►
From YouTube: Digital Diatoms: biophysics and information processing
Description
One hour lecture on "Digital Diatoms: biophysics and information processing". Presenter: Bradly Alicea, April 2023.
A
Hello,
my
name
is
Bradley
ellisam
I'm
from
the
open
worm,
Foundation
I'm,
orthogonal
research
and
education,
laboratory
and
I'm,
the
head
of
the
diva
worm
group
and
I'm,
going
to
talk
to
you
about
our
work
on
digital
diatoms,
biophysics
and
information
processing.
So
at
the
right
here
you
can
see
some
examples
of
the
digital
diatom
which
we
will
be
talking
about
today.
A
A
As
we
do
this
work
in
conjunction
with
some
of
the
developmental
work,
we
do
with
c
elegans
and
other
model
organisms
as
well.
So
diatoms
are
algae,
they're,
single
cell
eukaryotes,
with
plant-like
characteristics,
so
they
have
a
chlorophyll
and
a
silica
cell
wall.
So
the
cell
wall
is
actually
important
because
it
keeps
it
rigid
and
in
fact
it
can
keep
its
rigidity
in
a
number
of
shapes.
As
you
can
see
at
the
left
make
sure
at
the
left,
you
have
circular
phenotypes,
you
have
Rod
shaped
phenotypes.
A
A
We
can
look
down
at
the
bottom
and
see
that
there
are
these
Rod
shaped.
Flake
shaped
and
3D
shaped
phenotypes
and
they're
all
about
10
to
80
microns
in
size,
which
is
not
very
large,
so
they're,
very
small
organisms.
Sometimes
they
live
solitary
and
sometimes
they
live
in
colonies
and
we'll
be
talking
about
a
colonial
diatom
called
basilaria.
That's
the
genus
name
we'll
be
talking
about
that
today.
A
A
So
along
the
edges,
it
has
a
structure
and
then
it
has
rafae
which
are
in
the
middle
and
then
an
epival
which
are
on
the
sides,
and
so
you
can
see
that
there's
a
phenotype,
that's
actually
quite
complex
and
we'll
see
electron
microscopy
images
of
this
in
a
little
bit.
This
is
important
for
digitization
because
it
gives
us
landmarks
to
use
four
different
techniques,
so
these
are
two
examples
of
the
Vaso
area
diatom.
A
So
this
is
a
bunch
of
these
rod-like
cells,
they're
in
physical
contact
in
a
colony
and
they
they
tend
to
slide
against
one
another
quiescent.
Of
course
they
sit
like
this,
as
you
can
see
in
this
sort
of
Stack,
sometimes
that's
elongated,
sometimes
it's
sort
of
in
a
short
stack
that
is
just
one
cell
next
to
another,
but
sometimes
they
slide
against
one
another
and
they
make
this
accordion-like
movement.
So
these
are
some
examples
of
diatoms
in
the
wild.
A
As
you
can
see
there
here
for
sliding
back
against
one
another,
and
then
they
reach
a
maximum
limit,
so
they
don't
quite
come
apart
and
then
they
come
back
and
they
slide
back
around
this.
This
becomes
important
later
in
some
of
our
measurement
techniques,
because
we
do
get
this
oscillation
first
of
all
that
where
they're
fully
extended
and
then
they
slide
back
together
and
then
they
come
back
and
extend
out
the
other
direction,
but
also
because
they're
these
physical
limits
to
the
colonies
extend
so
as
they
slide
apart.
A
They
don't
slide
apart
all
together,
and
this
is
going
out
of
focus
a
little
bit
kind
of
have
a
place
where
they
stop
and
there
there's
limited
contact
at
that
point,
and
then
they
come
back
together
and
there's
some
there's
some
Physics
of
movement
during
that
the
end
of
that
slide
that
are
interesting
to
characterize.
A
So
this
is
a
clearer
view
of
one
County,
that's
isolated
from
the
rest
of
them,
so
you
can
see
here
it's
a
much
cleaner
view
of
the
movement,
so
it
slides
apart
and
the
cells
are
almost
like
coming
out
of
contact.
There's
a
small
area
where
the
remaining
contact
at
the
very
end
of
that
extension
and
then
they
slide
back
in
the
other
direction
and
you
can
see
they
don't
quite
come
out
of
contact.
A
A
So
we
can
analyze
bacillary
movement
and
culture.
This
is
an
accordion-like
Movement.
We
can
model
this
as
a
series
of
oscillators.
So,
as
you
can
see
here
as
we
as
we
look
at
this
oscillation,
we
can
actually
look
at
this
as
oscillators,
both
cell
by
cell
oscillators,
or
sometimes
we
can
measure
movement
relative
to
different
cells
in
the
culture.
So
there's
a
leading
cell,
as
you
can
see,
there's
a
Leading
Edge
of
the
movement
and
then
there's
a
trailing
edge
of
the
movement.
A
So
the
one
cell,
in
the
middle
of
the
frame,
isn't
moving
at
all
the
rest
of
the
cells
are
moving
back
and
forth
and
you
can
see
the
Cell
at
the
top
or
the
cell.
Sort
of
to
the
farthest
right
is
the
leading
edge
of
this.
So
it's
really
driving
the
oscillation,
the
rest
of
the
cells,
except
for
the
bottom
cell
and
the
stack
are
following,
and
it's
producing
the
sliding
motion
so
you'll
see
that
the
different
phases
of
movement
there
are
different
ways.
We
can
model
this.
A
A
Sometimes
the
movement
can
vary
due
to
stimuli
and
we're
not
varying
the
stimuli
here,
but
sometimes
these
colonies
will
become
quiescent
or
they'll
change,
their
oscillation,
the
rate
of
oscillation,
based
on
the
amount
of
light
that
they
encounter
based
on
their
chemical
environment
and
so
on,
and
so
you
can
see
here.
This
is
a
little
bit
more
natural
set
of
oscillations
in
that
you
have
a
little
bit
more
variety,
so
you
can
see
that
there's
variation
in
the
county.
It's
not
just
that
smooth
oscillation
of
a
couple
cells.
A
A
So
here's
some
electron
microscope
images
of
Basil
area
companies,
so
at
the
top
you
can
see
one
of
these
stacks
and
they're
a
raid
kind
of
a
semi-open
array
here.
So
there's
a
Leading
Edge,
that's
kind
of
moving
out
and
there's
a
trailing
Edge,
that's
sitting
stationary
and
you
can
see
how
it's
kind
of
spreading
out
like
an
accordion,
but
you
can
actually
zoom
in
on
single
cells
and
look
at
some
other
features
so
like
we
talked
about
in
the
first
slide,
we
have
this
rafae,
which
we
can
use
as
a
landmark.
A
We
have
other
things
along
the
surface
of
the
cell.
We
can
use
as
a
landmark
and
depending
on
the
resolution
or
the
image
input
images
we're
using.
We
can
actually
characterize
some
of
these
things
and
use
them
as
landmarks,
for
you
know,
marking
certain
parts
of
the
cell
relative
to
other
cells,
and
so
this
is
a
very,
very
high
resolution
image.
This
is
one
micron.
This
line.
This
is
two
microns,
so
you
can
see
the
scale
spatial
scale
this
the
entire
cell
is
about
10
microns.
A
So
that's
a
very,
very
tiny
cell
and
you
can
see
how
small
these
features
are.
These
feature
sizes
are,
in
many
cases
less
than
a
micron.
So
these
these
you
know
these
parts
of
the
buffet
here
are,
you
know,
maybe
two
microns
at
most
and
some
of
these
perforations
are,
you
know
a
fraction
of
a
micro,
it's
a
very
small
feature
sizes.
A
So
one
set
of
experiments
we
did
was
to
apply
deep
learning
and
do
a
deep
learning
analysis
to
these
basilarious
colonies,
and
a
lot
of
the
technical
work
here
was
done
by
usual,
saying
nazmat
Singh,
who
have
been
contributors
to
the
diva
one
group,
especially
back
in
2018
2019
2020,
then
that
Arrow,
and
so
we
can
use
deep
learning
techniques
to
Define
individual
cells
in
a
bacillary
colony
for
movies
of
isolated
colonies,
and
you
saw
the
movies
previously.
We
could
isolate
the
colonies
and
those
are
the
best
sources
of
data.
A
We
have
to
look
at
these
Colonial
behaviors,
but
we
can
also
use
videos
that
you
saw
where
you
have
multiple
colonies.
You
have
some
diversity,
but
in
this
case
we
just
use
severe
some
isolated
colonies,
some
very
simple
movements,
smaller
colonies.
Just
so
we
can
get
a
handle
on
the
the
sort
of
structure
of
these
colonies,
the
structure
of
the
cells
and
place
them
in
a
spatial
reference
room,
but
the
cells
themselves.
You
can
see
that
we
can
also
characterize
different
parts
of
the
cell.
A
This
becomes
important
for
movement
because
remember
I
said
that
these
cells
slide
against
one
another,
but
they
don't
come
out
of
contact
there's
a
little
bit
of
the
cell
that
remains
in
contact
even
at
their
maximum
extent
of
movement.
So
you
see
that
we've
defined
a
number
of
landmarks
on
each
cell,
so
we've
defined
three
landmarks
where
you
have
each
end
and
then
the
centroid.
A
So
the
centroid
is
just
the
midpoint
between
the
two
ends
and
we
can
identify
those
through
computer
vision
techniques
and
then
we
can
define
a
bounding
box
defined
by
one
two,
three
four
and
five
four
and
five
being
each
side
of
the
cell.
So
basically
we're
creating
a
bounding
box
around
this
elongated
cell,
we're
defining
the
ends
of
the
cell,
we're
defining
you
know
the
left
and
right
edges
of
the
cell
and
then
the
centroid,
which
helps
us
to
sort
of
register.
A
This
allows
us
to
build
graphs
from
the
data
to
sort
of
characterize
the
shape
of
these
colonies
in
terms
of
numbers,
but
also
to
create
higher
order,
shape
parameters.
So
we'll
talk
about
that
in
a
little
bit,
but
basically
to
define
the
shape
of
these
colonies
as
they're
moving
to
do
this
work.
We
use
d-blat
version
three
and
we
used
our
open
Devo
cell
platform,
which
is
a
platform
internal
to
our
group,
with
some
mask
pre-processing.
So
this
is
a
little
bit
older
work.
A
It's
not
this
Cutting
Edge
state
of
the
artwork
that
you
might
expect,
but
again
it
was
a
pretty
decent
result.
So
if
people
are
interested
in
participating
in
this
and
doing
another
version
of
this
we'd
be
happy
to
talk
to
you
about
that.
So
this
is
what
the
result
looks
like
after
segmentation,
where
we
had
these
bounding
boxes
that
represent
these
cells.
A
So
here's
an
example
of
how
we
put
this
into
a
spatial
reference
frame.
This
is
this,
this
Colony,
it's
it's
expanding!
It's
it's
in
movement!
At
a
certain
rate,
you
can
see
it's
not
a
clean
movement.
You
have
some!
You
actually
have
a
little
bit
of
a
it's
none
directly,
an
accordion
like
thing.
It
looks
like
it's
kind
of
like
moving
in
a
two
different
directions.
So
it's
not
really
clear
what
the
input
data
is.
A
So
we
can
set
tell
you
where
the
centroids
of
the
cells
are,
and
you
can
see
that
sometimes
the
centroids
are
close
together
and
sometimes
they're
far
apart,
and
that
approximates
the
movement
across
this
County
or
the
position
across
this
County
and
differences
in
movement.
So
the
differences
in
these
centroid
positions.
If
we
take
the
first
derivative
of
position,
we
say
that
that's
some
sort
of
motion
in
that
direction.
A
So
if
we
take
different
images
from
different
time
points,
we
can
plot
these
out
and
see
the
motion
or
the
the
changes
in
position
over
time.
So
we
can
look
at
another
example
where
we
can
see
the
same
thing
here.
We
have
this
cell
and
you
know
we
have
to
orient
these
cells
in
different
ways,
so
we
can
turn
the
orientation
around.
A
We
tried
to
orient
them
in
the
same
at
the
same
orientation,
so
there
was
some
print
processing
done
there,
but
basically
we
have
the
same
situation
here
where
we
have
the
centroids
that
are
far
apart
and
close
together
and
the
close
together
ones
are
when
the
colony
is
folded
together
and
the
ones
that
are
farther
apart
indicate
there's
some
sort
of
movement
there
or
extension
of
movement,
Across,
The
Colony,
and
so
this
is
a
third
example
where
we
have
much
more
regularly
regular
spacing
between
the
centroids,
and
this
is
in
when
the
colony
is
expanded
in
this
manner.
A
So
in
this
case
there
are
two
leading
edges
cells
that
are
two
leading
edges
and
they're
kind
of
there's
a
Center
of
mass
up
here
where
it's
closed,
and
so
you
can
see
that
movement
but
they're
much
more
evenly.
Spaced
centroids
are,
and
so
you
can
see
that
it's
reflected
what
you
see
in
the
images
reflected
on
the
graph
and
then
a
final
example
is
where
we
have
this,
which
is
a
similar
type
of
configuration,
but
it's
flipped
in
in
another
Direction.
A
A
So
one
of
our
other
group
members,
Richard
Gordon,
Dr
Richard
Gordon.
He
published
a
book
chapter
in
2021.
It's
called
the
Whimsical
history
of
proposed
motors
for
diatom
motility,
and
this
is
chapter
14
in
the
book.
Diatom
gliding
motility.
A
So
this
chapter
focused
on
the
different
mechanisms
that
have
been
proposed
for
diatomy.
As
I
said
before,
we
know
that
they
move
they
know.
We
know
that
they
have
these
complex
movements,
but
we
know
they
also
don't
have
a
brain
and
they
don't
have
muscles.
They
have
actin
filaments,
but
they
don't
have
muscles,
and
so
what
is
driving
this
emotion?
What
is
it
sort
of
resemble?
What
is
it
approximating?
A
And
so
people-
and
this
goes
back
to
like
the
17th
century
or
maybe
the
18th
century,
where
people
have
been
proposing
different
types
of
mechanisms
for
movement
and
so
they've
used
these
sort
of
analogies
of
use.
Other
organisms
like
snails
and
humans,
they've
used
human
activities,
like
you
know,
rowing
on
a
kayak
they've
used
a
very
dated
references
like
there's
one
example
where
there
are
a
bunch
of
people
rowing
a
rowboat,
and
you
have
another
example
of
like
a
jet
engine,
and
you
know
there
is
another
example
of
opinion
on
a
rack.
A
That
are,
you
know,
doing
work
and
so
we're
you
know
just
moving
around,
and
so
there
are
a
lot
of
different
potential
mechanisms
or
I
guess
you
could
call
the
metaphors
for
movement
in
this
book.
Chapter
Dr
Gordon
talks
about
21
different
mechanisms
that
have
been
proposed
for
movement,
but
we're
still
in
a
closer
really
to
understanding
movement.
A
A
So
these
are
not
too
bad.
You
know
they
look
like
they're,
pretty
good.
You
know
ways
to
experiment
with
these
colonies
to
sort
of
break
down
individual
cells
and
the
relationship
between
the
cells,
the
movies
that
we
had
in
the
Deep
learning
experiments
were
collected
from
the
web,
so
people
amateur,
diatom
observers
have
taken
movies
of
this,
and
some
of
them
are
not
of
high
quality.
These
movies
are
much
higher
quality,
so
Thomas
as
a
retired
physicist.
He
actually
has
access
to
some
of
these
physics
tools.
A
These
bio
biomechanical
tools
and
biomechanical
models
to
analyze
the
data,
so
this
is
from
our
contribution
to
the
diatomic
lighting
motility
book
chapter
10.
This
is
from
towards
the
digital
diatom,
and
this
shows
how
we
did
this
modeling.
So
each
diatom
cell
is
independent
and
you
measure
the
distance
between.
If
we
take
the
extent
of
motion
outward,
we
can
take
the
distance
between
the
neighboring
cell,
the
end
of
the
neighboring
cell
and
the
end
of
the
cell
that
we're
measuring
so
diatom
number.
A
One,
for
example,
is
a
57
point:
Micron
distance
between
the
far
end
of
diatom
number
one
and
the
farm
of
diatom
number
two.
Then
we
can
do
the
same
for
diatom
number,
two
relative
to
diatom
number:
three.
We
get
it
a
little
bit
different
measurement,
it's
a
little
bit
larger
59.14
microns.
Now
we
can
do
this
all
the
way
down
the
colony,
so
we
can
we're
not
using
landmarks
like
we
were
in
the
last
paper.
A
A
We
can
also
do
really
interesting
things
that
are,
you
know,
have
their
home
and
statistical
physics
and-
and
so
one
thing
we
can
do-
is
look
at
a
phase
portrait
of
diatom
oscillations.
So
when
we
do
these
measurements,
we
can
actually
get
these
record
like
a
measurement
that
goes
from,
say,
57.5,
microns
down
to
zero,
microns
and
then
57
microns
in
the
other
direction.
So
from
that,
we
can
look
at
two
neighboring
cells,
in
this
case,
diatom
number
one
diatom
number
two
and
draw
phase
portraits
that
describe
the
motion
relative
to
one
another.
A
So,
as
you
can
see,
we
have
a
signal
here
that
goes
back
and
forth
between
one
side
and
the
you
know
one
one
extent
and
then
the
other
extent
as
you
go
in
the
other
direction,
and
you
can
see
this
face
portrait,
and
so
you
notice
that
the
face
portrait
is
nice
and
round
nice
and
circular
it's
a
nice
orbit.
But
you
do
have
these
points
here
which
are
squiggly
as
opposed
to
the
rest
of
the
face
portrait.
A
Actually,
you
see
some
squiggle
in
it,
there's
variation
in
the
movement
and
the
reasons
for
that
which
are
somewhat
on
top,
which
are
partially
anatomical
and
partially
because
of
the
the
medium
that
the
diatoms
are
in.
But
you
have
these
this
sort
of
area
where
it's
really
gets
squiggly
and
what
these
are
is
when
you
get
to
that
extent
of
the
motion,
the
the
largest
extent
of
the
colony
you
end
up
with
this
end
of
the
oscillation,
so
the
oscillation
extends
to
an
alert.
A
You
know
it
extends
all
the
way
out
and
then
it
stops
and
moves
in
the
other
direction
when
it
stops
and
moves
in
the
other
direction.
This
is
a
signature.
We
expect
that
there's
some
sort
of
signature
of
jerkiness,
which
is
something
like
the
fourth
order.
Fourth
derivative
of
position,
so
we
have
these.
We
can
calculate
position
at
any
one
point.
When
we
put
that
together
we
get
a
phase
portrait
and
then,
as
that,
as
that
changes
we
get
acceleration.
So
we
get
acceleration
that
we
can
measure
here.
A
We
can
integrate
over
a
certain
number
of
time
points
and
get
acceleration
and
then
at
the
at
the
fourth
derivative
of
position.
We
are
this
jerkiness,
which
is
very
quick.
It's
not
even
it's
it's
not
acceleration,
it's
actually
something
more
higher
order
than
that.
So
actually
it's
a
third
derivative
of
physician,
I'm,
sorry!
So
this
is
something
that
we
can
look
at.
A
We
need
to
have
apply
better
tools
to
this,
but
in
this
paper
we
just
basically
suggested
that
this
was
what
was
going
on
here,
and
so
you
know,
you
see
a
lot
of
there's
a
lot
of
things
you
can
extract
in
these
phase
portraits
and
you
can
also
plot
them
against
sine
waves,
so
we
can
have
a
sine
wave
at
a
certain
frequency.
A
We
can
look
at
these
oscillations
as
something
that
occurs
over
a
certain
frequency,
so
you
can
see
that
in
in
the
full
sort
of
extent,
when
the
thing
is
oscillating
like
an
accordion,
this
is
going.
You
know
it's
pretty
much
consistent
with
the
sine
wave.
Sometimes
you
get
shorter
position.
You
know
get
shorter
changes
in
position
here,
so
it
doesn't
track
exactly
with
the
sine
wave,
but
this
could
be
like
changes
in
Direction.
This
could
be
just
a
slowing
down.
There
are
a
lot
of
things
that
could
explain
this.
A
If
we
look
at
this
mapping
to
a
sine
wave
more
specifically,
we
can
you
know
we
can
see
that
you
have
these
variations
in
the
amplitude,
but
not
in
the
phase,
and
so
that's
that's
good.
That
suggests
it's
basically
a
sine
wave
and
we
can
actually
look
at
different.
You
know,
maybe
you
know
we
could
take
the
just
a
little
bit
further
and
say
if
we
expose
the
diatom
to
different
types
of
stimulus
or
you
put
them
in
different
media.
What
does
this
look
like
versus
the
sine
waves?
A
The
sine
wave
is
like
a
null
hypothesis
or
a
just
a
random
model,
so
you
can
actually
compare
it
against
the
sine
wave.
What's
interesting
as
well
is
you
can
watch
them
changes
in
in
behavioral
phase?
So
you
know
these
diatoms
are
oscillating
back
and
forth.
We
can
compare
that
to
a
sine
wave.
A
Sometimes
there
are
changes
in
the
amplitude
of
the
signal,
but
sometimes
the
signal
ends
or
it
attenuates
down
to
zero,
and
what
this
means
is
that
the
diatom
is
starting
to
stop
oscillating,
basically,
it's
oscillating
to
swim
and
it's
becoming
quiescent.
So
you
can
actually
observe
this
in
the
data
where
it
becomes
quiescent.
A
A
This
actually
can
be
applied
to
other
types
of
things
as
well.
So
there
are
these
things
in
physics
that
people
are
biophysics.
People
are
exploring
called
micro
swimmers
and
so,
like
I
said,
the
oscillation
and
the
patterns
of
oscillation
have
to
do
with
swimming
behaviors
for
these
colonies,
they're
swimming
through
a
water
column
and
the
water
sometimes
changes
its
opposition.
Sometimes
it's
you
know
salinity
it's
brackishness
Etc.
A
A
You
know
there
are
different
things
you
can
explore
here.
This
is
a
magnetic
helical
micro
swimmer.
So
this
is
a
a
magnetic
micro,
swimmer,
that's
made
of
a
metal
and
then
it's
put
in
a
magnetic
field
and
it
can
move
helically.
So
it's
swimming
heliquely.
It's
not.
There
are
no
muscles,
there's
no
actin,
there's
no
brain,
but
it's
it's
responding
to
a
magnetic
field
and
it's
moving
accordingly.
Through
a
medium
on
the
left.
We
have
thermoelectric
guidance.
A
So
we
have
these
little
micro
swimmers
which
again
are
about
the
size
of
a
diatom
and
they
move
around
these
little
balls
and
they
move
through
this
water
column
or
this
liquid
column,
and
this
is
what
they
call
thermoelectric
guidance.
So
you
can
use
this
to
to
guide
micro
swimmers
to
different
targets
as
well.
Micro
swimmers
are
very
useful
for
different
things
at
the
at
the
Nano
scale
or
the
micro
scale
things
like
their
medical
applications.
A
Now
that
we've
talked
about
some
of
the
work
of
characterizing,
the
phenotype
of
diatoms
in
particularly
basil
area.
The
second
part
of
this
talk
will
get
into
the
sensory
world
or
the
habitats
experienced
by
diatoms,
and
so
you
can
see
that
diatoms
they
exist
in
these
title
or
Coastal
ecosystems.
So
you
have
estuaries,
which
you
see
on
the
upper
left.
You
have
these
tidal
areas
where
they're
near
the
surface
of
the
water
or
near
like
the
cup,
the
shoreline,
the
river.
You
know
there
are
a
lot
of
rivers
and
streams
where
they
live.
A
So
some
of
the
specimens
that
Thomas
Herbert
got
were
from
the
Necker
River
in
Germany,
so
these
are
temperate
streams
and
late
in
rivers
and
lakes
that
Harbor
diatoms.
You
can
also
Harvest
them
from
estuaries
or
from
anything
near
the
surface
of
the
of
the
water
column,
and
so
you
can
go
in
and
you
can
get
a
jar
of
water
and
you
can
isolate
these
diatoms.
If
you
go
into
any
one
of
these
environments,
you
should
be
able
to
find
them,
isolate
them
and
culture
them
on
your
own.
A
They
are
difficult
to
culture,
but
not
as
difficult
as
other
cell
lines
and
organisms
so
and
they're,
probably
about
as
difficult
to
isolate
in
culture
to
see
elegans,
which
is
you
know
to
say
that
you
need
to
have
some
equipment
for
it,
but
not
too
much.
You
can
easily
they're
they're
plentiful
in
the
environment,
so
this
is
the
kind
of
sensory
world
that
they're
used
to
they're
used
to
light
they're
used
to
changes
in
the
light
column,
because
water
wants
light
from
the
environment
hits
the
water.
It
there's
a
degrading.
A
A
So
one
thing
we're
interested
in
is
their
responses
to
light
their
responses
to
chemical
variation.
In
a
lot
of
these
brackish
environments,
you
have
salt,
water
and
fresh
water.
So
you
have
this
mix
of
things
and
so
there's
a
mix
of
salinity
gradients
that
are
kind
of
interesting.
There's.
Also,
you
know
differences
in
light
differences
in
water
temperature
and
other
things.
So
there's
all
sorts
of
you
know,
stimulus
changes
that
you
could
observe
and
that
these
diatoms
have
a
response
to
so
one
interesting
area
of
research.
A
That
kind
of
is
in
the
same
vein
as
this
is
psychophysics
and
so
psychophysics
has
been
applied
to
a
number
of
animal
bottles.
This
referenced
by
Stebbins,
which
was
a
1970
talks
about
animal
psychophysics,
so
designing
experiments
either
optimized
for
Animals.
They
mean
non-human
animals,
but
of
course
you
can
apply
psychophysics
to
humans.
A
You
can
also
apply
psychophysics
to
Mouse,
so
this
is
where
mice
are
looking
at
cats
and
other
mice,
and
then
you
have
goldfish
and
then
maybe
even
diatoms.
Now
no
one's
done
working
diatoms
in
terms
of
psychophysics,
but
that's
kind
of
what
we're
proposing
happen
and
so
saris
published
something
in
2006
on
the
relational
psychophysics
in
humans
and
animals.
A
A
So
there
are
a
number
of
papers
that
this
that
discuss
the
idea
of
non-normal
cognition,
and
so
we
can
think
of
cognition
as
this
information
processing
mechanism
or
this
information
processing
capacity
that
doesn't
necessarily
rely
on
a
brain.
It
relies
on
complexity,
you
know
biological
complexity
and
it
it
relies
on
different
mechanisms
for
information
processing.
A
So
in
this
paper
by
lion
in
2015,
this
is
the
cognitive
cell,
bacterial
Behavior
reconsidered
they
consider
bacterial
behavior
and
bacteria
also
do
not
have
a
brain
as
non-oral
customer
and
they
Define
non-neuronal
cognition
is
the
total
set
of
mechanisms
and
processes
that
underlying
information
acquisition,
storage,
processing
utilization.
So
we
have
information,
acquisition,
information,
storage,
information,
processing,
information,
utilization,
it
doesn't
require
brain,
it
doesn't
require
a
brain,
it
doesn't
require
neurons.
A
It
requires
certain
you
know,
features
in
both
the
individual
cell
and
in
a
comedy
to
do
this,
and
so
we're
going
to
see
some
of
those
mechanisms
later
talk
about
the
extracellular
Matrix
being
ideal
is
a
memory
medium,
so
they
talk
about
the
extracellular
Matrix
being
important
as
memory
and
there
being
an
internal
model
in
the
cell,
that
that
interacts
the
neutral
cellular
matrix
that
produces
stigma,
Gene,
which
are
the
physical
traces
that
can
be
referenced
in
red
in
the
future.
A
So
if
you're
familiar
with
insect
colonies
like
ant
colonies
or
tournament
mountains,
they
use
the
pheromones
and
they
lay
pheromone
Trails
down
in
their
environment
in
other
con
specifics.
Other
other
insects
can
read
those
signals
and
follow
those
trails
and
they
have
an
internal
model
inside
of
their
physiology.
That
tells
them
to
recognize
that
and
to
follow
that
trail.
So
these
are
all
things
that
we
can
use
in
in
diatoms
as
well,
so
in
diatoms
there's
an
interesting
extracellular
sort
of
stigmergic
mechanism
and
those
is
miselage
Trails.
A
So
these
Nissan
trails
are
secreted
polymers
that
completely
encase
the
cell
and
are
responsible
for
adhesion
and
interactions
at
the
internal
environment.
So
these
miselage
trails
and
strands
they
work
to
connect
these
cells
together,
so
they'll
set
the
cells.
When
you
looked
at
those
movies
of
the
motion,
they
appeared
to
be
sticky,
so
they
didn't
fly
apart
when
they
extended
themselves
out,
but
they
also
didn't
move
smoothly.
A
There
was
some
noise
and
that
noise
was
because
you
have
these
miselage
strands
and
trails
in
between
the
cells,
but
also
they're
laid
out
in
the
environment,
so
that
the
cells
other
cells
know
where
to
go.
They
have
a
signal
in
the
environment,
they're
responsible
for
a
number
of
interactions
of
the
external
environment,
so
we
can
actually
also
look,
though,
at
the
wave
in
which
diatoms
process
environmental
signals-
and
so
you
know,
like
I,
said,
there's
a
lot
of
variation
in
the
environment.
A
They
actually
do
take
in
environmental
signals
like
light
and
chemistry
and
and
other
things.
And
so
what
are
the
you
know?
There's
a
a
pretty
rich
literature
in
the
diatom
community
of
some
of
these
sensory
interactions.
So
this
first
paper
is
the
effects
of
elevated
CO2
on
the
natural
diatomic
community
in
the
subtropical
Northeast
Atlantic.
So
this
is
in
just
a
standard,
diatom
Community.
There
isn't
a
lot
of
work
done
specifically
ambassadoria,
but
the
idea
is
to
get
characterized
dependent
in
diatoms
and
then
applied
to
Basil
area.
Basilary
may
vary
in
the
way.
A
This
is
done
there,
but
you
know
it
gives
us
a
model
to
go
forward,
so
the
diatom
cells
can
sense,
CO2
sensation
and
concentration,
and
that
affects
the
growth
of
the
cells
and
growth
of
the
colony
in
cases
of
high
CO2,
there's
reduced
grazing
pressure
and
an
effect
on
photosynthesis,
and
then,
in
this
paper,
Hayden
light
intensity
modulates
the
response
of
two
Antarctic
diatom
species,
the
ocean
ocean
acidification.
A
This
there
are
two
things
that
we
see.
The
first
is
that
CO2
and
light
availability
affect
diatom
physiology.
They
tested
three
light
intensities
and
they
saw
that
low
low
light
effects,
physiology
and
distinct
physiological
traits.
They
also
observe
localized
sources
of
chemokinetic
and
chemotactic
Mobility.
So
chemokinetic
is
where
you
have
chemical
signals
that
drive
movement
and
then
chemotactic
is
where
chemical
signals
Drive
movement.
In
a
certain
direction,
so
there
are
two
different
things
that
they
can
they're
observing
here
they
actually
observed
attraction,
pheromone,
guided
surge
to
find
mates.
A
So
this
is
interesting
because
there's
this
pheromone
guided
search
to
find
mates
and
to
restore
cell
size
and
that
times
have
this
weird
reproductive
capacity,
basically
the
accrete
material
to
restore
their
cell
size.
They
also
have
they
also
do
mating
in
in
some
ways.
So
there's
there's
this
relationship
between
these
chemical,
chemical
sensation,
movement
and
even
other
types
of
behaviors.
A
This
paper
by
bondoc
decision
making
physiology
of
the
benthic
diatom
seminovus
Robusta,
searching
for
inorganic,
inorganic
nutrients
and
pheromones,
and
this
other
paper
by
felsiatory
perception
of
environmental
signals
by
a
marine
diatom.
This
talks
about
diet
times
having
sensory
systems
for
detecting
and
responding
to
fluid
motion,
or
specifically
Shear
osmotic
stresses
and
iron
abundance.
So
they
have
these
senses
for
chemical
imbalances
in
the
environment,
but
also
physical
forces
in
the
environment.
So
this
is
important
when
you're
in
a
water
column
the
nutrients
are
fluctuating,
but
also
you're.
A
You
know
in
different
parts
of
the
water
column,
there's
different
Shear
forces
happening,
and
so
it
guides
your
motility,
so
Mal
we're
interested
in
building
a
model
around
propulsion
as
a
mode
of
movement.
So
we
want
to
build
a
psychophysics
model
that
is
focuses
on
propulsion.
It
focuses
on
you
know,
maybe
having
like
input
sensory
inputs
or
other
types
of
inputs
that
result
in
motion
and
variation
of
motion,
so
we're
interested
in
environmental
inputs
and
actin
filament
outputs,
acting
filaments.
A
Being
the
things
that
drive
motion,
we
have
a
solitary
move
movements
as
a
small
subset
of
behavior,
but
this
allows
us
to
view
what
we'll
call
this
COPD
or
Collective
pattern
generation
circuit,
we'll
talk
about
what
that
is
in
a
minute
and
then
psychophysics
models
account
for
a
wide
range
of
behaviors
outside
of
oscillatory
movements.
So
we
can
actually
look
at
other
types
of
behaviors
outside
of
this,
we're
just
going
to
focus
on
the
oscillatory
movements
for
now
to
sort
of
propose
a
psychophysics
set
of
psychophysics
models
for
this
work.
A
So
this
is
movement
generation.
This
is
the
ref
a
wall
and
there's
a
substratum.
So
if
you
put
the
the
ref
the
cell
on
a
substratum
with
the
wall
on
the
outside,
of
course,
it
can
move
against
that
wall.
The
substrate
could
be
water,
it
could
be
another
cell,
so
another
Rafa
wall,
you
can
vary
the
substratum
and
it
gives
you
you
know
different
results,
but
basically
it's
moving
across
the
substrate
in
One
Direction,
and
this
is
gliding
Locomotion,
so
they're
actually
looking
at
the
mechanics
of
this
in
this
paper.
A
This
is
an
example
of
how
this
works.
So
there's
a
cell
wall
substrate
in
the
movement
as
we
can
see
as
it's
moving.
There
are
a
lot
of
interactions
here.
We
have
the
actin
filaments
inside
the
cell.
We
have
this
myosin,
which
is
moving
against
the
actin
filaments.
We
have
a
membrane
complex
at
the
surface
of
the
Rafael
wall
that
transmit
this
transmit
the
forces
downward
these
misoage
adhesives
and
then
those
mucine
mesolage
adhesives
transmit
those
forces
against
the
substrate.
A
So
you
have
this
interaction
between
the
raphae
wall
and
the
substrate
through
these
adhesive
Mesa
lodges,
they're,
transmitting
forces
from
inside
the
cell,
and
that's
how
you
get
this
movement,
but
also
you
get
Collective
movements
when
two
cells
are
moving
against
one
another
when
the
others,
when
the
substrate
is
another
cell
wrapping,
we
can
use
a
series
of
coupled
sinusites
to
model
an
ideal
Collective
pattern
generator
so
again,
we'll
talk
about
these
Collective
pattern.
A
Generators
and
you
will
recall
those
sine
waves
that
we
modeled
for
basic
basilaria
movement
and
we'll
consider
that
for
some
of
our
Collective
pattern
generation
work.
So
we
have
these
half
phase
or
90
degrees
out
of
phase
movements.
We
have
quarter
phase
or
45
degrees
out
of
phase
movements
and
those
can
be
used
to
simulate
this
coordination.
So,
in
the
cases
where
you
know,
maybe
the
The
Colony
breaks
apart
or
it's
not
moving
in
the
way
that
maybe
we
expect
it
to
given
the
stimulus
or
when
we
change
the
stimulus
it
becomes.
A
A
Additive
noise
can
also
be
used
to
simulate
this
coordination.
So
in
the
graph
on
the
right
we
can
see
additive
noise.
This
is
kind
of
an
extreme
version
of
additive
noise,
but
we
have
this.
We
can
add
this
in,
and
this
is
what
it
looks
like.
So
we
can
actually
have
two.
You
know
half
phase
or
quarter
phase
signals
with
noise
and
that
can
be
used
to
model
movement
as
well
and
then
finally,
the
speed
of
movement
or
the
derivatives
position
can
also
be
varied
according
to
the
effects
of
the
stimulus.
A
So
here
we
have
an
example
of
a
central
pattern
generator.
This
is
something
we
find
in
nervous
systems.
This
is
a
series
of
neurons
that
are
interacting.
You
have
these
excitatory
synapses
and
inhibitory
synapses
that
are
driving
this
oscillation.
So
basically,
you
have
neurons
that
are
signaling
each
other
in
an
excitatory
or
inhibitory
fashion
and
they're
driving
this
oscillation
in
the
circuit.
A
So
you
can
see
that
there's
an
input
from
C1
and
C2
and
their
feedbacks
between
X1
and
X2,
X1
and
X2
are
both
being
modulated
by
C1
and
C2,
as
well
as
being
co-modulating
each
other
and
responding
to
external
inputs.
So
you
can
see
how
that
works.
That's
a
collective
or
that's
a
central
pattern
generator,
and
we
see
this
in.
Like
the
spinal
cord.
We
see
this
in
other
parts
of
like
insect
nervous
systems.
We
see
this
in
mammalian
nervous
systems
where
you
need
to
generate
a
rhythm
for
something.
A
So
you
see
this
in
the
cardiac
system
and
other
places
it
generates
a
rhythm,
that's
continuous.
So
it's
just
the
activation
of
this
circuit
and
it
generates
a
rhythm,
that's
a
central
pattern.
So
this
is
a
central
pattern.
Generator
it's
been
talked
about
in
the
Neuroscience
literature
for
decades.
A
What
we're
talking
about
here
is
a
collective
pattern,
generator
which
is
different
in
a
collective
pattern.
Generator.
We
have
these
inputs
to
the
Leading
Edge
cell,
as
we
saw
in
our
movies,
and
we
discussed
already
so
this
Leading
Edge
of
cell
will
take
in
these
in
sensory
inputs,
and
then
we
have
these
they're
landmarks
they're,
not
giving
their
just
sites
where
you
have
sensory
inputs
and
then
the
sensory
inputs
are
linked
to
other
cells,
Down
The
Colony.
So
you
have
these
links,
these
mechanical
links.
A
You
have
other
sensory
links
between
cells
and
so
that's
what
this
network
is.
So
it's
just
a
collection
of
sites
where
either
you're
have
an
input
of
sensory
information
or
you
have
an
interaction
with
another
cell,
and
so
we
use
the
W
script
for
w
subscript
for
these
to
identify
these
different
sites.
And
then
we
have
a
network
of
different
relationships
between
these
points,
so
we
can
see
that
they,
because
these
cells
move
against
one
another
and
they're
all
experiencing
the
environment
in
different
spatial
locations.
A
The
you
have
a
lot
of
potential
interactions
between
them,
and
so
that's
what
these
lines
are.
Their
interactions,
it's
almost
kind
of
like
a
neural
network,
but
it's
not
quite
because
it's
structure.
First
of
all,
it's
spatially
inhomogeneous,
but
also
because
it's
actually
more
of
a
interaction
model
than
a
neural
network.
So
we
can
model
this
and
we
can
look
at
the
variation
in
these
different
interactions,
the
strength
of
the
interaction
over
movement,
over
stimulus
intensity
and
over
time.
A
So
our
sensory
inputs
are
I
sub
n.
These
can
be.
You
know
these
can
be
varied.
I
just
put
three
here
as
an
example.
These
are
sensory
inputs
at
inertial
points
on
a
surface
of
the
leading
cell.
Then
we
have
movement
weights,
so
the
movement
weights
are
these
weights
of
interaction.
These
are
the
way
we've
modeled.
A
This
is
largely
about
movement
and
about
the
interactions
between
Rapha
walls,
but
we
can
also
model
some
of
these
W
sub
MN
points
as
environmental
inputs
as
well
depending
on
the
movement
of
the
colony,
depending
on
which
ones
are
exposed
to
the
environment
at
different
times,
and
then,
of
course,
you
know,
they're
also
interaction,
points
for
mechanics
and
movement,
so
this
is
what
we're
talking
about.
A
So
the
idea
is
that
this
model,
if
we
modeled
it,
should
describe
these
interactions
and
should
describe
the
movement
so
using
our
COPD
model,
we
can
look
at
differences
in
stretch
phase
between
coupled
cells,
so
on
left
we
have
this
model,
this
idealized
model
of
our
copg,
and
then
we
have
directional
movement
back
and
forth,
and
then
we
have
our
microscopy
images
where
we
have
an
example
from
these
images.
A
So
we
can
see
that
we've
kind
of
labeled
the
weeding
Edge
in
this
case
I
think
we've
labeled
the
trailing
edge
here,
but
we're
kind
of
pretending
that
that's
the
Leading
Edge
of
the
cell,
so
whatever
wherever
we
could
have
in
environmental
inputs,
on
the
trailing
Edge
as
well.
But
in
any
case
we
have
one
cell
where
these
inputs
are
coming
in
and
of
course,
ideally
they
could
come
into
any
cell
and
then
we
are
moving
back
and
forth.
A
So
you
can
see
that
these
inputs,
these
environmental
inputs,
are
transmitted
Across,
The
Colony,
as
it
moves
back
and
forth,
and
so
based
on
the
configuration
you
can
see
the
spatial
extent
of
the
oscillation,
these
inputs
on
a
single
cell,
and
then
these
networks
being
arbitrary
points
within
this
within
this
Colony.
So
there's
a
network
that
forms
within
the
colony,
that's
a
network
of
information
processing,
and
so
that's
basically
the
idea
and
then,
as
you,
vary,
the
environmental
signals.
This
network
changes
in
terms
of
information
processing.
A
It
drives
the
oscillation
which
is
not
autonomous
movement,
but
it
affects
that
in
different
ways.
We
can
also
look,
then,
when
we're
talking
about
environmental
inputs
at
sensory
thresholds
or
just
noticeable
differences.
So
this
is
another
core
attribute
of
psychophysics.
This
describes
differences
in
light
intensity,
for
example,
measured
by
output
movement
weights.
So
what
this
means
is
that
you
have
a
light
source
again:
it's
I
have
the
sun
there,
but
you're
going
to
probably
have
it
attenuated
by
the
water
column.
A
These
light
sources
are
going
to
come
into
the
different
points
at
different
intensities,
and
we
expect
that
there's
near
the
sensory
thresholds,
where
the
light
is
intense
enough
to
trigger
a
change,
sometimes
light
will
be
you
know
normal
or
it
will
be.
You
know
not
noticeably
different
from
normal,
so
it
might
fluctuate
a
little
bit,
but
it
might
not
affect
the
movement
that
much,
but
sometimes
it's
going
to
be
above
that
noticeable
difference
threshold
and
affect
the
information
processing
Network,
and
so
we
can
measure
those
differences
in
light
intensity.
A
By
looking
at
the
structure
of
this
network.
Looking
at
the
motion
and
even
modeling
the
the
diatom
and
proposing
that
they're
different,
you
know
noticeable
difference,
thresholds
or
other
types
of
information
processing
properties
of
this
network.
A
We
can
also
get
the
signal
to
noise
ratio.
So
one
of
the
reasons
I
put
this
noisy
signal
in
is
to
point
out
that
there
is
a
lot
of
noise
in
the
environment,
and
so
you
know
the
ability
of
the
diatom
to
separate
out
that
signal
from
the
noise,
especially
in
terms
of
measuring
forces
or
measuring
chemical
inputs,
is
important.
So
there's
this
Dynamic
sensing
at
the
water
column,
the
physical
properties,
measured
by
output
movement
weights.
So
when
there's
an
input
again,
there's
a
lot
of
noise
to
that
input.
A
Can
this
information
processing
networks
separate
the
signal
from
the
noise
and
operate
on
the
signal
and
that's
another
thing
we
can
think
about
in
terms
of
psychophysics
because
signal
the
noise
ratio
is
something
we
can
play
around
with
and
see
if
it
makes
it
effect.
It
makes
a
difference
on
the
movement.
I
should
note
that
when
we
looked
at
those
phase
portraits
of
the
movement
between
two
cells
that
we
saw,
these
noisy
fluctuations
in
the
in
those
orbits-
and
some
of
that
is
due
to
sort
of
the
extent
of
the
The
Colony
moving
against.
A
You
know
to
its
largest
extent
space
initially,
but
the
other
part
of
that
is
noise
inherent
in
this
flux
in
this
movement
against
one
of
the
cells
moving
against
one
another,
and
perhaps
some
of
that
is
due
to
this
information
processing
within
the
colony,
and
so
that's
something
we
can
look
at
more
deeply
as
well.
Using
this
kind
of
method.
A
So
one
example
of
psychophysics
is
Weber's
law,
so
this
is
where
we
look
at
the
intensity
of
the
light
stimulus
versus
the
ability
to
discriminate,
different
light
sources.
So
sometimes
the
light
isn't
coming
from
One
Direction.
As
you
saw
in
these
environments,
you
know
you
have
light,
is
being
diffracted,
and
so
it's
coming
from
different
directions.
There's
this
sort
of
effect
when
you're
underwater.
That
is
where
the
light
is
indirect.
A
It's
kind
of
I
can't
remember
the
name
of
the
effect,
but
it's
it
diffuses
the
light
source
enough,
so
that
the
diatom
needs
to
be
able
to
discriminate
between
different
the
True,
Light,
Source
or
different
light
sources,
and
this
holds
true
with
chemical
inputs
and
physical
inputs
as
well,
especially
with
chemical
inputs.
When
you
have
a
lot
of
different
chemical
compositions
in
the
environment
that
you
to
be
discriminated
against,
so
we
can
use
something
like
Weber's
law
to
quantify
these
more
directly
and
do
experiments
like
that.
A
So
I've
talked
about
a
lot
of
things
in
this
talk.
This
is
our
roadmap
for
the
digital
basil
area.
So
far,
I've
talked
about
modeling
I've
talked
about
data
segmentation
extraction,
I've
talked
about
characterization,
so
we
have
this
roadmap
for
the
digital
Massillon
area,
we're
operating
mostly
from
microscopy
data,
but
some
of
that
also
involves
you
know,
behavioral
data
and
maybe
even
simul,
you
know
simulating
conditions,
but
let's
take
the
microscopy
data
is
this
is
the
root
for
our
roadmap?
A
What
we
do
need
is,
we
do
need
a
lot
of
microscopy
data
data
on
movement,
on
movies,
data,
on
static
observations
and
so
forth,
and
those
lead
to
do
on
two
different
paths.
We
have
a
deep
learning
path
and
a
physical
modeling
path
and
the
physical
modeling
path
is
heavily
dependent
on
dynamical
systems
in
the
Deep
learning
path
allows
us
to
look
at
things
like
Colony
shape
and
within
Colony
shape.
We
have
a
number
of
things
that
we
would
like
to
look
at
we'd
like
to
get
accounts.
A
If
the
positions
that
these
cells
take
with
relation
to
a
colony,
we
have
a
variation
of
that
shape
or
of
the
colony
shape.
So
we
saw
those
images
where
you
have
a
number
of
different
colonies
in
the
same
image
and
they
all
look
different.
They
all
have
different
compositions
and
they
all
had
different
properties
and
as
well.
You
know
we
also
want
to
look
at
things
like
topological
properties
of
the
colony
and
its
shape,
so
using
a
topological
data
analysis
approach.
That's
with
respect
to
time,
so
we
can
use
deep
learning
for
that.
A
We
can
also
use
the
psychophysical
modeling.
For
that
part
too,
in
terms
of
the
physical
modeling,
we
are
directly
addressing
dynamical
systems,
so
there
are
a
number
of
things
that
are
interesting
and
dynamical
systems
land.
Some
of
those
are
like
the
collective
pattern,
generators
the
psychophysics
themselves
in
the
moments
of
position,
and
you
saw
the
examples
of
phase
diagrams
and
other
work
that
you
know.
A
We
have
information,
processing
networks
and
you
know
other
types
of
things
that
are
very
interesting,
so
that's
kind
of
our
road
map
for
developing
the
digital
basso
area,
and
hopefully
you
know
in
doing
this,
we
build
a
better
sort
of
account
of
diatoms
of
bacillary,
and
why
do
we
care
about
diatoms?
Well,
diatoms
are
very
important
for
the
first
of
all.
They
represent
a
lot
of
biomass
centers.
A
You
know
they're
also
applicable
to
things
like
micro,
swimmers
or
nanotechnology,
and
you
know
they're
also
they're
they're
important
in
like
biofuel
production
and
other
things,
so
understanding
how
we
can
culture
diatoms.
How
we
can
you
know
address
their
behavioral
variability
in
their
physical
variability
is
important.
A
So
thank
you
for
your
attention.
This
work
was
done
within
the
medieval
one
group,
so
you
can
visit
our
website
or
GitHub
repository,
and
we
also
have
a
a
deep
learning
platform
called
Divo
learn
which
involves
graph
neural
networks
and
free
trained,
deep
learning
models
and
other
types
of
work,
and
that
can
be
found
at
that
Repository
Thomas
harbick
has
run
a
website
called
Observation
of
diatoms,
where
he's
put
out
a
lot
of
his
observations
on
culturing,
diatoms
and
taking
images
of
them
and
then
check
out
our
weekly
Diva
meetings.