►
From YouTube: DevoWorm (2021, Meeting 34): Digital Bacillaria Roadmap, Model Organism Zoo, Connectome/Projectome
Description
Roadmapping Digital Bacillaria for further development and Hacktoberfest. Bacillaria as a digital model of motility, colonial dynamics, Diatom biomechanics, and aneural cognition. Digital models and model organisms. Planning for Neuromatch 4 Conference. Papers on a C. elegans connectome inspired robot and the projectome of the bumblebee. Attendees: Bradly Alicea, Jesse Parent, Susan Crawford-Young, and Assaf Wodeslavsky.
A
A
C
A
A
Otherwise
we
have
a
good.
We
have
a
good
number
of
things
to
talk
about
today.
So
let's
get
started,
I'm
gonna
share
my
screen.
A
All
right
so
first
thing
is,
is
that
we
have
a
couple
of
things.
We
we've
got
a
bunch
of
materials
on
the
digital
bassleria
project,
we're
going
to
revisit,
but
first
we're
going
to
talk
about
this
repository
that
I
found
in
on
github,
and
it's
something
that
you
might.
I
think
people
in
this
group
would
enjoy
hey
susan.
How
are
you
you're,
muted,
oh
okay,.
A
So
people
in
this
group
might
enjoy
it's.
This
awesome
deep
gnn.
So
gnn
stands
for
graph
neural
networks.
So
this
is
a
repository
of
papers
on
the
topic,
so
we
I
think
we
talked
about
gnns.
A
Previously
we've
talked
about
how
they
work
a
little
bit.
We've
talked
about,
maybe
how
to
implement
them
in
our
group
for
different
problems.
This
is
a
nice
set
of
reviews
or
papers
on
contemporary
topics.
It
looks
like
they're
keeping
it
up,
2021,
2020
and
before
2020..
That's
what
counts
in
graph.
Neural
networks
is
like
kind
of
a
historical
review
but
looks
like
we
have
a
it's
a
fairly
new
area,
so
we
have
a
lot
of
different
papers
on
it.
I
don't
think
they
have
any
software.
A
C
My
eyes
are
working
better
now
I
have
some
really
weird
glasses.
You
can
see
that
this
one
has
the
lens
yeah
so
currently
trying
to
find
some
better
glasses
so
that
I
can
see
better,
but
yeah
I've
been
able
to
use
reading
glasses.
So
I'm
now
I
can
do
my
signal
processing.
C
Yeah
I
have
to
do
my
research
project.
No
I've.
C
Review
my
signal
processing
so
that
I
can
do
the
math
for
it
like
I.
I
know
the
physical
process
and
I'll
know
the
equations
when
I
see
them,
but
I
I
know
the
physical
process.
A
B
C
A
Yeah,
it's
it's
tough,
like
graduate
school
is
pretty
tough
like
just
trying
to
keep
up
with
some
of
the
things
you
know,
trying
to
master
some
area
and
new
areas
that
you
haven't
really
been
familiar
with
in
the
past.
So
yeah.
A
Go
back
to
this,
I
don't
think
there's
any
any
sort
of
like
soft.
A
There
are
a
lot
of
papers
that
you
know
kind
of
give
you
a
good
idea
of
this
area,
and
so
you
know,
maybe
we
should
talk
more
about
graph
neural
networks.
I
think
christian
is
interested.
I
I
don't
know
if
he's
going
to
join
us
today,
but
that's
something
we
should
follow
up
on
okay,
so
then
I
guess
the
other
thing
I
wanted
to
talk
about
was
this
digital
basil
area
repository,
so
I
met
with
oswal
and
asmet
and
thirun.
On
friday
we
talked
a
little
bit
about
this.
A
A
We
have
a
paper
that,
or
we
actually
had
a
paper
that
was
like
machine
learning
and
bio
and
for
biomechanics,
and
it
involved
looking
at
basil
area
as
a
digital
organism,
and
so
this
is
just
recently
been
published
as
a
book
chapter,
and
so
this
is
the
book
chapter
here.
It's
chapter
10
in
this
book
towards
a
digital,
diatom
image,
processing
and
deep
learning.
Analysis
of
bacilluary
paradoxes,
so
in
this
paper
to
recap,
we
took
back
basilar
paradoxa,
which
is
actually
a
species
that
is
not
very
easy
to
define
from
the
other
baso
area
species.
A
A
So
in
this
paper
we
took
we
mined
a
lot
of
data
from
actually
from
videos
that
were
available
on
the
web
and
we
took
these
videos
and
we
took
them
apart
and
analyzed
them,
so
that
was
ozuwa
and
asmet.
I
think
we're
mainly
involved
in
that
and
it
kind
of
characterized
all
the
cells.
So
you
know
basil
area
are
these
cell
colonies
that
have
these
long
rod
shaped
cells
and
they
move
together
in
a
certain
way
that
we've,
I
think,
we've
talked
about
in
this
group.
A
So
these
rod
shaped
cells
are
stuck
together
in
colonies
and
the
colonies
configure
themselves
in
certain
ways,
and
so,
when
you
get
a
still
image,
you
get
all
these
different
cells
that
are
in
different
positions
relative
to
one
another.
You
have
to
define
the
cell,
the
cell,
centroid,
etc,
and
then
so.
This
is.
This
is
an
example
of
a
bacillary
colony.
You
have
these
chains
of
cells
and
they're,
either
elongated
or
they're
stretched
out
a
little
bit
like
this,
and
you
get.
You
know,
they're.
A
Of
movement
so
they're
in
this
picture,
they've
this
was
of
mueller,
who,
back
in
the
18th
century,
you
know,
was
able
to
identify
at
least
eight
modes
of
movement
or
eight
modes
of
sort
of
this
phenotype,
and
so
here's
an
example
here
of
a
colony
that,
where
there's
one
cell
that's
stretched
out
against
another
cell
and
then
the
rest
of
the
cells
in
the
colony
are
aligned
with
one
or
the
other
cell
here.
So
this
is
an
example
of
stretching,
but
only
local
stretching
not
global
stretching.
A
And
then
this
is
a
close-up
of
the
rod
shaped
cells.
They
have
a
morphology.
You
know
there.
There
are
a
lot
of
actin
filaments
inside
and
the
cell
walls
are
silicate,
so
they're,
actually
an
interesting
type
of
cell
that
you
don't
see
in
say
like
animal
cells.
So
this
is
a
rod
shaped
cell
here.
This
is
10
microns
for
a
sense
of
scale,
so
it's
a
very
small
organism
and
these
colonies
are
also
likewise
very
small,
and
this
is
done
by
scanning
electron
microscope.
A
So
this
is
actually
pretty
high
resolution
for
this
spatial
scale
and
you
can
see
that
we
have
these
horizontal
slits
the
rougher
here
and
then
you
have
different
these
valves
that
take
in
water
and
they
do
other
types
of
liquid
and
gas
exchange,
and
then
you
have
different
structures
within
the
cell
and
so
they're.
Actually
you
know
this
is
part
of
their
movement.
They
have
these
actin
filaments
that
move
and
cells
move
against
one
another.
So
it's
almost.
A
In
some
ways,
and
so
we're
able
to
identify
these
cells
digitally
extract
them
from
video
and
then
thomas
herbic,
who
we
work
with,
he
was
able
to
take
colonies
and
do
this
analysis
on
different
colonies.
This
was
not
a
high
throughput
analysis,
like
usual
and
azimut's
analysis,
but
this
is
something
where
he
took
the
cells
under
a
light
microscope.
This
time
look
mark,
put
landmarks
on
the
cells
relative
to
their
extent
of
their
movement
against
one
another.
So
you
can
see
these
cells
are
are
extending
themselves
in
the
colony,
but
the
cells.
A
You
know
they
don't
reach,
they
don't
come
apart.
You
would
think
like
if
they
stretched
against
one
another,
that
they'd
fly
apart
and
have
to
reassociate,
but
they
don't.
They
come
to
sort
of
this
region
where
they're
sort
of
in
close
proximity.
They
don't
quite
come
apart,
so
you
know
they
can
stretch
in
each
direction
like
this.
There's
this
little
overlap
area
here
that
overlap
area
is,
as
it
turns
out.
A
You
know
there's
a
size
where
the
cell
isn't
overlapping,
and
so
you
can
define
all
those
parts
of
the
cell
that
you
know
define
the
movement,
and
so
you
have
you
measure
the
length
you
can
find
centroids.
You
can
do
other
things
and
then
you
can
build
a
model
of
movement
which
is
in
a
graph
further
down.
This
is
actually
an
example
of
what
azmid
and
noswell
did
for
each
of
these
video
frames
for
a
colony.
This
is
a
basilaria
cell.
A
A
A
So
there
were
some
things
that
were
tried
and
failed.
This
is
like
a
good
example
of
some
methods
that
failed,
so
they
tried
a
couple
of
methods.
They
tried
candy
edge,
detection
and
watershed
segmentation
on
these
colonies,
and
they
didn't
work
at
all,
but
the
method
that
they
finally
settled
on,
which
was
a
deep
learning
model.
Deep
lab
version.
Three
actually
worked
quite
well
for
this,
so
this
is.
A
Of
how
these
you
know,
models
and
methods,
kind
of
work
or
don't
work.
A
And
then
I
think
we
have
some
more
information.
This
is
actually
plotting
out
some
of
the
features
and
they're
in
space
and
then
there's
this
there's
this
graph
here
of
the
stuff
that
thomas
did
with
the
biomechanics,
where
it's
actually
looking
at
the
movement
over
time.
So
you
have
your
cells,
they
form
this
oscillatory
movement
back
and
forth,
so
he's
looking
at
that.
A
They
tend
to
settle
down
over
time
in
this
example,
and
they
form
this
attractor
base
in
here
where
you
have.
If
you
look
at
like
two
cells
that
are
moving
against
one
another,
you
can
plot
it
out
as
an
attractor,
and
you
can
see
that
like
in
this
case,
this
is
a
nice
way
to
show
and
I'll
zoom
in
this
is
a
nice
way
to
show
some
of
the
noise
that
happens
when,
as
the
cells
stretch
out
and
come
back
together.
A
So
as
they
kind
of
stretch
to
their
maximal
point,
there's
a
sort
of
noise
that
happens
where
they
kind
of
stop
and
come
back.
So
it's
gliding
it's
fairly
smooth
gliding
in
these
parts
of
the
attractor
and
then
it's
a
little
bit
more
there's
a
little
bit
more
variation
on
these
ends,
and
these
ends
are
where
the
the
cells
stop
and
then
come
back
together.
A
So
this
is
velocity,
and
so
this
is,
you
know,
like
I
think,
the
first
order
of
well
actually
the
second
order
of
position
or
actually
the
first
order
position,
and
then
you
can
have
so.
This
is
meters
per
second,
and
so
that's
the
first
derivative
of
position
and
you
can
get
like
you
know,
acceleration,
which
is
the
second
or
derivative
of
position
which
is
going
to
show
more
of
this
sort
of
noise.
A
And
then
you
can
get
even
higher
derivatives
of
position
which
really
kind
of
break
out
some
of
the
stuff.
To
you
know
some
of
the
higher
order,
movement
dynamics
so,
for
example,
lighting,
isn't
necessarily
smooth.
It
looks
smooth
here
relatively
smooth,
but
you
can
see
there's
some
variation
in
the
trace
here.
I
would
like
to
like
pull
pull
apart
and
take
a
look
at
it,
and
so
you
know
you
can
do
you
can
do
this,
but
you
need
higher
resolution
data
than
what
we
have
here.
A
The
sampling
rate
here
was,
I
don't
remember
what
the
sampling
rate
was,
but
it's
really
the
sampling
rate
is
you
know
your
video
rate.
I
think
it's
like
twice
the
video
rate,
so
whatever
video
rate
was
which
wasn't
very
high,
and
you
know
beyond
that,
you
get
effects
like
aliasing
and
other
effects.
A
We
actually
talked
about
this
and
there
we
did
talk
about
a
paper
once
where
you
can
actually
up
sample
video
data
where,
if
you
have
like
a
movie,
that's
sampled
at
a
thousand
frames
per
second,
you
can
sample
it
at
10,
000
frames
per
second
just
so
that
you
can
get
that
extra
resolution.
So
you
can
do
things
like
this,
where
you
can
look
at.
You
know
higher
derivatives
of
of
position
and
emotion.
A
A
Second,
derivative
of
of
position
and
so
forth,
and
so
we
can
do
a
lot
of
these
types
of
things
and
then
we
have
so
you
know
we
can
also
look
at
the
changes
between
the
position
of
cell
2
relative
to
cell
1..
So
if
you
have
a
pair
of
cells
that
are
moving
against
one
another,
you
can
see
the
relative
position
as
it
changes.
A
So
in
this
case,
I
think
we
had
a
different
discrepancy
here
later
in
the
time
series
as
the
cell
was
moving
or
as
the
cells
were
slowing
down
in
terms
of
their
movement
against
one
another.
So
this,
like
you
know
this
accordion
pattern
that
you
have
of
movement
of
the
colony
it
slows
down
and
even
stops
and
as
it
stops,
there's
some
instabilities
in
the
two
cells
moving
against
one
another.
So
we
can
quantify
all
this
using
these
methods
we've
introduced
in
this
paper,
so
this
is
in
this
book.
A
A
The
diva
worm
repository
or
the
digital
basil
area
repository
this
is
in
the
diva
worm,
github
organization,
and
so
this
was
the
repository
that
we
made
for
this
project.
When
we
were
writing
this
up
so
there
you
know,
some
of
these
repositories
are
fairly
old
they're
over
a
year
old,
and
some
of
these
things
were,
you
know,
we've
had
this
out
for
probably
as
a
preprint
for
probably
about
a
almost
two
years
now.
A
So
you
know,
some
of
the
analyses
are
maybe
kind
of
dated,
but
you
know
we
did,
I
think
for
our
audience.
That
was
a
pretty
good.
I
was
a
pretty
advanced
analysis,
but
we're
talking
now
about
going
back
and
actually
taking
data,
that's
a
little
bit
more
controlled
because
we
took
a
lot
of
the
data
off
of
the
web,
so
they're
different
source
materials,
which
made
it
a
little
bit
hard
to
do
a
lot
of
the
segmentation
when
you
have
different
backgrounds,
for
example,
in
different
movies
and
different
experimental
setups.
A
It's
really
hard
to
like
build
a
data
set
and
pull
all
those
images
together
into
a
single
data
set.
A
So
that's
something
that
we
had
to
overcome
in
the
first
study
and
the
second
study,
the
the
stuff
that
we're
going
to
do
now,
we're
looking
at
doing
some.
You
know
getting
some
data
from
a
standardized
source
and
thomas
harvick
has
been
collecting
data
on
diatoms
he's
been
culturing
them
and
taking
videos
of
them,
and
so
those
videos,
then
we're
going
to
use
those
we're
going
to
you
know
analyze
them
in
in
similar
ways
to
what
we
did
here
now.
A
The
reason
I
also
bring
this
repository
up
is
because
hacktoberfest
is
coming
up,
so
hacktoberfest
is
coming
up
this
in
october.
That's
why
they
call
it
oktoberfest,
it's
supposed
to
be
like
oktoberfest
but
they've,
given
it
the
name
hacktoberfest
and
we
we're
going
to
participate
as
an
organization,
and
so
it
means,
from
october
1st
to
october
31st
we're
going
to
have
activities
for
hacktoberfest
and
really
the
way
this
program
works.
This
is
their
github.
A
This
is
a
program
that
they've
run
with
different
organizations.
I
think
digitalocean
was
the
first
organization
host,
but
other
organizations
host,
and
the
idea
is
that
you
go
to
an
open
source
repository
with
a
badge
and
we'll
put
a
badge
on
the
repositories
we
want
to
mark
and
then
we're.
You
know
if
you
like,
do
a
certain
number
of
pull
requests
in
that
repository,
you
get
some
sort
of
you
know
recognition
or
a
t-shirt
or
something.
A
So
you
know
we
we
haven't
given
any
we've
done
a
lot
of
recognition
of
this,
but
we
haven't
given
any
swag
out,
so
I
don't
know
if
we
have
any
swag
to
give.
So
I
don't
know
we'll
we'll
deal
with
that,
but
the
I
think
the
idea
here
is
to
get
people
involved
in
it,
and
you
know
maybe
right,
you
know,
give
recognition
for
it
and
things
like
that.
A
It's
supposed
to
motivate
people
to
participate
in
open
source,
so
you
know,
hopefully
we're
also
doing
the
divo
learn
repository
as
well.
So
if
people
want
to
people
are
contributing
to
evil
or
nevo
learn,
I
think
I
talked
about
that
last
week,
where
we
were
talking
about
some
of
the
people.
You
know
who
have
been
contributing
to
divalern
and
were
you
know
in
the
process
of
producing
a
new
version
on
that?
A
So
maybe
if
people
want
to
participate
in
october
for
divalern,
then
you
know
you'll
be
part
of
this
new
version
as
well,
but
this,
but
this
is
something
that
we're
going
to
promote
in
the
next
next
couple
weeks,
I'm
going
to
put
a
badge
on
the
different
repositories
and
I
think
our
repositories
that
are
going
to
be
digital
basil
area
and
then
the
diva
learn
project
which
I
don't
have
up
here,
but
anyways
this
is
going
to
be.
A
This
is
a
little
bit
different
type
of
repository
because
it's
really
centered
on
some
of
the
work
that
was
done
in
that
paper
and
we're
trying
to
do
some
more
work
here.
Like
I
said,
we
have
video,
you
know,
there's,
there's
image,
processing,
there's-
and
I
I
was
talking
to
about
this.
I
have
some
data
for
people.
If
they're
interested
in
doing
this,
we
may
have
to
set
it
up
where,
if
people
want
to
analyze
data,
we'll
have
to
create
a
pipeline
for
them
to
access
it.
A
So
this
is
going
to
be
maybe
or
maybe
that
the
people
who
participate
in
hectoberfest
are,
you
know,
sort
of
offline,
and
we
might
you
know,
give
them
certain
tasks
surrounding,
like
maybe
some
of
the
you
know,
digital
processing
or
maybe
just
maintaining
things.
I
don't
really
know
how
we
want
to
work
this,
but
it's
a
little
bit
different
because,
most
you
know
repositories
are,
you
know,
fix
this
issue
of
code
or
whatever,
and
so
they're
very
clear
issues
in.
A
C
A
You
can
label
issues
here,
so
I
might
label
it
with
hacktoberfest
and
you
know
give
people
some
guidance
on
where
to
go
so
I'll
be
working
on
that
from
now
until
the
end
of
the
month,
and
hopefully,
on
october
1st,
we
can,
you
know,
start
off
with
some
people
contributing
hello
jesse.
How
are
you
I
see
you're
here
october,
susan.
C
C
Yeah,
you
already
showed
it
to
me.
Oh.
A
A
Okay,
good
yeah,
so
good,
and
so
the
idea
of
a
you
know
a
digital
diatom.
It's
kind
of
a
lot
like
open
worm
in
a
lot
of
ways
like
in
open
worm,
they're
trying
to
build
this
digital
c
elegans.
A
So
the
idea
is
to
take
the
data
that
exists
and
in
a
lot
of
what
we've
done
in
open
worm
has
been,
like.
You
know,
really
based
on
microscopy
data
or
other
types
of
data
like
you
know,
lineage
trees
and
things
like
that
and
just
putting
them
into
numbers,
creating
models
of
say,
like
neurons
and
simulating
that
in
this
case
we're
creating.
A
You
know,
models
of
cells
so
we're
doing
some
cell
modeling
we're
doing
some
movement
modeling
and
we're
doing
some
actually,
as
I'll
show
you
in
a
minute
we're
also
doing
some
modeling
of
sort
of
the.
I
wouldn't
want
to
say
the
mind
but
like
something
called
non-neuronal
cognition
or
a
neuronal
cognition,
a
neural
cognition,
and
so
that's
something
that
you
know.
We
are
kind
of
putting
pieces
together
on
that,
so
we're
very
early
on
and
and
macillaria
or
where
diatoms
in
general
are
much
simpler
than
even
c
elegans.
A
So
we're
you
know,
I
think,
we're
moving
along
on
that
we're
going
to
get
a
take
another
pass
at
this
movement
and
cell
modeling,
and
then
we're
also
going
to
deal
with
some
deeper
issues,
and
then
you
know,
hopefully
at
some
point
we'll
put
this
all
together
into
a
you
know
something
that
looks
like
open
worm
and
you
might
ask
yourself
what
is
what
are
diatoms
have
to
do
with
development?
Well,
I
mean
diatoms
have
a
very
interesting
sort
of
life,
history
and
and
growth
set
of
growth
patterns
they
accrete
their
cells.
A
Instead
of
you
know
the
cells
growing
like
a
a
typical
animal
cell,
so
they
accrete
layers
of
their
cell
and
they.
You
know
they
have
a
lot
of
interesting
growth
dynamics
and
then,
of
course,
they
can
associate
themselves
in
colonies,
which
is
also
interesting.
It's
very
different
from
what
animal
some
animals
do.
Other
animals
do,
of
course,
form
colonies,
but
they
do
in
a
slightly
different
manner,
and
so
this
is,
you
know.
This
is
kind
of
an
interesting.
A
Digital
modeling
exercise,
it
always
does
rely
on
more
data,
and
it's
one
of
the
things
I
mean
about
open
worm
is
that
we
have
we
sort
of
have
this
idealized
organism.
You
have
this
adult
or
it
started
out
with
an
adult,
and
then
you
know
you
have
the
hermaphrodite
in
the
male.
So
we
don't
really
differentiate
between
those
in
open
worm
too
much.
A
We
don't
spend
a
lot
of
time
on
the
larval
stages.
So
yeah
I
mean
it's
going
to
be.
You
know
this
digital
modeling
exercise
of
modeling
an
organism.
You
know
it's
like
what
do
you?
What
is
your
conception
of
the
organism?
You're
trying
to
model?
Is
it
like
just
an
adult
or
a
static
part
of
their
life
history?
Is
it
like
one
stage
of
their
life
span,
because
some
organisms
have
very
different
phenotypes,
say
in
different
parts
of
their
life?
A
C
The
diet
homes
are
an
important
part
of
the
ecosystem.
They're
like
the
base
food
for
a
lot
of
things,
yeah.
C
Small
fish
like
they're,
something
like
20
percent
of
the
ecosystem,
relies
on
them.
So
yeah,
that's
another
aspect
of
it.
A
Yeah
and
they
also
use
them
for
biofuels-
a
lot
there's
a
lot
of
research
being
done
so
they're,
they're
algae,
so
they're
they're.
You
know
they
live
in
water,
they
have
this
and
so
that
they're
used
the
people
think
they
can
use
them
for
biofuels
and
other
applications
like
that.
Some
material
science.
I
think,
as
well
so
they're
very,
very
interesting
to
a
lot
of
people,
there's
a
community
of
people
who
are
interested
but
like
with
open
worm
it
might
take
on
a
life
of
its
own.
A
A
A
Welcome
back
yeah,
so
I
wanted
to
go
over
this
paper
then,
and
I
I
told
you
about
the
bacillaria
anaeronal
or
non-neuronal
cognition,
and
this
this
title
is:
you
know
it's
kind
of
a
working
title,
the
actual
title
of
this
paper
and
it's
almost
finished.
This
is
why
I
bring
it
up.
Is
the
psychophysical
world
of
the
motile
diatom
basil
area
paradoxa?
A
A
A
You
know
this
ranges
from
like
people
looking
at
cells
and
human
body
and
building
a
model
of
decision
making
to
you
know
people
doing
things
here
in
this
part
of
the
introduction,
where
they're
looking
at
different
organisms,
so
they're,
looking
at
all
different
types
of
small
organisms
like
slime,
molds
and
other
types
of
you
know,
organisms
that
don't
have
a
brain,
and
so
you
know
they're,
maybe
doing
things
like
collective
behavior
or
they're
doing
things
like
responding
to
stimuli,
and
so
you
know,
even
though
they
don't
have
an
nervous
system,
they're
exhibiting
something
called
cognition,
and
maybe
it's
worth
thinking
about
cognition
a
lot
more
broadly
than
say
like
the
human
brain
and
think
about
what
cognition
kind
of
means,
which
is
information,
processing.
A
And
so
that's
I
mean
that's,
not
the
sort
of
the
latin
root
of
cognition.
But
it's
what
people
have
come
to
mean
by
when
they
say
cognition
is
that's
information
processing.
So
in
this
case
you
could
use
a
brain
or
you
could
use
some
sort
of
cellular
structure
like
this,
where
the
cells
have.
A
You
know,
different
components
so,
for
it
turns
out
that
each
of
these
bacilluria
cells
can
sense
their
environment,
they
can
sense
water
flow,
they
can
sense
light,
they
can
sense
temperature,
they
can
sense
water
chemistry
and
they
can
do
all
this
and
they
respond
to
it.
A
So
there's
a
respond,
a
characteristic
response
that
they
make,
and
so
in
this
paper
we
review
some
of
the
the
evidence
for
this,
especially
like
light
sensitivity-
and
you
know
they
can
do
all
this
and
they
respond
in
kind
now
they
can
do
this
in
a
colony
or
they
can
do
this
individually.
In
this
case,
we
have
a
colony
of
cells
that
are
responding
to
these
cues
and
they're,
doing
it
in
a
spatially
dependent
manner.
A
So
you
know
if
you're
on
one
end
of
the
colony
versus
the
other,
and
so
you
know
that's
part
of
their
decision
making,
but
also
that
they're
connected
to
one
another
through
this
colonial
structure,
and
so
these
colonies
are
held
together
by
secreted
polymers.
So
there's
this
extracellular
component
that
allow
you
know
that
sort
of
enables
a
lot
of
the
dynamics
of
this
colony.
A
A
So
this
is,
you
know.
This
is
an
interesting
thing
and
most
people
have
explained
this
using
biophysical
models.
So
they've,
you
know
used
metaphors
of
explosion
like
jet
engines
and
things
like
that
very
mechanical
metaphors.
Very
you
know
you
know
things
that
are
chemical
or
biophysical
in
nature.
A
What
we're
doing
here
is
we're
turning
that's
head
and
we're
saying.
Well,
this
is
actually
information
processing
like
something
you
would
might
see
in
the
brain,
but
you
also
see
in
other
organisms
that
don't
have
a
brain
and
there's
this
whole
area
of
a
neuronal,
cognition
or
non-normal
cognition
that
allows
us
to
interpret
this.
A
So
we
get
into
some
of
the
the
reviewing
some
of
the.
What
non-normal
cognition
is
it's
basically
breaking
the
behavior
down
into
some
of
this.
Some
of
the
components
of
the
movements
how
those
things
are
put
together
and
then
you
know,
we've
introduced
models
for
all
of
this.
A
So
you
know
we
use
this
sort
of
paradigm
of
psychophysics,
which
has
been
applied
to
a
wide
range
of
animals,
a
wide
range
of
organisms,
and
it's
basically
this
idea
that
you
have
sensory
inputs
and
you
have
characteristic
outputs,
and
these
characteristic
outputs
can
be
characterized
by
some
quantitative
law
or
some
quantitative
rule.
So
we
use
the
example
in
this
paper
of
of
the
weber
fechner
law,
which
is
a
law
that
determines
you
know
if
you
have
a
certain
level
of.
A
If
you
you
have
a
sensitivity
to
changes
in
a
stimulus,
so
it
could
be
changes
in
a
white
stimulus
or
changes
in
a
temperature
stimulus,
and
the
idea
is
that
you
can
discriminate
between
two
different
levels
of
intensity,
and
so,
if
you're
better
at
this,
you
can
discriminate
between
smaller
and
smaller
magnitude
differences.
A
So
you
know
it's.
I
think
I've
shown
this
in
the
group
meetings,
but
you
know
you
might
have
an
array
of
like
10,
dots
versus
11,
dots
or
10
dots
versus
100
dots
and
the
idea
is,
you
have
to
be
able
to
distinguish
whether
those
are
the
same
or
different,
and
you
might
think.
Well,
that's
easy.
You
know
you
just
take
the
the
10
versus
100.
Dots
is
really
easy.
The
10
versus
11
dots
is
really
hard
and
you
would
be
right.
That's
true,
and
that's
usually
what's
found.
A
You
know,
I
don't
know.
Maybe
in
humans
use
grouping
strategies
to
sort
of
make
this
a
little
bit
easier
for
smaller
magnitudes,
but
in
general
they
define
this
sort
of
relationship,
and
so
that's
kind
of
what
we're
drawing
from
here,
but
we're.
A
We're
looking
at
signal
detection
theory
we're
looking
at
something
called
a
collective
pattern,
generator
which
is
based
on
a
a
central
pattern,
generator,
which
is
often
found
in
in
nervous
systems
where
you
have
a
network
of
neurons
and
they
produce
a
rhythmic
output.
So,
like
your
heart,
is
a
collective
pattern,
generator
you
find
them
on
insects
and
even
c
elegans,
where
you
have
periodic
movements
that
are
generated
by
groups
of
neurons.
A
A
Each
cell
is
taking
an
accounting
of
that
stimulus
and
differences
between
intensity
of
the
stimulus
and
then
producing
a
pattern
with
this
connectionist
model
that
generates
this
collective
pattern
generator
and
then
you
know
also
even
algorithmic
information
theory
for
other
types
of
you
know
looking
at
different
modes
of
movement
and
changes
in
the
mode
of
movement,
and
then
it
kind
of
gets
a
little
bit
into
dynamical
systems
to
tie
that
together-
and
I
have
to
finish
that
part
well,
there
are
a
couple
things
that
need
to
be
finished,
we're
pretty
close
to
the
end
here,
and
so
this
is
a
very
sort
of
broad
overview
paper,
and
you
know
it's
not
like
based
on
any
well
it's
based
on
some
data,
but
we
don't
do
any
rigorous
data
analysis
in
it,
so
the
other.
A
So
the
thing
I
want
to
do
from
this
is:
I
want
to
get
some
feedback
on
it
from
place.
We're
going
to
be
submitting
it
because
it's
dreadfully
late
in
terms
of
where
I
was
going
to
submit
it,
and
then
you
know
we'll
keep
working
on
this
as
a
theme
in
this
digital
basil
area
project,
and
so
hopefully
that
we'll
get
some
feedback.
A
Is
this
paper
still
open
to
contributions?
Yes,
it
is.
Let
me
send
you
a
link-
and
I
can
put
this
in
the
slack
as
well,
and
the
permission
should
be
open
on
this.
If
you
want
to
go
ahead
and
take
a
look,
this
is
the
document.
It's
a
google
docs,
so
you
can
just
kind
of.
If
you
have
comments,
please
leave
a
comment.
A
Don't
just
put
in
something
in
the
text.
It's
the
way
look
kind
of
like
managed
comments,
but
yeah
I
mean
I
would
appreciate
comments
or
you
know
if
you
want
to
add
some
things
in
because
we're
getting
close,
but
I
can't
seem
to
get
to
the
word
limit
that
I
need
on
that,
because
I've
been
we've
been
working
on
it
for
months
and
I've
been
trying
to
massage
it
for
about
at
least
two
months,
and
you
know
getting
getting
a
lot
of
that.
All
in
order
is
very
difficult,
but
anyways.
A
I
want
to
eventually
incorporate
some
of
this
into
the
digital
basil
area
repository
as
well.
So
some
of
these
models
that
we
talked
about
in
the
paper
you
know
these
might
be
something
that
could
be
implementable
as
digital
models.
So,
for
example,
you
know
some
of
the
stuff
on
predictive
processing
and
heavy
and
learning
you
know
those
are
things
that
you
know.
A
We
have
to
kind
of
come
up
with
a
way
to
model
that,
like
you,
know
some
software,
some
code
or
you
know
some
sort
of
simple
simulation
where
you
can
simulate
the
sort
of
thing
without
having
a
nervous
system.
You
know
if
we're
using
this
vasel
area
model.
You
know
what
kinds
of
things
could
we
model
with
this
or
how
would
we
build
a
computational
model
that
would
demonstrate
this
without
a
nervous
system.
A
A
I
may
need
to
rewrite
this
part
a
little
bit,
but
it's
this
idea
of
active
inference
and
there's
this
idea
of
frequent
energy
minimization
for
cognition,
and
this
actually
ties
in
well
with
the
free
energy
minimization
models
that
we
use
in
like
in
in
biophysical
movement.
So
if
we
just
use
the
biophysical
example,
we
would
have
like
a
biophysical
model
where
there's
energy
minimization.
A
That
tells
you
what
the
behavior
should
look
like
and
then
you
would
have
this
more
sort
of
cognitive
or
information
processing
model
which
would
do
the
same
thing.
So
it's
I
think
it's
interesting
how
there
are
parallels
there
that
you
can
build
upon
and
then
the
free
energy
principle,
if
you
use
that
or
use
predictive
processing,
which
is
related
to
that.
Those
sorts
of
things
then
give
you
an
sort
of
a
pathway
into
other
sort
of
more
cognitive
models
where
you
can
maybe
say
that
there's
information
processing
going
on.
A
So
you
can
use
these
metaphors
and
some
of
the
computational
mechanisms
to
draw
parallels
and
bridge
between
things
that
don't
look
at
first
glance
to
be
related,
but
but
anyways
ultimately
we'd
like
to
put
this
into
the
digital
basil
area
repository
as
well.
So
we
have
a
behavior
section,
but
there
isn't
much
here.
It's
just
some
images
and
files
that
some
code
that
was
put
in
here,
but
there
isn't
much
structure
to
this.
A
Yet
so
we
really
need
we're
really
trying
to
work
on
this
part
to
fill
it
in
and
then
eventually
we're
gonna
have
this
as
like
a
resource
for
people
yeah.
So
this
this
this
description
is
dated,
but
I
think
we're
you
know:
we've
got
a
lot
of
stuff.
I
need
to
update
this.
This
was,
I
think,
created
at
the
beginning
of
the
project,
so
yeah
we're
going
to
be
lujul
wants
to
revive
this,
as
mitt
wants
to
revive
this.
I
also
talked
to.
Who
else
did
I
talk?
I.
A
Wants
to
do
some
work
with
this
as
well,
but
I'm
not
sure
you
know
we're
gonna
kind
of
go
over
this
and
in
upcoming
meetings
and
kind
of
hash
out
some
of
these
things
and
then
we'll
be
doing
things
for
october
october
festivus
as
well,
so
so
that
was
a
lot
of
talking
about
that,
but
I
think
I
hope
people
get
a.
A
We
can
like
pick
up
activity
on
this
again,
maybe
next
next
google
summer
code,
we
can
actually
get
this
funded
so
like
last
year
we
had
a
project
on
this.
We
had
a
lot
of
applicants,
but
we
didn't
get
it
funded,
so
we
had
like
seven
or
eight
applicants
for
that
didn't
get
anywhere
with
that,
so,
but
they
did
actually
come
up
with
some
really
interesting
proposals.
A
So
we'll
follow
up
on
this.
I
wanted
to
go
to
the
submissions
page,
the
document
that
we
have
and
again
we
have
all
these
things
that
are
kind
of
hanging
in
the
air.
We
have
this
boring
billion
book
that
you
know
this.
This
early
life
embryo
is
an
early
life
idea
that
we've
kind
of
yet.
C
A
Realize
anything
with
we
have
this
kindle
book
idea,
which
is
also
kind
of
hanging
out
there.
We
have
a
mathematics,
a
diva
worm,
that's
something
that
I
talked
about,
I
think.
Last
week
I
talked
about
sort
of
the
mathematics
of
open
worm
and
showed
that
document
where
we
had
the
you
know
the
equations
and
we
had
some.
A
You
know
methods,
that's
kind
of
what
we're
looking
for
for
this,
but
for
diva
worm,
and
so,
if
you
want
to
find
out
more
about
that,
go
back
to
the
video
from
last
week,
I
can
well
I
have.
I
did
put
it
in
the
slack
channel,
so
you
can
check
that
out.
A
There
are
also
some
opportunities
to
do
some
molecular
level
simulations.
So
I
was
talking
about
diatom
motion
and
the
derivatives
of
position.
A
Derivatives
of
motion-
that's
something
that
this
is
so
jerkiness
is
where
you
get
really
high
derivatives
of
of
motion,
and
it
shows
that
sort
of
variation
at
a
very
high
scale
or
high
level
of
you
know.
So
you
can
see
a
lot
of
movement
variation.
A
If
you
look
at
something
called
jerkiness,
which
is
like
you
know,
positioned
or
like
I
guess,
distance
per
second
per
second
per
second-
and
that's-
you
know
that'll
it's
interesting
because
you
know
a
lot
of
this
higher
level
sort
of
variation,
especially
when
these
cells
change
their
mode
of
behavior.
Their
position
is
really
interesting,
so
we
have
a
couple
of
lines
on
that.
I
don't
know
where
that's
going
now
and
then.
Finally,
we
have
this
narrow
match,
4.0,
which
the
submissions
are
coming
up.
A
A
We
also
did
some
stuff
on
artificial
neural
networks
and
biological
neural
networks
this
year.
Actually
I
was
thinking.
Maybe
we
could
do
something
more
oriented
towards
diva,
worm
and
kind
of
a
maybe
something
where
we
have
an
event
during
neuromatch,
which
is
in
descent.
The
actual
conference
is
in
december,
and
I
don't
know
if
people
are
interested
in
this,
but
this
is
something
that
we
could
do
where
we
have
like
a
meet
up
of
people,
maybe
even
across
to
open
worm.
I
may
bring
it
up
at
the
annual
meeting.
A
A
So
you
know
I
definitely
like
to
submit
some
abstracts
and
if
you
have
anything
you'd
like
to
submit,
let's
talk
about
how
to
do
that.
How
to
make
that
you
know
interesting
to
this
audience
and
you
know,
maybe
we
can
make
some
diva
worm
tie-ins
here
so
yeah.
I
think
they're
interested
in
a
lot
like
there's,
there's
a
attract
on
cellular
molecular
work,
there's
also
a
track
on
some
other
types
of
computational
work.
So
there
are
a
number
of
tracks.
We
could
fit
into
computational
tracks
for
machine
learning
and
deep
learning.
A
So
there
are
a
number
of
opportunities
here
and
then,
of
course,
the
nurips
workshops
and
neurops
is
coming
up
in
december.
So
if
you're
interested
in
machine
learning,
there
are
a
lot
of
opportunities
to
learn
from
that.
I
think
a
lot
of
the
workshops.
Their
submissions
are
closed,
but
we
should
check
this
out
as
well.
B
I
have
a
few
ideas
about
mirror
match.
Okay,
I
think
it
would
be
good
to
do
what
you
said
about
the
open
worm
like
get
everybody
to
present
what
they're
doing,
and
I
think
as
far
as
specific
presentation,
either
by
I
think,
it'll
be
good
in
general
to
do
like
a
bigger
thing,
but
also
I
I
would
particularly
be
interested
in
doing
something
along
the
lines
of.
B
B
You
know
things
things
like
like.
I
think
I
think
there
could
be
something
because
I'm
thinking
a
little
bit
back
to
I
didn't
quite
this.
This
was
something
that
never
happened
at
princeton
envisioned
logan
logan,
fraser
collins,
who
was
very
into
like
at
the
time
he
sort
of
shifted
off
from
that
now.
B
But
in
the
time
he
was
very,
very,
like
you
know,
in
favor
of
getting
to
like
being
able
to
fully
model
a
a
like,
the
the
nervous
system
and
the
biology
of
a
honey
of
a
bee,
as
opposed
to
a
fly
and
like
how
how
being
able
to
make
advances
in
modeling
and
simulating
those
organisms
could
help
us
and
also.
B
Why
why
why
look
at?
Why
look
at
non-normal
cognition?
You
know
why
look
at
these
things
that
we
can
model
in
different
ways
and
some
kind
of
a
thing
along
those.
B
Have
it
fully
fleshed
out,
but
like
sort
of
a
a,
maybe
not
necessarily
like
I
could
see,
this
is
very
much
a
typical,
like
jessie
thing,
to
walk
or
say
a
lot
of
my
projects
fall
in
the
category,
but,
like
you
know
what
sort
of
the
history
of
this?
What's?
Where
are
we
at
right
now
and
we
will
be
good
in
the
future
like
what
what's
the
cutting
edge
of
this,
and
what
can
I
provide
this
kind
of
thing?
B
This
is
very,
like
all
my
presentations
have
that
theme
anyway,
but
like
just
in
the
comics
of
the
simulation
of
all
organisms.
That
comes
to
my
mind,
and
I
don't
know
if
we,
I
imagine,
there's
been
things
that
have
been
already
been
done
here
or
warm
or
deep
or
warm,
but
like
that's
just
what
like
I,
I
that's
what
comes
to
my
mind
as
yeah.
This
one
comes
in
my
mind
as
it's
a
very
dressy
thing
to
say
and
suggest,
but
but
you
know
I'm
curious
what
what
fits
with
that.
A
Yeah,
so
I
think
actually,
the
model
organisms
part
is
interesting,
because
I
know
we
have
the
devo
zoo
stuff
and
that's
of
course
something
we
can
always
go
back
to
and
you
leverage,
but
I
was
at
a
workshop
this
summer,
where
the
you
know
it's
like
biologists
and
modelers,
and
one
of
the
things
that
people
were
some
of
the
modelers
were
really
intrigued.
By
is
you
know
all
these
model
organisms
that
the
biologists
talk
about
like
you
know
they
have
different
model
organisms
for
different
systems,
and
I
think,
last
week
I
had
a
paper.
A
I
think
this
was
in
the
other
group
where
we
on
where
they
looked
at
organoids
and
other
model
organisms,
and
they
looked
at
how
they
like
sort
of
recapitulate
biology,
which
means,
I
guess
be
it.
You
know,
be
close
to
something
like
human,
behavior
or
biology,
and
so
the
biology
you
know
they.
They
made
this
assessment
of
it.
A
A
So
those
are
all
things,
and
I
think
that
idea
model
organisms,
I
think,
is
something
that
they're
very
few
resources
on
like
kind
of
reviewing
model
organisms
and
their
usefulness,
but
especially
like
when
we
get
into
the
digital
organism
space
where
like.
If
I
want
to
build
a
digital
model
of
this
organism,
you
know,
first
of
all,
you
know:
how
is
it
useful,
how
does
it
fit
into
like
other
model
organisms
if
we.
A
We
understand
a
little
bit
more
about,
like
you,
know
the
diversity
of
biology,
or
you
know
different
processes
like
you
know,
if
we
had
different
models
of
you
know
like
c
elegans
and
bees-
and
you
know
fruit
flies,
could
we
understand
some
process
in
common?
A
We
actually
did
a
paper
on
that
several
years
ago,
where
we
looked
at
like
developmental
data
or
or
tracking
data
of
embryos,
and
looking
at
like
some
of
the
common
potentially
common
processes
across
different
types
of
embryogenesis,
so
I
mean
those
sorts
of
things
you
know
you
can
make
comparative
things
going.
You
know
you
have
comparative
work
going
on
there,
and
so
it's
very
interesting.
I
think
we
should
try
for
that.
A
B
Yeah,
I
would
be
curious,
like
if
you
want
to.
B
Maybe
you
could
like
refine
what
I've
said,
what
I
brought
and
what
you
said
and
like
mention
it
in
the
slack
and
like
tag
me
in
it
please,
because,
because
I'm
fleshing
out
like
this
isn't
on
the
fly
thing
and
I
was
like
oh
that
would
be
cool,
and
I
would
I'm
definitely
interested
in
that
because
I
feel
like
that's.
B
I
haven't
done
anything
really
with
diva
worm
for
a
while,
and
that
would
balance
out
some
of
the
things
with
me
on,
like
any
biologically
based
a
lot
of
the
things
I'm
doing
so
that'd
be
pretty
cool,
even
the
the
version
that
you
said.
The
other
thing
I
would
like
to
mention.
If
we
have
a
few
moments
I
put
in
and
it
could
actually.
B
Related
thing,
too,
that
we
could
do,
but
this
kind
of
is
it's
only
also
in
orthogonal,
but
but
just
to
kind
of.
A
B
Show
this
the
brainstorm
time
reading
club,
I'm
not
saying
we
can
initially
do
this
in
terms
of
under
a
natural
related
thing,
but
just
to
kind
of
show
it
if
it's
a
evolutionary
history
never
would
have
brain
and
discussion
between
disability,
neuroscience
and
so
on.
And
all
these
you
know
these
people
here
and
it's
starting
in
october.
So
it
wouldn't
really
be
like
a
lot
of
time
and
you're
gonna
do
weekly
meetings,
so
it
would
be
sort
of,
like
you
know,
it'll
be
later
on
and
a
shorter
time
thing
to
do.
B
But
I
could
see
I
may
be
interested
in
presenting
something
like
discussing
this
at
mirror
match
in
some
form
or
like
making
it
a
direct
like
making
an
adjacent
presentation
or
go
along
with
this
and
kind
of
thinking
about
the
the
paul
kaiser
stuff
that
we
talked
about
in
johannes
group.
Mostly,
I
think
not.
I
don't
think
I
mentioned
that
in
diva
world,
but
just
just
sort
of
that
sort
of
evolutionary
stages
of
development
for
how
brains
and
nervous
systems
have
come
across.
B
So
again,
I
don't
have
a
specific
project
set
in
mind,
but
the
two
things
that
immediately
come
to
mind
about
neuromatch
and
divorce.
What
we
just
talked
about
and
something
like
this
and
whether
or
not
we
I'm
probably
gonna,
do
this.
This
group,
like
100,
either
way
in
combination
with
some
things
going
on
with
orthogonal
and
just
cognition
features
in
general,
but
also
like,
I
think,
there's
a
lot
of
far
for
making
an
interesting
presentation
at
an
event
like
neuromax.
A
Yeah,
I
think
that's
also
yeah,
that's
something
that
would
be
also
another
narrow
match
candidate,
something
on
this
and
you
know
something
you
might
learn
from
this.
Like
you
know,
we
could
think
of
a
good
topic
that
we
could
explore
in
a
15-minute
talk,
there's
so
much
there
and
you
know
it's.
You
don't
have
to
wait
for
the
reading
club.
I
guess
you
could
just
go
and
look
at
some
of
the
resources
that
they
have
and
you
know.
Maybe
this
is
an
interesting
idea.
C
I
partly
know
why
axolotl
salamanders
are
a
model
organism
and
I
did
a
presentation
in
japan
on
it.
So,
okay,
yeah.
A
A
You
know
neuroscience
students,
they're
kind
of
aware
of
the
model
organisms
but,
like
you
know
how
much
are
people
aware
of
why
they
were
selected
or
you
know
how
they're
related
to
one
another?
That's
the
kind
of
thing.
Maybe
that
would
be
useful
because
I
don't
think
a
lot
of
people
appreciate
that
and
I
think
there's
this
missing
link
there
between,
like
I'm
using
this
model
organism
and
oh.
This
is
why
it's
important.
B
All
right,
the
only
other
thing
that
comes
to
mind
about
your
match-
and
I
was
gonna
mention
this
to
the
other
group
foundation
here,
because
it's
made
more
biology
based.
Is
there
a
lot
of
interesting
discussion
about
the
grossberg
book
that
came
up
with
the
ohana
and
other
stuff
like
that?
So
just
wanted
to
mention
just
wanted
to
mention
whoa
what
is
going
on
here.
B
Just
wanted
to
mention
that
I'm
not
saying
we
do
like
a
book
review,
but
more
like
just
just
mentioning
that,
for
the
sake
of
it.
A
Yeah
yeah
thanks
for
that.
That
was
a
good
overview
of
maybe
some
things
we
can
do
here,
because
it's
kind.
A
B
Yeah,
I'm
I'm
I'm
like
finally
reclaiming
my
mornings
and
I
know
our
saturday
morning.
The
meetings
are,
are
you
know
I'm
there,
but
like
having
functional
mornings
during
the
week
are
like
I
have
to
do
that
now
because
of
all
the
stuff
on
my
plate,
and
I'm
just
like.
Oh
I'm
here
and
I
can
check
in
with
you
there's
a
lot
of
cool
stuff.
B
So
yeah
nerdmatch
is,
on
my
mind
a
little
bit
and
I
think
between
this
group
and
orthogonal
and
everything
else,
I'm
I
think,
it'll
be
a
lot
of
fun,
but
I
think
I
think
it'd
be
a
really
good
target
to
center
some
of
our
upcoming
projects,
and
especially
for
me,
like
I
have
I
just
haven't.
I've
been
really
out
of
the
loop
with
the
one
for
pretty
much.
B
You
know
until
my
life
for
those
you
know,
I
haven't
seen
that
much
in
this
group,
but,
like
my
life,
changed
a
lot
in
the
last
six
months
or
so
so
like
being
here
in
india,
mcgowan's
like
ooh
like
okay,
you
know
yeah,
I'm
definitely
curious
about
the
the
model
organism.
B
Slash
simulation
stuff
that
I
mentioned,
and
in
general
like
even
for
like
neuromatch,
is
kind
of
a
nice
like
target
criticism.
Yeah.
We
can
make
this
and
do
it
and
make
it
very
it's
a
very
clear
objective
to
do,
but
in
general,
I'm
interested
in
stuff
like
that.
So
I'll.
Just
note
that
and
state
my
interest
and
so
on.
Yeah.
A
Very
good
so,
finally,
to
wrap
up
the
meeting
today,
I'm
going
to
go
over
a
couple
papers
and
let
me
go
back
to
the
okay,
so
we're
back
here.
A
Okay,
so
I
guess
a
lot
of
things
in
this
folder
and
actually
I
was
going
to
talk
about
two
things:
I'll
be
very
connectome
centric
today.
So
today
I'm
going
to
talk
about
two
papers-
and
this
is
linking
the
first
one-
is
linking
the
connectome
to
action,
emergent
dynamics
in
a
robotic
model
of
c
elegans.
A
So
this
is
something
that
open
worm
does
with
our
robotics
group
and
and
we're
doing
stuff
with
simulating
the
connectome,
and
this
is
linking
the
connectome
to
action
using
a
robot,
and
so
this
is
a
group,
valencia,
urbina,
canas
and
gleeser,
and
I
don't
they're
from
argentina
okay,
so
they
have
a
you
know
that
I
mean
I
don't
know
what
they're,
how
prevalent
they
are
in
the
literature.
I
don't
think
so,
but
you
know
this
is
something
that
they
did
here.
A
A
So
this
is
not
like
the
kind
of
thing
that
open
arm
is
doing
with
you
know,
modeling
ion
channels
and
and
picking
specific
cells,
but
they're
actually
looking
at
a
new
doing
a
numerical
simulation
based
on
c
elegans
nervous
system.
So
you
know
this
is
a
very
popular
connectome
to
work
with
for
people.
You
know
because
it's
it's
got
a
finite
number
of
cells.
We
know
what
the
cells
do.
A
They
have.
You
know
very
well-defined
circuits
that
actually
give
you
good
behavioral
outputs.
So
this
is.
This
is
a
common
thing
and
people
aren't,
you
know,
simulating
the
connectome
of
other
organisms
as
much
to
build
behaviors,
mainly
because
it's
harder
to
do
with
c
elegans.
You
have
a
lot
of
factors
that
work
in
your
favor.
A
B
A
The
robot
interacts
with
the
environment
through
a
sensor
that
transmits
the
information
to
sensory
neurons,
while
motor
neurons
motor
neuron
outputs
are
connected
to
wheels.
This
is
actually
a
lot
like
the
breitenberg
vehicle.
If
you're
familiar
with
that-
and
in
my
other
group
we
talk
about
this
a
lot.
It's
basically
the
idea
that
you
have
like
a
vehicle
with
wheels
the
effectors
or
wheels
the
sensors
are
the
sensor.
A
You
know
they're,
usually
up
into
the
front
they're
like
kind
of
like
eyes,
or
maybe
something
that
feels
the
environment,
and
so
you
can
make
this
connection
within
the
robot
and-
and
actually
you
know,
fratenberg
vehicles
are
very
small,
connectomes
they're,
not
even
connectomes,
in
fact
a
lot
of
the
time,
but
they
actually
can
produce
interesting
behaviors
that
are
very
similar
to
emotional
states.
A
A
Avoiding
collisions
with
obstacles
working
with
a
robot,
robotic
model
makes
it
possible
to
keep
track
simultaneously
of
the
detailed
microscopic
dynamics
of
all
the
neurons
and
also
register
the
actions
of
the
robot
in
the
environment.
In
real
time.
This
allowed
us
to
study
the
interplay
between
connectome
and
complex
behaviors.
A
A
So
this
c
elegans
stands
out
as
the
first
animal,
whose
complete
connectome
has
been
mapped.
That's
not
entirely
true,
there's
some
other
small
connectomes.
A
I
guess
there
are
some
small
connectomes,
like
parts
of
other
brains
that
have
been
mapped
out
and
there
a
couple
other
small
connectomes
have
been
mapped
out
that
are
even
smaller
than
c
elegans.
Although
the
range
of
behaviors
isn't
as
rich
as
c
elegans,
but
we
can
talk
about
that
later,
so
they
talk
about
this,
and
so
what's
also
interesting
about
c
elegans
is
that
it
has
302
nodes
or
neurons,
but
you
get
this
highly
specialized
structure.
A
So
you
get
these
network
motifs,
which
are
common
patterns
in
the
in
the
of
connection
that
you
know
lead
to
different
types
of
outputs.
So
you
know
you
might
have
like
a
wheel
and
spoke
motif
or
a
triangle
motif
where
the
nodes
form
through
their
connections,
form
some
pattern
and
that
usually
has
some
functional
significance.
A
C
A
These
again
they're,
not,
I
don't
think
they're
sharpening
the
the
neural
activity
in
the
term
in
terms
of
different
ion
channels,
but
they're,
actually
using
a
model
of
like
threshold
activity
and
so
they're
able
to
look
at
the
outputs
of
these
cells,
different
cells
and
they're
able
to
look
at
the
sort
of
do
some
signal.
Processing
as
susan
is
becoming
familiar
with
and
they're.
Looking
at
the
output
signal
and
the
frequency
of
the
output,
and
then
they
do
some
principal
component
analysis
to
look
at
this,
so
they
analyze
their
network.
A
Then
they
look
at
some
of
these
nested
neuronal
dynamics,
where
they're
looking
at
collective
oscillations
and
they're.
Looking
at
how
those
collective
oscillations
are
coupled
in
the
network,
so
we
talked
about
with
the
basilaria.
You
have
these
networks
of
of
nodes
that
produce
oscillations
that
are
stable
and
that's
what
they're
looking
at
here,
but
in
this
case
neurons
will
produce
oscillations.
They
can
produce
them
autonomously,
but
then,
when
you
connect
them
together,
they
result
in
more
complex
patterns,
more
complex
states,
and
so
that's
what
they're
doing
here.
A
A
So
these
are
things
that
you
know
if
you're
familiar
with
fireflies
and
the
firefly
model
of
synchronization,
where
fireflies
flash
their
luciferase
sack,
you
know
when
they're
they've
used
it
as
a
signaling
mechanism
to
signal
other
fireflies
and
over
time
they
can
synchronize
their
flashing
so
that
all
the
fireflies
are
flashing
at
the
same
time,
and
so
this
is
a
similar
thing
that
you
see
in
neurons,
whereas
they're
connected
together,
they
synchronize
or
they
order
their
oscillations,
and
so
this
actually
can
result
in
some
very
ordered
behaviors.
A
So
you
can
observe
these
nested
normal
dynamics
from
the
head
of
the
c
elegans
to
the
entire
nervous
system,
including
the
ventral
nerve,
cord
and
tail
ganglia,
and
then
they
show
some
more
examples
of
nested
neuronal
dynamics.
Here,
I'm
looking
for
some
of
their
robotic
work
here.
I
guess
they're
now
extending
this
model
to
global
dynamics,
as
it
relates
to
action.
A
They're
doing
some
dynamical
systems
modeling
here,
where
they're
looking
at
this,
these
ac
this
activity
and
up
on
a
couple
of
principal
components,
so
they're
able
to
like
take
this
activity
and
decompose
it
into
different
movement
patterns.
So
you
have
backwards
movement
forward,
movement,
rotate
left
and
right
movement
and
then
a
stopping
movement
and
they're
able
to
pull
this
all
out
in
this
principle
component.
A
So
this
is
largely
you
know,
this
is
okay,
so
this
is
something
that
it's
based
off
of
tim
buspas's
work,
so
tim
busbass
used
to
be
a
part
of
open
worm,
but
he
split
parted
ways
with
open
worm
and
for
a
number
of
reasons
and
they're,
so
they've
been
using
his
program
and
they're
running
it
on
a
raspberry
pi
computer.
So
this
is
interesting.
They're
running
these
very
simple:
to
build
robots,
then
they're
able
to
do
this
work,
so
this
is
very
interesting
work.
A
And
this
is
a
very
long
paper
in
e-life
journal
and
this
is
on
the
projectile
of
the
bumblebee
central
complex.
So
this
is
another
one
of
these
smaller
connectomes,
but
this
is
actually,
you
know,
part
of
a
larger
connectome.
So
this
is
in
the
bumble
bee,
which
I
think
jesse
talked
about
with
respect
to
logan
one
of
his
friends
from
another
group
and
is
interested
in
modeling,
the
bumblebee
and
so
in
this
paper,
they're,
actually
talking
about
getting
towards
sort
of
characterizing.
A
The
connectome
of
the
bumblebee
in
this
case
they're
actually
interested
in
what
they
call
the
projectile.
So
the
abstract
reads:
here:
insects
have
evolved,
diverse
and
remarkable
strategies
for
navigating
in
various
ecologies
all
over
the
world,
regardless
of
species.
Insects
share
the
presence
of
a
group
of
morphologically
conserved,
neural
pills,
known
collectively,
as
a
central
complex.
A
So
this
is
a
part
of
the
brain.
That
is
it's
this.
I
guess.
Neural
pills
are
a
certain
type
of
it's
it's
a
thing
in
in
neural
development.
So
I'm
trying
to
think
of
an
easy
way
to
characterize
it.
But
I
can't
right
now:
the
cx
is
a
navigational
center
involved
in
sensory
integration
and
coordinated
motor
activity,
despite
the
fact
that
our
understanding
of
navigational
behavior
comes
predominantly
from
ants
and
bees.
So
again,
our
model
organisms
are
limited
in
insects,
so
navigation
and
insects.
A
We
know
mostly
from
ants
and
bees
and
we're
you
know
we.
There
are
a
lot
of
different
insects
and
probably
a
lot
of
different
modes
of
navigation
behavior.
But
this
is
where
we
study
it
most
most
of
what
we
know
about
the
interlock
neural
circuitry
of
such
behavior
comes
from
working,
fruit
flies,
so
in
other
words,
we
know
a
lot
about
and
b
navigation,
but
the
neural
aspect
of
this
comes
from
fruit
flies.
A
So
we
don't
even
know
that
much
about
what
goes
on
in
the
minds
of
ants
and
bees,
but
we
try
to
make
that
extension
from
fruit
flies
and,
of
course,
that's
not
a
very
good
model
because
there's
a
lot
of
uniqueness
amongst
these
different
groups
of
organisms.
So,
that's
you
know,
that's
one
of
the
drawbacks
of
a
of
a
model
organism
here
we
aim
to
close
this
gap
by
providing
the
first
comprehensive
map
of
all
major
columnar
neurons
and
their
projection
patterns.
A
A
A
First
columnar
cells
are
central
to
the
most
computations
carried
out
by
the
cx.
They
link
the
four
adjacent
cx
neural
pills
and
provide
the
basis
for
intrinsic
computations,
but
also
generate
output
to
other
brain
areas,
and
so
this
kind
of
gets
into
some
of
this.
What
this
connectome
looks
like
it's
a
it's
quite
a
bit
different
from
c
elegans
connectome
and
that
has
different
structures.
A
It
has
different
types
of
neurons.
A
So
you
know
some
of
what
drosophila
has
and
what
we
study
is
actually
very
specific
to
their
ecology
and
what
they're
doing
behaviorally.
Now
we
don't
know
like
for
c
elegans.
The
same
is
also
true.
If
you
go
to
other
nematodes,
you
see
other
vastly
different
types
of
nervous
systems
and
connectomes.
A
So
what
we're
learning
from
ciolin's
connectome
is
nice
and
it
maps
the
behavior
cleanly,
but
it
doesn't
tell
us
very
much
about
anything
outside
of
the
specific
ecology
of
that
organism.
So
to
address
this
issue,
a
counterpart
of
the
drosophila
cx
connectome
will
be
needed,
so
they
provide
a
first
step
in
understanding.
This,
though-
and
it
looks
like
the
projectome-
is
something
that
is-
is
a
term-
that's
very
specific,
it's
a
type
of
connectome.
A
They
introduce
this
german
kesturi
and
leichmann,
and
I
guess
it's
just
a
matter
of
looking
at
how
things
project
from
one
part
of
the
brain
to
another
and
using
that
as
the
basis
for
a
connectum,
so
your
connectome
can
take
on
different.
You
know
you
can
construct
your
connectome
in
different
ways.
A
You
know
one
way
is
to
just
take
neurons
and
look
at
the
connections
between
them,
whether
they
be
you
know,
through
synapses
or
through
something
like
gap
junctions,
which
is
the
way
that
the
c
elegans
connectome
has
traditionally
been
constructed,
or
you
can
take
this
sort
of
projectomics
approach
where
you
look
at
how
one
area
projects
to
another
area
and
in
the
human
brain.
A
A
I'm
gonna
skip
that
for
now.
This
is
this.
Is
a
nice
set
of
images
here,
where
you
can
see
that
you
know
you
have
these
areas,
these
narrow
pill
and
they're
projecting
to
other
areas,
and
you
can
see
that
they're
looking
at
these
projections
here
and
how
they're
laid
out
and
so
they're
using
different
dyes
to
mark
these
connections.
A
So
you
know
this
is
a.
This
is
the
kind
of
method
they
use.
When
you
have
a
really
complex
connectome
is
they
have
to
use
tracing
dyes,
to
look
at
some
of
these
connections
and
see
where
they
go,
because
it's
not
immediately
apparent
from
like
just
doing
a
high
resolution
microscopy
image,
so
this
is
how
they
do
connectomics
and
more
complex
nervous
systems.
A
C
All
right,
what
was
the
raspberry
pi
reference
there?
I
oh.
A
So
yeah
that
was
for
this
paper
on
the
c
elegans
robots,
so
they
were
building
robots
using
a
c
elegans
connectome
and
they
were
actually
using
raspberry,
pi
technology
to
simulate
the
connectome.
C
C
Yeah
anyways,
so
I'm
here
and
I
guess
thank
you
I'll
try
again
next
week.
C
A
A
A
C
Let's
see
tom
portegas
gave
me
information
on
a
free
program.
You
can
get
to
the
point
so
I'll
I'll
maybe
send
that
information
to
you
yeah.