►
From YouTube: HTM Hackers' Hangout - May 4, 2018
Description
https://discourse.numenta.org/t/htm-hackers-hangout-may-4-2018/3863
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
A
So,
if
anybody's
watching
right
now
on
YouTube
type
questions
in
the
chat
and
at
the
end
of
this
address
anything
that
pops
up
in
the
chat
and
if
you
want
to
join
and
discuss,
you
should
know
how
to
do
it.
If
you're
somebody
who
has
the
once
wants
to
join
all
right,
ok,
so
the
first
thing
I'm
gonna
talk
about
is
this
share
my
screen,
which
screen
station
window
there
we
go.
A
So
this
is
go
back
to
the
very
top
from
Young
miry
who's
sort
of
a
recent
member
of
the
forum
post.
It's
a
big
question
that
Jeff
answered
on
a
month
or
two
ago,
and
it
was
a
this
type
of
about
v1
and
transformations
and
anyway,
he's
been
involved
in
a
lot
of
stuff,
so
I
think
you're
watching
right
now,
Hyde
young
you're
watching
great
stuff,
great
discussions,
so
I
just
wanted
to
go
over
this
because
it.
A
What
it
looks
like
to
me
is
that
originally
this
is
about
trying
to
model
a
real
v1
and
try
and
get
like
the
right
mapping
of
retinal
input
in
a
more
realistic
way
into
that
layer
and
doing
anything
that
might
be
v1
specific
or
so
I
just
wanted
to
encourage
that
go
for
it.
These
that's
what
this
discussion
sort
of
looks
like
you
know
here
discussing
how
different
things
should
be
represented.
You
know
we're
sort
of
a
you're
on
the
fringe
of
encoding.
A
C
A
A
Look
good,
and
hopefully
it's
helpful-
you
guys
are
helping
people
understand
this
stuff
I'm
also
interested
in
this
stuff
down
below
here,
especially
about
this
mapping
from
LG,
and
if
you
want
I,
think
that's
really
interesting.
This
holds
zebra
pattern.
I
mean
there's
so
many
problems
to
solve
here
anyway,
I'm
just
trying
to
summarize
and
get
a
handle
on
what
you
guys
are
interested
in
I
think
this
is
also
really
interesting
too.
You
know
the
phobia
stuff.
A
A
But
you
guys
that
are
doing
it.
These
experiments,
these
visualizations
I'd
love
to
hear
I'd
love
to
like
get
in
touch
with
you
guys.
Maybe
did
another
hack,
another
hackers,
hangout
and
just
discuss
this,
because
I
think
I
could
probably
crack
this
egg
a
little
easier.
If
we
just
had
a
discussion
about
it,
that's
just
me:
I'm
gonna
stop
presenting
now.
Oh.
A
A
A
B
A
A
However,
all
this
comes
across
over
the
video
but
I'm
wanting
to
try
and
make
all
these
visualizations
as
interactive
as
possible,
so
that
people
reading
the
text
can
identify
concepts
in
the
text
directly
to
visualizations
and
I'm,
also
working
on
some
grid
cell
stuff
too
again,
I,
don't
know
how
great
this
is.
Gonna
come
across
because
it's,
but
this
is
I,
took
some
code
from
Mirko
glucose
who's.
In
he's,
a
research
scientist
has
worked
been
working
with
us
and
he
wrote
this
like
little
simulation.
A
This
was
like
this
is
really
a
simulation
of
grid
cells,
and
so
we've
got
like
three
grid
cells
here
and
we
just
turned
the
red
one
on
and
this
thing
is
randomly
walking
across
it.
If
I
move
my
mouse
in
here,
I
can
I
can
I
can
take
control
of
it
and
you
can
see
obviously
where
the
the
firing
fields
are
and
then
you
can
turn
on
the
other
cells
and
see
their
firing
fields
so
I'm
working
on
a
blog
post,
it's
gonna
have
this
type
of
visualization
there's
also.
B
A
Other
I'm
trying
to
show
how
one
dimensional
projections
might
sort
of
look
like
or
it
might
look
like
this
is
a
one
dimensional
projection
of
a
grid
cell,
one
grid
cell.
It's
projection
to
one
dimensional
space
and
these
are
sort
of
represent
its
firing
fields
as
potential
to
fire
around
the
each
location
sort
different
view
on
it.
So
those
are
the
sorts
of
things
that
I'm
working
on
almost
entirely
right.
A
Now,
aside
from
the
video
like
I
interviewed,
a
neuroscientists
I
haven't
even
edited
it
yet
I
have
another
interview
with
David
Schneider
who's,
a
neuroscientist
at
the
spider
laboratory,
but
that'll
be
next
week.
I'm
going
to
interview
him
so
I'll
have
two
interviews
to
edit
and
then
other
than
that
I'm
working
on
this,
these
visualizations
for
more
educational
content
and
that's
what
Matt's
doing
that's
all
I
got
for
today,
so
I'm
gonna
open
up
the
floor
for
you
guys,
Paul
and
Glenn.
If
you
have
anything
you
can
fight
over
the
mic,
yeah.
B
A
B
Right
cool,
so
I
did
this
in
one
dimensional
space,
because
it's
you
know
easier
to
depict
on
a
page
here,
but
you
know
you
can
imagine
that
this
would
be
sort
of
you
know,
because
could
this
line
here
could
be
a
you
know,
a
hex
on
or
hexagon
or
something?
And
then
you
know
you
project
that
onto
a
sheet
of
many
columns,
but
the
basic
idea
here
is
that
nor
in
the
traditional
SP
algorithm
you're
you
hold
onto
your
potential
pool
and
your
permanence
is
from
a
mini
column
perspective.
B
So
each
mini
column
has
a
potential
pool
and
permanence
is
with
it
with
its
synapses.
But
the
other
way
you
could
think
of
this
is
instead
you
have
an
ex
axon
from
one
of
the
input
cells
and
it
holds
on
to
a
potential
pool
and
permanence
is
instead
of
the
mini
column.
So,
instead
of
sort
of
seeing
this
as
a
receptive
field,
is
kind
of
the
offices
like
a
transmission
field,
but
the
benefit
you
have
here
is
that
you
can.
B
B
If
that
you
know
if
the
if
the
synapse
was
connected
right
and
then
so
you
could
do
that
for
each
each
of
the
on
input
cell.
So
that
gives
you
a
couple
of
optimizations
one
is,
instead
of
every
mini
column,
sampling
from
the
input.
You
only
have
to
take
the
active
input
cells
and
then
they
basically
are
transmitting
the
score
to
the
many
columns.
That's
a
that's
one
of
the
optimizations
we
had
talked
about
in
the
past
with
that.
B
A
B
Yeah,
so
the
so
what
I
thought
yeah
so
there's
kind
of
a
couple
things
you
could
do
so
this
is
going
to
how
I
would
implement
this
in
temporal
memory,
which
is
instead
of
having
segments
you
sort
of
have
a
receptive
field
would
still
be
a
segment,
but
it'd
be
the
same
way,
would
have
a
receptive
field,
and
so
you'd
have
your.
Whatever
cells
are
transmitting
would
have
a
transmission
field.
B
The
receivers
would
have
a
number
of
receptive
fields,
and
then
you
need
with
the
thing
that
that's
that
I'm
still
figuring
as
you
some
sort
of
a
global
space
where
you
can
mirror
match
up
the
the
receptive
fields
with
the
transmission
fields.
And
so
you
know,
if
you
say,
think
of
a
cell
on
TM
that
has
three
segments.
There
would
be
three
receptive
fields
for
that,
so
yeah.
B
A
B
A
Interesting
well,
cool
thanks
thanks
for
showing
that
Paul,
that's
a
good
idea!
I
mean
there's.
So
many
ideas
like
this
that
people
could
try.
I
mean
that's
one
of
the
things
that
attracted
me
to
HTM
from
the
beginning
was
it
was
such
a
flexible
theory
and
so
many
things
you
might
do
with
it
and
as
we
learn
more
about
the
brain
and
how
things
work,
I
don't
see
a
I,
don't
foresee
anything
coming
up
through
the
experimental
neuroscience
research.
A
That's
going
to
invalidate
anything
that
you
found
so
far
about
each
Jim,
which
is
great
so
anyway,
there's
been
some
chat
activity,
so
a
bit
King
or
Mark
Brown
has
been
chatting
and
he's
he
just
mentioned
in
chat.
Cause
I
know
that
some
of
you
guys
aren't
going
to
read
this,
but.
A
Working
with
young
on
that
with
some
of
those
ideas
in
the
B
one
thread
and
he's
impressed
and
he's
trying
to
get
the
details
right
and
he's
hoping,
it
will
be
a
workbench
to
explore
how
HTM
matches
up
with
some
of
the
more
classical
papers,
and
hopefully
John
Paul.
Hopefully
he
explained
the
optimization
but
we're
essentially
talking
about
not
storing
all
those
permanence
azat
in
the
SPE
and
the
many
columns
attached
for
the
many
columns
but
those
permanence
values
to
the
input
somehow
to
the
acts.
Songs.
The
input
is
it's
another
way.
A
Let's
see,
there's
there's
a
little
bit
more
chat.
Go
look
down
in
the
YouTube
chance.
I
think
that's
retained.
If
you
want
to
see
some
of
the
details
of
mark
brown's
comments
about
the
v1
stuff,
we're
talking
about
STRs
are
amazing,
yes,
and
that's
it
Glenn.
Do
you
have
anything
we're
talking
about
I'm.
C
A
8020
rule
anything
else,
affirm
you
Paul,
nope
I,
don't
have
anything.
Okay,
everybody
thanks
for
joining
in
the
hack
out
and
thanks
for
watching,
live
taking
part
in
it's
a
chat
and
the
HTM
community
don't
forget
to
to
join
our
forums.
If
you
haven't
already
at
discourse,
Numenta
org,
that's
where
all
of
the
communication
happens
for
the
most
part,
take
care
have
a
wonderful
weekend.