs
are
generated
from
the
simulation
yeah.
A
So
then,
this
is
more
simulation
data.
The
simulations
predict
self-sorting
and
rotating
vertices,
so
they're
able
to
to
replicate
that
finding,
and
then
you
see
these
vortex
size
fluctuations,
so
they're
not
static,
they
don't
just
kind
of
sit
in
a
deterministic,
you
know
swarm
or
whatever
you
want
to
call
it
collective,
but
there's
a
pulsating
along
the
edges
and
sometimes
they
break
apart.
I
imagine
because
that's
what
collectives
can
do
sometimes,
and
so
you
can
see
the
different
shapes
here
over
time
of
these
collectives.
A
Okay,
so
that's
pretty
much
all
I
want
to
talk
about
the
papers.
This
is
something
that
was
put
out
on
social
media.
This
is
a
picture
from
this
work
where
they
said.
We
then
used
agent-based
modeling
to
identify
the
required
physics
elements.
In
particular
his
analysis.
The
authors
demonstrated
that
the
propulsion
of
the
single
spores
sporozoites
is
converted
into
elastic
energy
of
the
vortices,
making
them
a
model
system
for
active
elastic
matter.
So
this
is
something
we've
talked
about
in
the
context
of
embryos,
but
you
see
another
example
of
elastic
active
elastic
matter.
A
Being
these
collectives
of
very,
very
simple
organisms-
and
here
you
have
this
bridge-
between
experiments
and
models-
was
quantitative
image
processing.
So
this
is
something
that
you
know
you
put
these
things
under
a
microscope.
You
look
at
the
collective
under
a
microscope
and
you
can
actually
track
the
cells.
You
can
track
the
spore
sporozoites
in
in
their
sort
of
active
state
as
they're
moving
around
and
you
can
track
them
and
then
use
those
data
to
build
the
simulations.
A
So
the
microscopy
was
really
the
linchpin
of
all
this.
To
bring
this
to
something
that
we
could
actually
simulate
and
measure
for
that
matter,
and
this
is
able
to
show
that
the
sporos
sporozoites
and
the
vortices
are
sorted
according
to
speed
and
curvature
that
the
vortices
undergo
during
their
shape
oscillations
or
that
it.
B
A
Thank
you
for
joining
me,
so
I'm
going
to
finish
up
here
with
I'm
going
to
go
over
this
repository
a
little
bit
of
the
different
papers,
and
I
already
krishna
will
be
able
to
see
this
later
so
wataru,
but
this
is
the
open,
sour,
open
source
and
open
science
reading
repository
that
I
posted
in
the
slack-
and
it
has
like
some
books
like
there's
some
digital
books
in
here,
which
are
you
know
really
for
your
reference.
A
I
don't
mean
for
you
to
read
them
cover
to
cover,
but
I
just
want
you
to
have
this
as
a
reference
and
then
these
these
folders
that
have
you
know
different
topics
that
you
might
be
interested
in
and
then
this
is
a
laboratory
contribution
philosophy
which
is
maybe
less
relevant
to
you,
but
is
maybe
interesting
if
you
want
to
look
through
it.
This
book
is
a
particularly
a
particular
interest
to
maybe
you
know
when
you're
thinking
about
long
term,
about
your
projects
and
what
you're
producing.
A
A
So
if
you
think
about
like
linux-
or
you
think
about
some
of
the
other,
like
you
know,
firefox
or
other
types
of
free
and
open
software,
they
have,
you
know,
have
to
have
successful
structures
to
run
their
contributions
and
so
in
open
norm.
For
example,
you
know
we've
had
no
problem
attracting
people
into
the
project,
like
people
always
come
by
the
slack
and
check
it
out,
but
it's
you
know
it's
a
little
harder
to
turn
that
into
product.
You
know
into
like
contributors
who
are
producing
something
of
value.
A
A
So
this
book
has
a
lot
of,
and
this
again
is
for
your
reference-
you
don't
need
to
like
read
a
cover
to
cover,
but
there
are
a
lot
of
things
in
here.
So
you
know:
there's
the
history
of
open
source
software,
sort
of
the
difference
between
free
and
proprietary
software,
free
versus
open
source.
So
if
you're
open
sourcing-
something
it's
not,
you
know
locking
it
away
like
you
know,
you
might
have
like
the
windows.
A
System
is
proprietary,
you
can't
freely
modify
it,
you
can't
freely,
distribute
it,
but
with
open
source
software,
both
of
those
things
are
true.
Now
open
source
isn't
necessarily
the
same
thing
as
free.
You
can
have
something
where
you
open
up
the
code
to
people,
but
it's
definitely
not
free
in
the
sense
that
you
distribute
an
executable.
I
mean
you
know,
there
are
different
ways
that
people
do
this.
A
You
know
they're
open
source
organizations
that
have
software
that
isn't
quote
unquote:
free,
there's,
free
software,
that's
open
source,
there's
free
software
that
isn't
open
source,
and
then
you
know
that.
So
that's
that's
something
to
keep
in
mind
when
you
think
about
licensing
your
software.
Now
we
have
in
this
group
we
tend
to
stick
to
things
like
the
mit
license,
which
is
an
open
source
license
that
you
know
enables
different
people
to
do
different
things
of
the
software,
and
it
sort
of
you
know
keeps
people
from
using
it
in
different
ways.
A
That
might
not
be
what
we
would
want
to
see
in
the
world.
So
you
know
there
are
a
lot
of
different
things
you
need
to
think
about.
When
you
work
on
projects,
you.
B
A
Know
maybe
think
of
a
good
name,
something
you
won't
necessarily
have
to
worry
about.
What
we
work
on
with
in
in
the
divo
learn
area
is
version
control
and
versioning.
So
how
do
you
make
a
version
of
something
a
new
version
of
something?
So
these
are
all
things
that
you
know
you
might
learn
about
and
apply
to
your
work
later
in
life
or
you
know
we
can
talk
about
this
with
respect
to
diva
learn,
and
you
know
there
are
all
sorts
of
different
things
like
choosing
a
license.
A
You
know
there
are
all
sorts
of
open
source
licenses.
There's
also
you
know
discussing
things
having
discussion
groups
surrounding
your
software.
How
do
you
make
that
productive?
A
You
need
technical
infrastructure,
you
need
social
and
political
infrastructure,
so
you
know
there's
in
a
lot
of
open
source
projects.
There
are
different
types
of
leadership.
You
have
benevolent
dictator
types
of
leadership,
you
have
totally
democratic
types
of
leadership,
consensus-based
democracies-
and
this
you
know,
shapes
how
these
open
source
projects
actually
come
to
fruition.
If
you
know
sometimes
the
benevolent
dictator
model
was
good
for
driving
forward
innovation,
but
not
so
good,
for
you
know
managing
changes.
A
You
know
there
are
other
ways
that
you
can
make
money
off
of
open
source,
so
there's
a
whole
section
on
economics
of
open
source
and
how
those
things
get
financed
if
you're
interested
in
open
source
as
a
career,
you
know
a
lot
of
companies
are
looking
for
open
source
people
or
people
who
know
open
source
and
a
lot
of
companies
are
investing
in
open
source,
not
necessarily
on
their
own,
but
to
use
those
tools
in
different
ways.
So
there
are
different
things.
A
You
know
to
keep
in
mind
long
term
about
what
you're
learning
in
this
program.
You
know,
then,
there's
you
know
communications
which
involve
like
how
do
you
communicate
to
people
what
you've
done?
We
definitely
are
going
to
be
writing
up
the
results
of
our
projects
for
submission
to
google.
So
this
is
maybe
something
to
pay
attention
to.
A
You
know
there's
managing
participants.
How
do
you
get
people
to
say?
How
do
you
get
a
maintainer?
For
example?
How
do
you
get
people
to
maintain
the
platform
so
that
it
stays
current?
A
So
these
are
things
you
need
to
think
about.
When
you
make
you
release
open
source
software
into
the
world,
how
does
it
get
maintained?
And
so
you
know
that's
and
then,
of
course
you
know
you
have
people
just
contribute.
How
do
you
get
contributors
to
come
in
and
contribute
to
your
software?
Instead
of
just
you,
we've
done
things
like
hacktoberfests,
which
are
work,
organized
things
that
github
is
involved
with,
and
these
these
sorts
of
activities
drive.
People
to
you
know,
contribute
and
they
reward
them
for
contributing
and
there.
A
So
there
are
a
lot
of
ways
to
get
people
involved
in
your
open
source
platforms
and
how
to
get
it
maintained
and
how
to
keep
it
relevant
and
all
these
things.
So
I
just
wanted
to
bring
that
to
your
attention
because
I
think
that's
important
to
think
about
long
term
in
terms
of
what
you're
learning
here
and
what
you
might
do
in
your
career.
A
So
any
questions
about
that.
A
Okay,
well,
if
that's
all
the
questions,
you
know
we
can
ask
questions
later
in
slack
or
via
email,
but
I
think
we'll
wrap
it
up
for
today,
and
I
hope
that
you
have
a
good
week.
I
know
for
the
gsox
students
I'll
be
continuing
to
give
you
information
in
the
slack.
I'll,
probably
give
you
some
more
things
for
community
period
and
I'm
going
to
encourage
you
to
go
into
the
open
worm
organization,
open
room
foundation
and
learn
more
about
what
they're
doing
that
might
be
useful.
A
Otherwise,
let's
just
keep
working
on
getting
our
projects
ramped
up
and
hopefully
by
when
the
coding
period
starts,
we'll
be
able
to
hit
the
ground
running
on
things
and
get
our
you
know
stick
to
our
schedules.