►
From YouTube: HTM Hackers' Hangout - Apr 6 2018
Description
HTM Hackers’ Hangout is a live monthly Google Hangout held for our online community. Anyone is free to join in the discussion either by connecting directly to the hangout or commenting on the YouTube video during the live stream.
If you have something specific you’d like to discuss, or if you just want to learn more about the HTM Community, please join HTM Forum at https://discourse.numenta.org. We have active discussions about HTM theory, research, implementations, and applications.
More info on all these topics at http://numenta.org.
A
Hi
welcome
to
HTM
hackers,
hangout
I'm
Matt
Taylor
from
the
mid-south.
It's
April
6th,
2018
I'm,
going
to
go
over
a
few
topics
and
we'll
cover
some
things,
came
up
on
the
forum
and
then
we'll
open
it
up.
If
anyone
else
wants
to
talk
about
anything
at
all
apologize
up
front,
if
my
computer
fan
suddenly
gets
loud
but
keeps
happening
ok.
So
if
you
guys
oh
I'm,
going
to
share
my
screen,
it's.
A
Obviously
all
the
different
types
location
cells,
I'm,
not
all
of
them
and
specifically,
but
why
there's
so
many
different
types?
What
do
we
have
to
model
every
single
type?
You
know
stuff
like
that?
How
are
they
expressed
in
HTM?
How
are
they
expressed
nation,
neurons
and
dendrites?
However,
they
expressed
in
layers
what
do
they
express
in
layers?
What
are
the
processes
of
creating
layers
that
depend
on
grid
cells
or
location
cells?
So
those
sorts
of
questions
I'm
going
to
ask
them
about
different
brain
areas.
A
Like
the
hippocampus
and
thalamus,
what
zip
campus
doing
how's
it
contributing
to
neocortical
function,
others
who
think
that
they're
relating
together
when
things
are
actually
happening,
same
thing
with
the
thalamus
any
new
clues
on
what's
going
on
with
that
relay
or
what
we
think
it
might
be
doing,
and
then
I'm
gonna
talk
to
about
oscillations,
because
I
think
oscillations
are
interesting.
Is
that
a
process
is
a
signal?
Is
it
a
signal,
processing
itself?
I,
don't
know.
Is
it
some
indication
of
attention
the
different
types
of
oscillations
in
the
brain?
A
Why
are
there
different
types
more
they?
What
could
they
be
doing
so?
Some
high-level
stuff
like
that
and
then
I
again
I'll
talk
about
object,
representation
because
for
the
longest
time
I
described
it
as
it's
very
tempting
to
want
to
want
to
describe
how
object
recognition,
representation
in
your
brain
works
by
saying
it's
everything
that
you've
sensed
at
some
location
in
space
you
know
and
you're
against
a
model
in
your
brain
and
I
wanted
I
wanted
it
to
be
like
that.
But
it's
just
not
like
that,
because
I
think
it's
so
intricately
tied
to
motor.
A
A
A
You
don't
eat
myself,
talk
by
sharing
my
screen
again
now:
yeah,
okay,
so
the
other
couple
things
was
I
came
up
here.
That
mark
asks
that's
mark
brown
yeah
anachronism
from
earlier
versions
of
HTML
sort
of
like
what's
old
and
what's
new,
what's
replaced
it.
So
this
is
a
it's
it's
hard.
What
I
can
definitely
tell
you
is
we
don't
use
the
backtracking
temporal
memory
at
all
and
any
new
stuff?
A
A
So
the
old
white
paper,
the
hold
old
white
paper,
I,
think
we're
just
we
tried
to
remove.
You
know
links
to
that.
For
the
most
part,
that's
misleading,
but
that's
just
because
it's
all
but
I,
don't
think
anything.
Do
you
think
of
anything
Marcus,
that's
hanging
out
there
that
sort
of
Holden
faded
well
yeah
I,
couldn't
tell
by
the
question
if
it
was
asking
about
about
before
the
about
what
has
changed
between
the
white
paper
and
Bamie
or
between
Bamian
now
between
Bamian
now
pretty
much,
nothing
has
changed.
A
It's
just
we're
researching
more
things
that
will
eventually
end
up
in
back
as
editions
between
the
white
paper
and
family
they're,
really
quite
similar,
they're.
Really,
the
biggest
difference
is
that
I
would
say
what
was
what
used
to
be
called
the
temporal
cooler.
You
know
called
temporal
memory,
the
the
term
temporal
cooling.
We
just
think
of
that
as
something
different
than
what
would
what
new
meant
to
people
thought
in
2011.
A
Otherwise,
the
main
really
the
only
change
in
the
algorithm
is
that
it
used
to
be
that
the
temporal
memory
which
used
to
be
called
the
temporal
cooler
would
try
to
do
multi-step
prediction.
It
would
try
to
learn
if
something
was
predicted
correctly.
It
would
try
to
get
where
it
could
predicted
one
step
earlier
and
then
one
step
earlier
such
that
you
could
look
at
like
the
predictions
of
the
temporal
memory
to
figure
out
what's
gonna
happen,
same
steps
for
now
or
five
steps
from
now.
A
They
used
to
do
that,
and
that
was
in
the
white
paper
that
it
did.
That
would,
and
that's
no
longer
true
so
really
just
like
that
difference
in
the
algorithm
and
just
like
the
naming
of
them
and
that
those
are
the
main
changes
we
tweaked,
how
boosting
works
and
the
spatial
cooler
since
we
did
that
are
like
roughly
a
year
and
a
half
ago.
You
a
did
some
good
work
on
that
for
the
while
working
on
this
facial
cooler
paper,
but
that's
really
not
a
that's,
not
like
a
fundamental
algorithm
change.
A
Sorry,
we
updated
Bamie
with
that
tweak
and
to
boosting
to
I
believe
yeah
we're
trying
to
keep
a
me
up-to-date,
but
yeah
I
would
say
it's.
It
is
up
today
to
my
knowledge.
For
the
most
part,
everything
coming
out
of
research
is
an
extension
to
HTM.
There's
nothing
crucial.
Our
core
is
changing,
I
mean
you'd,
say
everything
we
have
right
now
up
that's
implemented
up.
It's
a
new
pic
is
just
foundational
to
what
comes
next.
So
no
major
changes.
A
A
I
haven't
really
gone
over
all
these
different
HTM
systems
on
a
hackers
hangout
before,
but
there's
a
bunch
of
different
HTM
systems,
and
some
of
these
are
pretty
cool
like
like
this.
One
is
not
too
old,
I
mean
two
years
old
is
not
too
old.
Some
of
these
have
been
around
for
seven
years,
but
I
remember
when
this
came
out.
It
had
really
nice
documentation
so
I
liked
late
to
that,
and
some
of
these
are
black
implementations
of
before
we
even
went
open-source
like
I
know.
A
A
couple
of
these
old
C++
ones
were
five
years
old.
You
know
this
was
before
we
even
went
open-source.
Several
of
them
are
and
the
guys
that
were
working
on
these,
like
that
was
Michael
Faria.
He
eventually
came
over
and
started
talking
on
the
forum
and
working
on
other
things,
there's
a
few
interesting
JavaScript
ones,
including
HTM
J,
at
which
Paul
amrohi's
on
the
on
the
call
right
now-
and
these
are
cool
because
I've
got
a
ton
of
JavaScript
visualizations
that
I
used
in
HTM
school.
A
That
I
can
just
attach
to
this
pretty
easily
so
I'm
investigating
some
of
these,
especially
that
there's
another
JavaScript
one
that
somebody
just
posted
that
I
don't
have
linked
here,
but
and
there's
all
these
Java
based
ones,
here's
cortex
I,
don't
you
guys
haven't
used
this
for
a
while,
but
it's
a
very
cool
closure
version
of
library,
the
Felix
Andrews
and
recently
Furcal
was
working
on
it.
I
know:
Marcus
has
worked
on
this
before
you
know.
A
You're
kind
of
mixing
two
up
for
bulled,
mostly
worked
on
cortex
Felix
built
Kim
cortex,
which
is
further
down
the
list.
I
helped
felix
with
that
he's
the
main
person
who
really
did
comport
X
both
of
us
did
a
lot
work
together
on
comport
X
was
the
visualization
for
it,
though
he
originated
that
as
well.
Thank
you
for
correcting
me.
That's
absolutely
right.
This
is
the
closure
one,
the
other
one
was
furgus
cortex
and
I.
Don't
think
that
cortex
one
is
complete,
but
the
comport
x1
certainly
is
there's
a
ruler.
A
That's
where
a
bunch
of
cool
visualizations
came
from
and
sort
of
the
story
of
how
we
got
Marcus
involves
from
there
to
anyway.
Oh
the
reason,
I
brought
this
up
is
because
I'm
working
on
a
project
I'm
going
to
show
you
real,
quick.
It's
an
interactive
document
called
building
HTM
systems
and
I'm
going
to
go
over
from
from
scratch.
A
C
The
thing
I
have
to
say
is
that
my
project
is
still
coming
along
great
I
work
on
it
every
single
day,
but
it's
still
months
away
from,
but
you
guys
are
gonna
love
it.
It's
so
good,
it's
so
great,
but
it's
still
a
little.
It's
too
far,
I
want
it
I.
It
always
feels
like
it's
three
months
from
done,
and
it's
felt
like
that
for
two
years
now,
but
I'm
hopeful
that
someday
will
actually
be
done,
but
I
already
have
great
great
great
stuff.
C
A
About
you,
Gary
I
saw
your
really
interesting
program.
I
was
talking
to
the
guys
before
this
started
out.
I
just
I
think
that
having
a
pet
project
like
that
is
really
intriguing,
and
you
can
tell
you
put
a
ton
of
work
and
thought
into
it
and
and
that
the
visualizations
are
great
and
all
the
configurations
I
mean
that's
all
custom
stuff
I.
Obviously,
as
a
software
developer,
I
recognize
the
work
that
goes
into
that
that's
pretty
cool,
pretty
cool
stuff.