►
From YouTube: NuPIC Development Progress Review - Sept 4, 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
Right
no
comments
from
the
oh
hello,
dybbuk
welcome
to
the
September.
Fourth
new
pic
development
progress
review:
I
am
Matt.
This
is
david.
This
is
the
rest
of
the
limited
engineering
team.
Let's
talk
about
stuff
that
got
done
so,
first
of
all,
I
guess.
I'll
share
my
screen
and
okay.
So
I
took
down
some
notes
from
a
meeting
we
had
earlier,
where
we
talked
about
progress
in
windows
and
serialization.
A
So
God
is
sort
of
leading
both
of
these
two
efforts,
so
I'm
just
going
to
run
through
what
we
talked
about
Scott.
If
you
have
any
comments,
you
are
very
welcome
to
jump
in
so
we're
working
on
Windows,
Richard
Crowder's,
helping
us
with
the
windows
stuff,
but
it's
a
very
complex
landscape
on
windows,
because
some
of
their
some
of
the
compilation
environments,
don't
support
C++
11,
there's
some
confusion
about
whether
we're
doing
two
or
three
and
what
the
conflicts
are.
I
honestly,
don't
understand
the
landscape,
but
it
seems
that's
gotten.
A
Austin
have
a
good
understanding
of
it,
I'm
going
to
let
them
move
forward,
but
our
current
plan
is
to
try
and
create
Python
binaries
that
will
be
installed
on
windows.
So
we
can
run
new
pic
on
top
of,
and
once
we
get
those
created,
we
will
make
sure
that
they're
available
to
the
community.
So
if
anyone
in
the
community
wants
to
run
them
on
on
windows,
they
can.
Is
that
a
fair
assessment
for
you
guys.
B
Listen
will
be
a
update
on
the
windows
stuff
from
Richard
in
next
several
days.
Great
love
that.
A
Great
so
also
serialization,
so
we've
had
a
bunch
of
bill
problems
over
the
past
few
weeks.
I
can't
blame
them
all
on
cat
pee,
but
kathy
has
always
been
the
whipping
boy.
Every
time
we
have
a
build
failure,
a
lot
of
is
because
it's
it's
a
complicated
build.
So
we've
Scott
and
I
discussed
this
and
we
decided
we'll
we'll
disable
it
for
a
while.
A
I've
tested,
install
and
others
have
to
I
haven't
got
any
reports
of
the
failing,
so
everything
was
good
and
the
plan
is
to
reenable
the
captain
proto
and
the
Duca
poor
bill
are
likely
going
to
bundle
that
source
with
it
and
which
will
produce
the
new
pic
findings
that
have
Captain
p,
bundled
so
developers
o
need
to
worry
about
installing
it
themselves.
They'll
just
get
a
binary
that
that
will
have
it
if
they're
working
on
new
pic
and
as
well,
users
will
not
have
to
worry
about
installing
comfy.
So
that
makes
me
happy.
A
A
Okay,
so
I
have
some
issues.
I
want
to
make
sure
a
highlight,
but
I
think
if
I'd
go
through,
the
road
map
will
get
all
these,
and
so,
let's
hit
some
of
these
super
issues
here.
This
was
a
ticket
that
I
was
putting
together
all
of
some
of
the
problems
that
I
was
seeing
with
builds
and
test
failures
and
stuff
like
that
that
we've
been
fixing
over
the
past
week
or
two.
So
a
bunch
of
these
have
been
fixed
up.
There's
still,
a
few
that
are
important.
Like
Scott
has
mentioned
this
a
few
times.
A
B
A
Okay,
there's
a
few
other
issues:
I
want
to
clean
up
before
I
close
this
out,
I
want
to
remove
our
dependency
on
our
support
repositories,
and
then
we've
got
a
few
little
things
that
just
need
to
be
cleaned
up
in
our
builds
like
we're
still
releasing
a
wheels
to
s3
for
no
reason
after
we've
moved
all
of
our
C
over
to
do
before
we
I
don't
think
we
need
to
be
doing
that
on
our
Lennox
fields,
pretty
big.
So
those
are
cited
me.
A
Moving
onward,
I
did
add
tickets
to
add
an
anaconda
support.
This
came
out
of
discussion.
We
had
with
somebody
trying
to
use
our
system
and
just
kept
run
into
problems.
We
really
should
support
an
econo.
We
run
anaconda
on
our
Santos
builds.
We
just
don't
have
any
decent
documentation
and
we
don't
have
any
CI.
The
builds
in
anaconda,
so
I
think
this
is
something
we
might
as
well.
Try
an
act
so
I'm
going
to
work
on
that
I,
don't
know
exactly
when,
but
I'd
love
to
get
them
at
it.
A
Because
there's
been
a
lot
of
interesting,
I
found
the
support
and
we've
never
really
helped
anyone
with
it.
What
is
what
is
the
port
grand
economy?
Yeah?
It's
a
good
question.
Let's
see,
we've
talked
about
this.
I
think
I
created
a
first
issue,
a
dreadful
discussion,
but
in
my
opinion
it's
not
Condit
install
new
pic.
It's
just
that
you
can
compile
and
build
new
pic
on
an
anaconda
system
or
do
pip
install
new
pic
on
organic
on
a
system
either
there's
instructions
on
making
it
work
for
it
just
works.
A
A
Know
this
might
be
really
simple
to
do.
The
answer
may
be.
We
do
have
support
for
it.
It's
just
we
haven't
highlighted
it
and
I
haven't
it's
not
in
a
place.
That's
visible
enough
for
me
to
point
people
to
the
answers
and
if
that's
the
case,
then
great
it'll
be
really
easy,
but
honestly
I
tried
running
and
building
an
anaconda
I,
don't
know
it
was
a
year
ago
or
more
and
failed
miserably.
So
at
that
point,
I
thought
well
I'm
just
going
to
give
up
move
on
so
an
edge
case,
but
it's.
A
A
A
So
let's
talk
real
quickly
about
this
proposal
for
stricter
api's
through
interfaces
and
marks
different
issues
for
creating
these
different
interfaces.
I'm
fine
with
all
of
this
and
I
think
discussions
are
ongoing
for
each
one
of
these,
but
I
want
anybody
and
the
audience
to
have
an
opportunity
to
comment
on
this
if
they
like
anyone.
A
One
one
ticket
here
I
wanted
to
note
is
this:
I
could
close
this
out,
probably
if
we
make
a
decision
on
this
vision
stuff.
Oh,
this
has
been
hanging
around
for
a
long
time
and
I
think
there
was
a
PR.
Well,
there
was
a
PR
that
was
closed
because
I
think
it
removed
too
much,
but
I
don't
know
what
to
do
about
this.
We
should
we
should
we
move
vision,
code
out
or
or
or
not,
it's
been
around
for
too
long.
A
B
Question
there
was
the
pc,
a
node,
whether
that
name
or
anything
vision
but
specific,
some
visions.
Okay,
so
then
that's
what
was
decided
originally
and
then
there
was
a
PR
where
they
were
moving
trying
to
move.
The
test
will
remove
the
desk
or
easy
a
note
from
new
day,
because
it
depended
on
the
image
sensor
and
so
I
said
we
should
added
in
moving
the
task
and
easy
netted
two
new
pic
vision
or
we
should
be
seen.
You
know,
Lamont's
new
pic.
Then
we
need
to
change
the
test,
so
it
doesn't
rely
on
the
intercessory.
A
So
I
found
I
found
the
new
pic
issue.
Wait,
move
vision
coat,
so
there's
two
issues
here:
the
10
effort
ones,
new
pic
quorum,
one's
different.
Sorry,
sorry
for
the
confusion,
but
I
think
it's
basically
the
same
thing
and
the
the
discussion
about
the
pc
may
know
thing
is
happening
on
this
new
fickes
shoe
thing:
was
it.
C
A
Bring
through
a
lot
of
the
stuff,
that's
done
so
that's
that's
good
I
think
I
mean
one
one
issue
that
you
probably
know:
tis
we're
not
really
following
this
milestone
thing
really
quickly
been
I
just
bumped
this
version,
since
we
did
it
did
a
release
recently
and
we
don't
really
pay
much
attention
to
this
I've
been
thinking
a
lot
about
what
views
we
need
to
look
at
at
this
meeting
and
I.
Don't
know
this
is
the
right
one.
C
It
gets
back
to
this
issue
of
Vita.
What
is
exactly
1
point
0,
and
is
it
well
defined
and
number?
Are
we
progressing
towards
it
and
it
sounds
like
there's
also
lots
of
stuff
that
comes
out.
That's
not
directly
related
to
one
point:
oh
so
the
exact
numbers
there
might
not
make
sense
percentage
complete.
C
C
C
C
C
C
C
C
D
C
D
D
D
D
Right,
well,
you
know
I,
don't
want
to
I,
don't
want.
You
know,
because
I
have
all
my
eggs
in
this
one
basket.
So
I
depend
on
the
outcome.
I
don't
want
to
walk
down
the
street.
You
know,
I
live
in
a
bad
neighborhood.
I,
don't
want
to
live,
walk
up
feat
say:
hey
was
it
you
that
made
that
jive
Turkey
AI
implementation,
I.
A
A
C
I'm
sorry,
but
that's
something
interesting
which
is
you
know
when
should
no
big
one?
Should
algorithms
move
over
from
the
Big
research
student
open
and
is
that
tied
to
one
point
0
at
all?
That's
what
David's
question
was
and
I
said
yeah
one
point:
it
was
more
about
the
code,
cleanliness
and
things
like
that
President
about
algorithms,
about
Vivian,
rather
than
having
hierarchy
or
feedback
or
something
like
that
it,
but
you
know
you're
doing
a
lot
of
research
into
big
research
which
anyone
can
look
at.
C
But
at
what
point
should
we
move
algorithms
over
to
new
packet?
David
was
saying:
well,
we
should
make
sure
is
working
well
on
some
real
world
problems
before
we
were
written
it
there,
that's
a
fairly
high
standard,
which
is
yeah
yeah.
That's
what.
D
D
You
know
we're
just
going
to
start
looking
for
some
benchmarks
and
some
and
some
real-world
problems
to
be
solved,
and
you
know
that
you're
going
to
want
to
compare
it
more
heavily
to
you
know
to
the
classic
ml
stuff,
and
so
you
know
I
think
we
we
have
I,
don't
want
to
make
a
false
start.
When
it
comes
out,
you
know
it
should
be
able
to.
You
know
to
perform
to
some
levels
so
that
we
can
say
you
know
you
know
HTM,
you
know
it
should
be
it
should.
D
A
A
So
there
is
an
issue
for
this.
That's
been
open,
first
of
six,
eight
months
or
so,
but
I'm
concerned
about
this
and
I
want
to
make
sure
everybody
else
is
concerned
about
this
and
it
doesn't
fall
off
of
our
radar.
That
I
don't
think
we
currently
work
with
4.9,
but
I
haven't
tried
in
a
while.
So
maybe
we
do
at
this
point.
Maybe
not,
but
there
is
an
issue
about
it.
It
is
a
fee
to
I
just
want
to
bring
it
to
everybody's
attention,
and
let
you
know
that
I'm
concerned
about
it
bad,
that's!
A
A
D
D
Somehow
there
are
some.
Some
problems
crept
into:
Java
8
and
the
coverage
tool.
Does
some
instrumentation
of
you
know
the
classes
in
order
to
produce
the
coverage
statistics
and
it's
the
coverage,
output
and,
and
so
I
got
in
contact
with
the
one
of
the
main
developers
from
jacoco.
It's
evgeniy
meant
man
jerk
off
and
he
you
know,
made
a
PR.
He
actually
helped
fix
the
build,
but
he
exposed
a
an
actual
bug
in
the
JDK.
D
So
we
have
a
you
know,
bug
up
on
Jake
on
the
locked
against
OpenJDK
on
grr
and
for
some
reason
it
came
into
effect
when
it
was
trying
to
instrument
some
of
the
support
libraries
I
had,
namely
the
trove
library
which
allows
you
to
do
you
know
native
data
structures
instead
of
objects.
So
now
I'm
doing
some
stuff
in
there
to
try
to
you
know,
give
it
some
performance,
enhanced
performance
a
little
bit
so
allowed
to
use
a
collections
that
use
native
or
primitive
data
structures.
D
So
anyway,
it
was
trying
to
instrument
that
so
we
kind
of
commented
that
part
out
of
the
bill
told
you
not
to
build
and
suddenly
now
it's
working
so
so
that
helped
out
a
lot.
We
we
are
ported
over
the
new
temporal.
Well,
the
changes
that
were
made
to
the
temporal
memories
is
the
last
time.
I
took
a
look
at
it,
which
was
left
last
August,
so
we
brought
it
up
to
date
with
the
current
version
of
new
pic
and
then
we
poured
it
over.
D
D
So
we
so
anyway,
we
have
this
monitor
mixing
framework,
so
we
can
basically
duplicate
the
integration
tests
that
are
a
new
pic
and-
and
we
did
that
and
by
doing
that,
we
actually
exposed
some
problems
with
the
new
pic
integration
test
was
just
over
slight
oversight
but
mom.
So
well
doing
this
port
kind
of
helps
out.
You
know
new
pic
to
sometimes
just
a
good
thing:
I
guess
what
else
do
we
do
so
so
the
temporal
memory,
the
monitor,
mixing
framework,
let's
see
trying
to
think
oh.
D
We
added
the
ability
to
input
binary
files
into
the
network,
API
so
actually
they're,
just
the
string
of
five
vectors,
so
the
bunch
of
ones
and
zeros,
so
you
can
have
you,
can
input
ones
and
zeros
into
you
know.
So
if
you've
done
any
like
pre
encoding
or
you've
done
your
own
encoding,
you
can
run
that
directly
into
the
network.
Now,
so
we
extended
the
functionality
of
the
sensors
to
handle
that
and
some
of
the
transformation
stuff
that's
going
on
inside
of
the
network
to
be
able
to
integrate
it
in
arm.
D
So
so
now
you
can
just
point
it
to
a
CSV
file
that
contains
ones
and
zeros
and
and
it'll
just
process
that
and
pull
it
on
through
the
network
and
process
it.
So
that
was
actually
a
community
member
who's
working
on
a
pet
project.
Josh
is
working
on
some
things.
He's
trying
to
do
for
the
indus
is
implementation
of
M
de
stand.
So
you
know
he
needed
that,
and
so
we
added
that
that
that's
about
it
I
think
they're
able
to
the
major
highlights
thanks.