►
From YouTube: Structural Plasticity in the Neocortex
Description
Paper review: https://www.researchgate.net/publication/26753678_Holtmaat_A_Svoboda_K_Experience-dependent_structural_synaptic_plasticity_in_the_mammalian_brain_Nat_Rev_Neurosci_10_647-658
Holtmaat A, Svoboda K. Experience-dependent structural synaptic plasticity in the mammalian brain.
Also: https://www.ncbi.nlm.nih.gov/pubmed/26354919
A
B
C
A
C
B
A
B
So
since
we
didn't
have
any
other
paper
this
week,
I
thought
I'd
just
do
a
very
quick
review
from
a
structural
plasticity
from
the
neuroscience
standpoint.
I
think
a
lot
of
this
we've
already
kind
of
talked
about
here
and
there
anyway,
so
shouldn't
be
a
big
deal.
They
will
go
through
fast,
maybe
not,
but
it
is
very
relevant
to
the
it's.
It's
the
kind
of
neuroscience
basis
for
a
lot
of
the
work
we're
doing
in
terms
of
dynamic
sparsity
right
now,
and
so
it's
it's
useful
to
at
least
so.
I'll
go
over
two
papers.
B
This
is
the
Holt
med
and
a
little
bit
of
paper,
and
then
I
will
briefly
go
through
the
the
Lowenstein
at
all
paper
as
well
and
I
thought.
This
is
a
pretty
nice,
concise
review.
It
might
be
hard
for
people
who
are
not
used
to
neuroscience
to
parse
through
the
paper
and
things.
So
what
I
thought
I
would
do
is
start
with
the
definition
of
structural
plasticity
from
the
Nernst
Einstein
point
and
then.
B
C
A
B
B
So
he
goes
through.
This
makes
me
sort
of
the
various
sections
he
has
in
his
papers
that
match
some
of
this
stuff
here.
So
I'm
not
really
going
to
cover
this
one
with
axons
growing
to
new
areas.
He
kind
of
shows
a
tuck
so
that
this
primarily
happens
when
there's
some
sort
of
injury
and
what
they
find
is
that
if
there's
some
big
injury,
like
some
part
of
the
brain
or
I,
think
there
can
be
really
massive
changes
in
the
absalon's.
B
So
and
basically
this
is
like
primary
motor
cortex
is
kind
of
level
1
of
the
motor
area
and
the
premotor.
It
can
be
level
2
level,
3
and
so
on.
So
you,
if
you
damage
level
1
now
you
can
get
axons
from
level
2
going
down
to
where
maybe
the
m1
primary
motor
cortex
axons
used
to
go
and
similar
things
here.
B
The
various
types
of
injury
like
if
you
amputate
one
of
your
limbs,
the
neurons
that
used
to
correspond
to
that
can
start
growing
axons,
that's
several
millimeters
difference
which
is
huge
in
the
in
the
neocortex.
So
when
there's
exhibition,
so
when
there's
injury,
you
can
have
very
large-scale
reorganization,
but
without
injury
it
seems
like
the
axonal
changes
are
relatively
small
and
you
don't
get
these
large-scale
changes
happening
and
I
think
the
same
basic
conclusion
happens
is
therefore
dendrites
that
dendritic
Arbor.
So
if
there's
injury
you
can
get
large-scale
changes.
B
B
B
Here's
an
example
of
that
happening
here,
so
these
are
days.
So
this
is
day
eighty
to
eighty
six.
Ninety
you
can
see
that
here
the
axons.
So
this
is
a
case
where
the
axon
changed
by
a
few
microns
is
not
large-scale
changes
that
small
scale
changes
still
occur
constantly.
So
here's
an
axon,
that's
now
growing
and
it
yeah
you've
got
these
new
Bouton's.
Our
synapses,
for
me
here
are
kind
of
synapses,
growing
and
drug
dying
over
several
days.
B
B
B
B
B
B
B
A
B
How
long
it
lasts?
It's
not
just
related
to
know.
Why
was
it
far?
That's
also
related
to
a
competition
yeah.
It's
a
competition.
Also
the
size
of
the
synapse
is
a
big
deal
and
actually
that's
as
the
synapse
has
become
more
and
more
active.
The
spines
grow
over
time.
So
the
size
is
a
proxy
for
like
how
often
it's
being
used.
B
That's
good
I,
don't
know
how
much
I'll
be
able
to
answer,
but
I
couldn't
try,
and
so
we
can
like.
We
could
I.
Remember
we
still
have
you
know.
Remember
you
a
he
had.
He
was
a
neuroscientist
who
was
here
before
he
had
a
whole
semester
long
course,
just
on
synapses.
It's
fine!
So
there's
a
ton
of
detail
here,
there's
a
lot
of
which
I
don't
know,
but
it
kind
of
looks
here's
a
better
picture.
B
What
I
was
drawing
here,
so
here's
the
green
thing
is
like
a
spine
and
the
red
thing
might
be
an
axon,
and
these
are.
This
is
a
pluton
that's
forming
and
then
in
here
you
have
these
vesicles
which
release
neurotransmitter,
and
so
whenever
the
synapse
become
active,
there's
neurotransmitter
released
from
one
to
the
other
and
the
more
nervous
transmen.
So
if
you
have
LTP,
which
is
like
increasing
the
weight
of
synapse,
then
you
get
more
of
these
vesicles.
They
release
more
of
these
neurotransmitters.
A
B
B
B
A
B
B
C
B
That's
a
separate
topic
in
the
HTM.
What
we've
done
is
we've
simplified
it.
We
just
said
it's
binary:
it's
either
on
or
off.
Okay,
and
that's
probably
not
completely
correct,
but
it's
a
safe
stopping
point.
Is
he
uses
a
word
wave
yeah?
How
much
I
think
it's
a
great
question,
I
think
I
would
recommend.
We
have
a
separate
there's,
a
lot
of
complexity
there
and
I've
seen
in
the
papers.
I've
seen
some
people
say:
no,
it's
really
like
a
machine
learning
in
between.
A
B
B
Here
so
where's
the
logic
for
adding
and
dropping
so
one
big
one
is
its
activity
dependent.
So
you
can
see
here
that
the
presynaptic
activity
and
glutamate
can
trigger
spine
growth,
so
that
gets
a
little
bit
to
the
chemical
side
of
it.
But
basically,
new
protrusions
seem
to
preferentially
grow
towards
axonal
bouton,
with
active
synapses,
often
ignoring
other
potential
elements
in
the
near
vicinity
of
the
dendrites
all
right.
So
this
is
the
basic
wire
to
get
a
fire,
get
our
water
together
philosophy.
B
A
B
B
C
A
B
B
B
C
A
Ladies,
the
accent,
fires
and
it's
got
a
blue
tag
on
there
might
be.
You
know
it
might
get
knit
a
little
bit
of
chemical
or
something
like
that
I
just
hypothesized,
but
that
would
you
know.
So
how
would
you
you
know?
Obviously
what
you
have
you
know:
inactive
synapses,
the
chemicals
once
the
best
will
come
across
they
kind
of
release
on
there
at.
Oh,
that's,
that's
really
cool.
But
what
could
imagine
that
there's
something
it's
basically
in
order
to
have
these
guys
rendezvous,
you
know,
there's
got
to
be
signals
back
and
forth.
Yeah.
A
A
A
C
B
Factor
for
me,
stasis,
which
is
another
factor,
but
the
primary
way
is
sort
of
activity
like
both
of
them
are
Far.
East
I
think
it
must
be
a
two-way
process
like
somehow
these
blue
tones
know
to
grow
because
there's
some
dendrite
nearby,
that's
active
and
somehow
these
dendrites
know
how
to
grow,
because
there's
blue
tones
that
are
active
for
sounds
that
are
active.
So
that's
I
think
it
has
to
be
a
two-way
process,
but
the
exact
chemical
basis
for
it
I,
don't
know
like
I
said:
that's.
A
B
A
B
B
B
B
B
Segments
there's
only
so
much
so
many
synapses
that
could
be
there
right.
So
if
you
start
growing
another
one,
you
have
you
eventually
you'll
have
to
drop
one
okay.
So
that's
it's!
This
is
not
directly
tied
to
the
activity
between
these
these
two,
but
it
sort
of
indirectly
something
will
disappear
because
there's
too
many
other
things.
So
then
it's
kind
of
maintaining
a
certain
sparsity
level
standpoint.
All
right
said
our
dynamic
sparsity.
What
week
we're
have
to
drop
something
in
order
to
add
something.
B
B
B
B
B
Attended
this
one
okay,
so
let
me
highlight
stuff
here,
so
one
really
interesting
picture
graph
I
think
is
this
one?
So
what
they
did
is
they
looked
at
three
thousand
six
hundred
eighty
eight
spines
from
eight
neurons
in
six
mice
over
multiple
weeks-
and
there
were
this-
is
some
amazing
actually
able
to
track
these?
Like
one
micron
level
things
over
multiple
weeks
in
lots
of
and
a
bunch
of
dendrites.
B
A
C
B
Every
day
or
every
four
days,
yeah
interval,
for
it
is
that
so
that's
a
huge
turnover.
They
also
contract
each
individual's
fine
and
they're
looking
at
the
size
of
it
and
how
long
it
stays
and
so
on.
So
I
found
these
two
graphs
quite
interesting.
So
here
what
so
there?
You
can
remember
each
spine,
seen
at
all
during
the
experiment
from
one
to
three
thousand
six
hundred
eighty
eight
they
track
kind
of
when
it
appears
and
when
it
disappears.
B
These
are
the
new
synapses
that
came
in
the
next
imaging
session,
so
there's
a
whole
bunch
that
appeared
and
then
the
next
time
a
bunch
of
them
disappeared,
but
some
do
stay
the
whole
time.
So
whenever
you
get
a
whole
bunch,
new
most
of
them
will
disappear,
but
then
some
will
become
will
be
stable,
yeah.
You
know
in
the
paper
or
something
yeah.
B
Yeah,
this
is
sort
of
what
you
wanted
to
measure
it
like.
How
long
do
these
things
say?
So
it's
okay,
most
of
them,
disappear
as
long
as
some
state.
This
kind
of
matches,
my
intuition,
with
the
temporal
memory
as
well
I
think
what's
happening,
is
when
we
are
in
the
temple
memory,
we're
constantly
adding
and
dropping
synapses,
but
there's
a
in
a
noisy
data
set.
Most
of
these
correlations
are
going
to
be
transient,
they're
they're,
noisy,
but
a
few
will
be
will
be
stable
and
persistent.
B
C
B
B
So
here
it's
looking
at
how
many
synapse
spines
are
on
each
neuron
I
think
is
that
there
were
eight
neurons
across
the
day
and
you
can
see
it's
very
stable.
So
even
though
you're
constantly
growing
and
adding
and
it's
very
dynamic,
the
absolute
number
stays
about
the
same.
So
this
is
this
homeostasis
again.
B
A
B
C
B
It's
a
different
neuron,
yeah
yeah
heard
Nora
indicated
by
different
donors,
and
a
finger
should
be
eight
yeah,
eight
of
them,
so
that's
pretty
cool.
This
is
like
how
long
a
given
synapse
lasts.
So
if
you
look
at
all
the
synapses
that
word
new,
it
follows
this
power
loss
and
a
probability
of
blasting.
This
many
days
drops
off
as
a
as
a
power
law.
It's
not
a
moment
can
happen.
Yeah,
and
these
are
the
these
are
the
existing
synapses
that
were
already
there,
the
very
first
time
they
measured
it.
B
So
there's
those,
obviously
they
also
follow
our
law,
but
these
are
like
synapses
that
likely
are
real,
strong
synapses,
so
they're
not
going
to
it.
Doesn't
it's
not
a
steepest
of.
B
B
B
Okay,
and
here
this
is
the
correlation
between
the
size
of
the
spine
and
its
age.
We
find
that
the
probability
that
a
spine
would
survive
is
a
monotonically
increasing
function
of
its
size.
Okay,
so
large
synapses,
large
spines
tend
to
stay
off.
So
that's
good
too
I
think
that's
pretty
much.
All
I
wanted
to
cover.
B
So
there's
a
lot
of
detail
about
what
a
spine
is
and
what,
how
it
grows
and
stuff
that
I
think
we
should
lead
to
a
lot
of
that.
I,
don't
know
and
I
think
we
should
leave
that
to
a
separate
session.
But
the
basic
results
are
that
the
network
is
very,
very
dynamic.
Actually
in
here
I
thought
he
had
some
nice
language
right
in
the
beginning.
B
Given
that
in
most
areas
of
the
brain
neurons
are
sparsely
connected,
structural
plasticity
could
provide
a
substantial
boost
in
the
memory
capacity
compared
with
plasticity
due
to
changes
in
synaptic
strength
alone,
structural
rearrangements
over
long
distances
allow
more
variability
and
therefore
a
large
number
of
potential
circuits
to
be
generated.
So
it
probably
allows
the
brain
to
be
very,
very
flexible
in
the
in
the
set
of
networks
that
are
possible.
You
mentioned
that
they
were
there.
B
So
that
backs
on
I
think
so
so
what
I
can
tell
you
is
there
are
several
different
types
of
reactions
that
I
know
of
so.
First
of
all
there
are
chemical
synapses
and
then
there
are
electrical
synapses,
okay,
so
everything
we've
talked
about
here,
being
chemical
synapses.
So
in
order
for
anything
for
communication
death
and
you
release
these
neurotransmitters,
which
have
to
kind
of
travel
to
the
other
side
and
have
an
impact
on
the
electrical
potential
on
that
other
side
and
that
takes
about
a
millisecond
from
what
I
understand.
B
So
electrical
synapses
are
much
faster.
Some
people
think
that
they
are
used
for
synchronizing
cells,
because
you
imagine
you
have
oscillations
and
each
one
part
of
the
brain
is
oscillating
at
a
in
another
part
of
the
on-state
brain
is
oscillating.
If
you
want
their
phases
to
be
locked
in
order
to
do
that,
it's
very
helpful
to
have
really
fast
connections
that
can
lock
things.
B
Within
the
chemical
synapses,
there's
different
types
of
reactions
that
occur
at
different
time
scales
and
have
different
impacts.
So
that's
another
thing,
so
we
typically
think
of
you
get
some
input.
You
make
a
weight
change,
it
happens
right
away
and
that's
it
I
think.
That's
the
that's
not
what
happens
in
the
brain.
There
are
different
timescales
of
plasticity,
so
some
time
scales
are
very
short.
Some
time
scales
are
very
long.
B
B
C
B
A
B
Because,
basically
it's
it
keeps
one.
One
theory
is
that
that
part
of
then
right
becomes
very
plastic
for
a
while
for
a
second.
So
if
you
have
a
high
frequency
burst,
and
then
it
invokes
this
medical
tropic
effect
now,
any
new
thing
will
get
learned
really
quickly
or
for
a
second
more
and
that's
kind
of
what
you
want
for
type
of
cooling.
B
Thank
you,
for
you
know
telling
me
anything
about
like
a
correctly
predicted
cell
and
it's
suppose
that
were
to
do
a
mini
burst,
then
that
could
invoke
them
that
a
petrale
receptor
that
could
cause
pooling
and
it's
kind
of
what
modeled
in
our
American
folklore.
That
idea
and
it's
very
complicated
histones
of
these
things.
B
B
A
B
B
B
A
B
B
But
he
says
things
like
if
you're,
if
you're
in
there
sensory
deprivation,
this
time
like
you're
in
just
in
darkness
for
a
long
time
and
then
you
starts
saying
like
then
there's
a
lot
more
growth
or
if
you're,
in
a
very
rich
environment
like
if
you're
in
a
really
boring
room
and
that's
your
entire
life,
you
don't
learn
much,
but
if
you're,
a
very
rich
environment
with
lots
of
stuff
going
on,
you
get
more
growth,
so
I
kind
of
lump
all
of
those
as
most
of
those
as
being
activity
dependent.
Would
you
see.