►
From YouTube: NuPIC Sprint Planning - Sept 27, 2013
Description
NuPIC Sprint Planning - Sept 27, 2013
A
A
Alright
well
start
with
the
roadmap,
which
is
still
focused
primarily
on
natural
language,
processing
of
the
new
of
the
hackathon
coming
up
in
September,
as
well
as
managing
external
dependencies
better.
Before
we
go
into
looking
at
our
current
sprint,
I
would,
like
I,
talked
to
Scott
nahson
about
this
earlier
so
I'd
like
you
guys
to
just
comment
on
how
we're
going
to
move
forward
tackling
our
dependencies,
hello,
you
Scott.
B
Yeah,
so
the
plan
is
still
the
same.
It's
just.
We
haven't
a
lot
of
time
to
work
on
the
best.
Basically
right
now
we
have
external
reconquer
that
I
problems
specific
that
have
the
dependencies
installed
in
my
body
and
those
aren't
quite
done,
but
they
help
the
Python
amenities
right
now
and
we
also
remove
the
mom
Python
dependency
side,
scrape
those
as
well
and
then
once
we
that
will
feature
of
our
continuous
integration,
two
views,
those
and
then
once
we've
done,
that
we
can
remove
all
the
dependencies
from
new
pic
and
then
that's.
A
Fun
so
the
end
result
will
be
Inuvik
repo
without
bundle
dependencies
that
someone
can
either
install
manually
or
use
our
kind
of
satellite
repository
that
contains
the
dependencies
if
they
have
that
platform.
Okay,
Reese,
prep,
okay,
so
that's
the
pen!
We
don't
know
exactly
when
that
plan
is
going
to
be
in
place,
but.
A
We're
working
on
it,
okay,
so
other
than
that
I
think
we're
making
good
good
progress
on
in
okies
stuff.
If
you
haven't
watched
this,
we
had
a
planning
video
with
Fergal
in
Francisco
and
which
was
we
have
shot
a
lot
of
stuff.
I
have
merged
in
some
work.
Matt
Keith
did
on
a
pastor
encoder,
which
I'm
going
to
try
to
use
next
week,
two
feet
in
st
ours
from
cept,
so
that
will
be
a
good
experiment.
A
So
that's
that's
our
focus
as
more
NLP
stuff
and
trying
to
knock
out
the
way
we
handle
dependencies.
So,
let's
take
a
look
at
first
off
are
open
for
classes
as
well
as
to
spend,
since
we
open
sourced
I
think
so
that's
great
there's
one
here
that
is
still
running
as
soon
as
this
one
passes.
I
will
be
able
to
merge
that
too,
that's
great.
A
A
C
A
B
A
Move
over
auto
see
what
to
do
and
then
two
infractions,
I
can't
I,
don't
have
permission
to
move
one
of
your
tasks
in
progress,
somehow
really
messed
up.
The
tier
configuration
Buffalo
you'll
have
to
move
it
so
once
you
do
have
the
different
environment
setup
with
dependencies
and
would
like
to
have
a
doc
somewhere
just
to
give
people
some
simple
instructions
how
to
use
it,
how
to
do
if
they
want
to
use
it,
and
then
this
build
script
that
assumes
dependencies
already
installed
is
that's
still
relevant,
yeah.
A
A
So
yeah
mark
and
David
of
any
going
back
and
forth
on
this
without
much
input
from
us,
so
it
might
be
a
good
idea
for
one
of
us
to
is
anyone
you
see
make
none
of
us,
you
see
make
so
I
would
say
as
long
as
this
doesn't
break
anything
for
us
and
they're
happy
with
it,
we
can
merge
it.
So
once
you
guys
come
to
a
consensus
that
you're
happy
with
give
us
the
thumbs
up
and
and
once
the
build
passes
will
merge
it.
A
All
right,
okay,
so
keep
that
as
well
he's
got.
It
is
blocked,
I
key
to
his
block
design
with
statuses
Gil.
His
internship
is
over.
He
got
all
of
his
c++
spatial
cooler
tasks
done
so
Josh
Rose
I
met
Josh
up
NASA,
I
believe,
and
he
was
working
on
figuring
out
the
best
way
to
feed
word
asteras
into
the
CLA.
I
reported
this
and
I
think
he
grabbed
it.
So
I
asked
anybody
who's
interested.
A
So
this
is
something
I
want
to
be
working
on
too,
because
I've
got
a
bunch
of
word
STRs
and
expand
in
mind
that
I
wanted
to
try
so
Josh.
If
you're
watching
I'd
love
to
collaborate
with
you,
so
the
encoder
has
been
merged,
so
we
should
be
able
to
try
it
out,
although
there's
no
sparsity
normalization,
yet
I
still
want
to
try
it
who
knows.
A
Okay
mark
also
is
a
ticket
on
stealing
report
or
si
Lang
support.
This
might
be
the
same
issue
mark
if
you're
watching
this.
Could
you
check
and
see?
If
is
the
same
issue
that
David
is
working
on
and
if
so
just
remove,
one
of
them
I'll
try
and
follow
up
with
you
on
that.
The
matplotlib
dependency
is
this:
this
PR
that's
been
around
for
a
long
time.
Once
we
get
our
dependency
situation
figured
out.
A
A
C
It's
finally
so
you're
going
to
feed
it
to
the
temple
pool
he's
going
to
predict,
have
multiple
prediction:
that
every
concept
that
those
are
on
the
org
together.
So
you
need
a
way
to
separate
those
out.
That
means
to
the
original
right.
So
I
can
talk
to
you
later
about
some
strategies
for
that
I
didn't.
A
C
A
A
A
A
A
A
B
C
C
A
A
A
A
Yes,
back
to
me
so
I
did
I
did
add
some
some
wiki
docs
to
this
NLP.
Oh
yeah,
so
I
flush
this
out
a
little
bit
based
on
the
work
that
I've
done,
but
I
and
I
was
going
to
send
this
out
to
the
community
to
ask
to
help
me
flush
us
out
a
bit
more
because
I,
because
a
and
I
have
started
keeping
history
of
things
that
have
been
done
within
the
community
to
to
working
towards
this
idea.
He
may
in
defining.
C
A
I
have
them
listed
so
in
the
in
the
pages
right
here.
I've
got
the
whole
meet
minutes
of
the
meeting,
there's
to
do
tasks,
so
they
are
in
the
wiki
and
I
think
I
made
tickets
for
each
task.
Okay,
wait
so
yeah,
so
here's
all
the
tickets,
most
of
them
are
hope
and
a
couple
in
progress,
so
they
should
all
be
represented.
A
I
had
an
expected
issue,
the
tooling
server
that
I
think
is
fixed
if
anyone
notices
issues
with
Travis
statuses
or
polar
bi
statuses.
That
don't
seem
right.
Please
tell
me:
let's
see,
I'm
this
providing
it'll
be
baseline,
I'm
still
kind
of
working
on,
because
I'm
working
on
a
project,
and
I
want
to
and
I'm
trying
to
make
sure
that
I
can
complete
the
project.
A
So
that's
kind
of
my
my
idea
of
a
baseline,
a
very
simple
project
using
word
s,
tears
from
stuff
so
and
I
think
we've
already
identified
some
things
that
still
need
to
be
done,
especially
to
extract
the
st
our
predictions
back
out,
I
mean
by
a
based
on.
Would
you
like
a
very
initial,
very
basic
result?
Yeah,
oh,
that
you
know
what
I
might
be
thinking.
This
was
a
different
task.
A
A
C
A
A
A
So
I
just
want
to
add
that
to
the
wiki
somewhere
to
saviors
of
ideas,
about
anomaly
detection,
and
then
Francisco
is
still
he
emailed
me
yesterday
and
asked
what
else
could
he
do
and
so
I'm
asking
him
to
get
a
simple
corpus
of
sort
of
the
most
common
English
words
that
are
used
just
so
we
can
have
that
offline.
So
nobody
has
hit
the
cept
API
that
can
we
can
just
post
it
on
our
website,
so
they
can
download
it
and
use
it.
I
mean.
C
A
Else,
yeah,
the
baseline
stuff.
Anybody
could
do
think
this
is
something
that
anybody
anybody
with
the
scent
API
key
could
do.
But
francisco,
I
think,
is
a
good
place
to
gather
that
the
most
common
english
words-
I
don't
know
what
tools
is
using
for
that,
but
I
could
probably
do
it
with
with
the
natural
language
toolkit
and
python,
but
I'm,
not
nuts,
not
really
familiar
with
it.
So
I'm
just
asking
him
to
do
it
seems
it
seems
like
it
wouldn't
take
too
much
effort
form:
okay,
Scott.
A
B
Plan
for
that
was
sit,
implement
a
really
small
example
in
a
bunch
of
these
different
things,
but
it
just
kind
of
time
consuming.
So
at
this
point,
there's
a
handful
time
looking
at
the
one
I
like
the
most
is
a
captain
proto.
It
has
some
cool
properties
like
you
can
memory
map
it
disk,
so
it
would
be
like
really
really
efficient
for
loading,
something
running
a
single
loading.
It
model
running
a
single
record
through
it
and
saved
it
back
disk,
but
it's
also
the
youngest
evolved,
the
projects
that
I've
been
looking
at.
B
So
it's
still
in
active
development
and
I'm
having
trouble
compiling
it
online
now
and
then
I,
I
I'm
I'm
leaning
towards
something
that,
like
protocol
buffers,
are
calibrated.
That
has
a
specification
rather
than
something
like
message
back.
That
just
takes
an
object
in
so
that's
it,
because
if
we're
going
to
be
using
these
across
languages,
we
don't
want
to
have
to
maintain
the
backwards
compatibility
stuff
in
each
of
the
different
languages.
B
A
B
A
A
A
So
this
is
what
all
of
these
have
points.
What
is
unassigned.
First
of
all,
you
know
p
baseline
and
I'm,
going
to
pull
this
to
the
top
of
the
backlog,
since
no
one
is
assigned,
and
I'm
pretty
sure,
I'm
not
going
to
get
to
it.
This
sprint
lungs.
Other
things
I
have
to
do
if
anybody
wants
something
interesting
to
do
with
NLP
use
would
be
a
great
one
to
pick
up.
There
are
some
ideas
in
the
comments
and.
A
It's
something
that
you
could
try
and
implement
with
other
technologies,
other
machine
learning,
algorithms,
other
NLP
techniques,
just
to
provide
something
simple,
especially
if
you
use
the
current
corpus
of
text
that
we're
working
with
that
we've
been
talking
about
in
our
mailing
list.
That
would
be
great
and
something
that
we
can
try
and
reproduce,
and
a
new
big
experiment
to
compare
to.
A
A
A
B
A
A
Superset,
you
don't
have
anything
which
is
probably
fine
for
the
spread.
We,
the
vision,
stuff,
is
not
necessarily
important
at
the
moment.
I'm
really
more
focused
on
NLP
for
hackathon
will
eventually
get
around
to
that
and
don't
think
we
have
any
assigned
tickets
and
awesome
to
think
you're
fully
engaged
with
internal
stuff.
So
we'll
probably
I'll,
probably
bug
you
occasionally
about
things.
A
A
A
So
so
the
main
thing
I'm
I'm
going
to
be
working
on,
as
I
still
have
this
NOP
experiment,
trying
to
figure
out
how
to
get
SDRs
out
of
the
CLA
starting
that
discussion,
probably
not
working
on,
but
at
least
starting
some
discussion
on
the
mailing
list.
Audit
and,
as
anybody
else
have
any
pictures
they
would
like
to
bring
up
before.
We
convene
okay,
great
thanks
for
participating
thanks
for
watching.