►
From YouTube: NuPIC Development Progress Review - Aug 7, 2015
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello
new
pic:
this
is
a
new
pic
development
progress
review,
August
seventh
2015,
my
name
is
Matt
Taylor.
You
look
at
my
screen.
That's
key
people
in
the
office.
Some
people
online
we're
going
to
bust
through
all
the
stuff
that
has
happened
over
the
past
month
because
we
didn't
have
a
meeting
two
weeks
ago.
So
hopefully
this
won't
be
too
much
longer,
but
there's
been
a
lot
of
things
merged.
A
So
let's
jump
right
in
so
first
thing:
I
want
to
talk
about,
I'm
probably
going
to
defer
to
Scott
about
this,
but
on
our
wall
board
you
can
see
where
new
pic
or
is
way
behind.
I'm.
Sorry,
new
pic
is
way
behind
new
pic
or
by
68
commits
they'll.
The
the
last
PR
that's
supposed
to
update
it
is
having
Travis
failures.
I
I
assume
you
guys
are
all
familiar
with.
What's
going
on
here
and
I.
Think
Aquila
has
been
the
one
that's
working
on.
A
B
Moved
some
over
new,
pic
or
and
as
a
PR
in
new
pic
that
will
bring
it
back
up
to
speed,
but
new
pic
core
is
suspended.
A
little
bit
of
flux
as
she's
been
trying
to
work
out
the
deployment
and
whatnot
so
she's
out
today,
but
I
think
we're
pretty
glow
skin
and
back
it
syncs.
If
I
Monday,
Tuesday
great.
A
A
You
can
see
in
our
build
times.
This
is
travis
for
new
pic,
and
this
is
new
pic
core
and
you
can
see
the
drastic
increase
in
the
bill
times
from
decor.
That
is
because
of
the
work
kill
has
been
doing
to
move
some
of
the
build
from
to
victory
of
the
core.
So
specifically,
issues
move
to
swig
compilation
from
new
pic
and
to
be
pakora
that
comes
along
with
a
bunch
of
tendencies
that
need
to
get
installed.
A
A
So
we've
got
that
cleared
up
and
let's
move
on
to
I
want
to
look
at
what
has
happened
over
the
past
month
with
the
pull
request
have
been
merged.
A
lot
of
these
are
Alex
and
Aquila
who
aren't
here,
but
we
can
at
least
talk
about.
What's
been
going
on
at
least
a
major
initiative,
so
one
of
which
alex
worked
on
was
the
Canon
classifier.
Accepting
multiple
input
categories,
which
really
was
a
bigger
deal
right,
could
gets,
and
you
mentioned
something
about
what
that
means.
It's
a
for
the
file
input
right.
B
By
the
Canon
classifier
part,
but
yeah,
so
currently
the
winds
currently
set
up
with
categories
in
both
the
encoders,
but
also
the
classifier
regions
and
whatnot.
It's
just
a
single
category,
and
so
for
some
of
the
work
that
out
soon
on
language
stuff.
They
want
to
be
able
to
do
multiple
classifications,
her
record
and
so
there's
a
new
field
type
in
the
in
the
file
record
stream,
which
is
a
list
type.
So
you
can
specify
multiple
integer
categories
as
poster
single
one.
A
Good,
it
boils
down
to
an
API
enhancement,
I
believe
and
that
we
can
we
can
do
more
with
categories
than
we
could
before.
So,
if
you
want
to
know
the
details
on
this,
there's
a
there's,
an
issue
2
335,
otherwise
we
should
move
on
all
my
pots
did
I'm
using
this
thing.
That
I
can.
I
can
write
on
web
pages,
but
every
time
I
do
anything
they
all
get
deleted.
So
there's
four
things
I
did.
I
updated
our
or
travis
builds
to
be
container
based
bills,
which
has
made
him
a
little
faster.
A
B
It's
just
our
projects
already
split
up
into
multiple
places,
with
the
new
pic
or
new
bag,
and
then
more
recently,
we
moved
the
vision,
stuff
out,
mosport
I'm,
not
very
supportive
of
separating
the
project
and
definitely
in
different
places,
but
with
different
at
made
sense,
because
there's
a
lot
of
heavy
dependencies
that
come
along
with
vision
that
for
people
not
doing
vision,
it's
a
bit
of
a
burden
and
so
anyways.
This
making
new
picking
names
first
package
allows
you
to
build
additional
functionality.
B
The
separate
project
then
saw
in
to
clean
API,
so
you
can
still
import
from
new
pic
vision
whatever
and
you
can,
and
similarly
with
the
by
no
means
now
that
we've
moved
the
build
entire
leave
it
to
new
pit
core
that
has
to
suck
up
that
pie
that
installs
new
pic
bindings
and
so
just
like
before
you'll
still
be
importing
from
the
pic
that
bindings.
But
that's
actress
ever
project
from
the
rest
of
you
pick
great.
B
A
B
Sounds
good
so
yeah
the
effort
is
to
make
hyper
sir,
just
an
old
library
right
now
it's
tightly
coupled
to
a
yoga
and
I
guess
there
was
an
issue
1753
that
I
think
you
opened
up,
which
was
just
a
straight
investigate
tasks
fractionating
that
thing
that
happened.
So
I
proposed
a
series
of
milestones
to
work
on
at
coming
weeks,
that
I
think
Scott
looked
over
and
approved
and
I
guess
Austin.
Also.
B
Just
recently,
I
saw
the
Gail
Smith
some
comments
to
elaborate
on
a
couple
things
but
largely
I
think
moving
forward
the
not
working
right
now
so
too.
We
really
try
and
separate
out
the
code
that
will
exist,
pull
up
basically
independently
Whip
It,
Up
research
and
also
to
kind
of
separate
stuff
that
will
exist
in
up
a
new
pic
specific
client
and
kind
of
stuff
read
the
code
that
exists
now
into
those
in
those
categories.
B
You
know
the
hypergate
of
existing
client
and
make
sure
that
things
of
the
entire
search
are
not
necessarily
hard,
not
depend
at
all
on
things
outside
of
packer
search.
Whereas
new
pity
the
client
can
actually
call
things
with
an
eye
research
and
also
within
do
pick,
but
we
want
to
try
and
separate
out
those
those
dependencies
and
then
afterwards
to
define
once
that's
done
to
actually
we're
going
to
find
a
more
appropriate
I,
/
search
api
that
and
eventually
my
reading,
our
design
into
to
fit
the
specification
to
that
again.
All.
A
A
B
C
So
this
was
sort
of
the
gist
of
the
comment
that
I've
made
on.
This
is
the
way
that
new
pic
uses
it
and
then
there's
a
library
itself.
The
library
itself
is
going
to
be
called
hyper
search
and
the
thing
that
it
does
effectively
is
swarm
and
so
I
think
it's
totally
okay
to
continue
using
swarm
to
describe
what
it's
doing
while
calling
the
library
itself
hyper
search.
Yeah.
A
Anything
else
major
in
here
I,
don't
I,
don't
believe
so
aside
from
things
that
are
being
done
to
move,
build
things
from
New
picked,
a
new
pic
or
I.
Think
that's
probably
it
so.
This
is.
These
are
new
pic
full
requests
over
the
past
month,
swap
switching
over
to
new
bacoor
pull
requests
over
the
past
month.
See
if
there's
any
big
items
in
here,
we
should
at
least
discuss
moved
the
swig
stuff
over
Richard
Crowder,
recently
updated
the
CMake
build
for
its
policy
behavior,
which
made
it
out
put
less
junk,
which
is
a
good
thing.
A
A
A
Excuse
me
other
than
that
registry
regions
yeah,
so
I
guess
a
yeah
Richard
saying
is
additional
work
of
the
port
from
Python,
so
so
something
that
had
to
be
done
in
this
Python
port
of
the
temporal
memory
algorithm.
So
that
which
brings
me
to
the
next
thing
to
the
TM
algorithm
progress
here.
Let
me
get
to
the
root
of
this
there's.
Still,
a
big
sort
of
super
issue
that
we're
tracking
Richard
is
is
plugging
through
these.
A
Some
of
those
are
kind
of
some
of
these
are
kind
of
Hong,
but
he's
he's
making
progress,
keeping
us
up-to-date
on
the
lists
and
these
issues.
So
a
lot
of
these
were
not
checked
off
the
last
time
we
had
this
meeting
so
at
this
point,
several
are
being
checked
off
and
others
are
under
implementation.
So
thanks
Richard
for
all
your
work
here,
yeah!
Ok!
A
So
that's
all
the
stuff
that
has
been
merged
over
the
past
month-
and
these
are
the
currently
open
PRS
starting
at
the
top,
the
oldest
I'm
not
going
to
go
into
some
of
these
them.
We
know
they're
just
in
a
bad
state,
we're
waiting
for
authors
to
add
to
make
changes
to
the
pr's.
We
now
do
have
an
automated
process
in
place
that
will
look
at
PRS
if
they
are
in
progress
and
have
not
been
touched
for
over
a
month,
they
will
be
automatically
closed.
A
So
a
lot
of
these
that
are
just
sitting
here,
just
waiting
for
that
to
happen
to
them.
Unless
there
is
some
movement
from
the
original
PR
authors
there'll
be
automatically
closed.
If
the
author
would
like
to
reopen
them,
do
some
more
work
on
them
and
reopen
them.
The
ones
that
are
marked
for
review
or
marked
as
ready
will
not
be
automatically
closed,
that
that
is
the
responsibility
of
the
nuclear
viewers
team
to
review
those.
A
So
that's
I've
gone
through
all
these
PRS
and
commented
on
those
that
warranted
comments,
but
otherwise
what
we've
got
here
is
a
bunch
of
serialization
stuff.
That's
either
a
tongue
or
blocked
on
other
serialization
stuff
that
needs
to
get
done,
got
a
PRS
that
Richard
is
working
on
for
temporal
memory
and
and
then
still
stuff.
A
keel
is
working
on
for
the
big
move
of
some
build
components
from
new
pic
to
nuclear
core.
That's
still
being
worked
out
and
then
a
few
sort
of
one
off
things.
A
A
C
C
A
A
A
C
It's
a
little
bit
if
you
were
to
take
the
code
as
it
is
and
translate
it
to
23.
The
Nakota
self
becomes
the
slow
mo
under
compatible
up
icon
27.
So
some
areas
you
need
to
take
advantage
of
six
so
that
it
automatically
will
use
the
Python
two
or
three
component,
depending
on
which
version
it's
running
in.
But
the
vast
majority
of
the
this
stuff
was
done
with
two
to
three
and
then
there's
a
handful
of
places
where
I
needed
to
go
in
and
use
six.
One
of
the
most
common
areas
is
in
string
I.
C
Oh,
so
we
have
a
lot
of
places.
We
use
the
IO
package
and
the
string
I/o
module
from
that
and
there's
some
Unicode
incompatibility
issues,
and
so
an
easy
way
to
fix.
It
is
to
use
the
string
io
that
comes
with
six
and
it
automatically
switches
back
and
forth
between
the
right
implementation
and
in
the
other
area
we're
using
six.
A
lot
is
wherever
we're
using
meta
classes.
It
makes
it
there's
a
whiff
metaclass
function
that
makes
it
easy
to
define
a
class
that
uses
it
in
this
canal,
boat,
203,.