►
Description
HTM Community Meetup - March 14, 2016
A
B
A
So
this
is
something
I'm
working
on
something
that
just
been
working
on
he's
been
working
on
a
different
ways,
so
you
know
I'm
going
to
talk
about
something
I
created
plus
HTM,
a
missionary
example
of
it,
but
I'm
going
to
talk
to
some
philosophy
as
well.
So
Jeff
audience
is
I
said
several
times
that
you
don't
have
five
senses.
You
have
millions
of
senses
because
from
the
brains
standpoint,
it's
just
millions
of
neural
things
and
it
doesn't
differentiate.
This
barely
between
sight,
sound
and
somatic
senses,
Sutter
so
the
world.
A
In
the
same
regard,
my
idiot
is
just
an
enormous
stream
of
temporal
did.
That
means
it's
a
busy,
and
if
we
can
create
intelligent
systems
that
operate
aesthetic
principles
of
the
brain,
we
can
potentially
understand
the
world's
data
streams
in
the
same
way
that
our
brains
understand
our
sensory
input
streams.
So
that's
sort
of
the
big
idea.
What
I
want
to
do?
What
I've
been
starting
to
do
for
a
while
is
find
ways
to
attach
live
fish
game
systems
to
real-time
data
streams,
sounds
familiar
right.
A
Well,
one
of
the
things
I
need
to
do.
That's
pretty!
Obviously
it's
the
point
where
you
can
do
this.
Is
you
need
to
experiment?
You
need
to
create
an
HTM.
We
do
a
bunch
of
data
tweet.
Some
model
parameters
go
around
chasing
aggregation
settings,
Frank
figure
out
how
to
make
that
initiative
system
understanding
that
data
shape.
A
A
A
In
addition
to
that,
we
need
to
figure
out
how
we
can
batch
process
yes,
batch
process
that
data
in
some
fashion,
as
well
as
live
stream
process
that
deal
so
the
first
thing
I
need
to
do
is
figure
out
how
to
take
one
of
those
isolated,
Dave
screams
it's
out
in
the
world.
Somehow,
and
let's
just
say
for
simplicity
sake,
this
is
just
a
scalar
data
feed
one
scalar
value
changing
over
time.
A
A
I
need
to
be
able
to
emit
that
back
out
somehow,
so
it
can
be
consumed
by
others
who
want
to
find
value
in
the
data
are
buying
my
agent
system
and
able
to
apply
navigation,
because
the
raw
rate
of
data
coming
from
a
lot
of
these
sensors
in
the
world
today
is
just
too
too
high
or
too
erratic,
or
it
needs
to
have
some
time.
Smoothing
your
memories
for
subsampling
or
something
so
I
need
to
be
able
to
do
that
with
paging
system
as
well,
and
this
needs
to
be
easy
to
rerun
low-cost.
A
Without
you
know,
losing
any
any
data
ready
for
automation,
because
I
definitely
adventist
and
easy
to
consume
the
results.
So
that's
my
goal.
That
was
my
goal
with
10
plus.
This
is
a
prototype
of
this
system
using
influx
vb
as
native
skin
sores
root
zone,
a
quick
aside,
I'm
in
trust
DD.
It's
a
tiny
serious
database
as
like
sequel
is
not
no
sequel
is
not
because
they're
not
time
series
a
totally
different
phase.
The
type
series
database
is
completely
built
around
the
idea
of
values
over
time.
B
A
By
the
time
associated
with
it,
I'm
not
going
to
details
of
its
box,
TV
I'm,
just
going
to
say,
fits
the
HTM
data
requirements
very,
very
well.
That's
why
I
chose
it
as
a
way
to
start
trying
to
process
building
HD
models
of
learn
on
the
streams,
so
this
is
interfaced
could
be
applied
to
a
lot
of
other
difference.
Extremely
datasources
I
want
to
hang
on
in
flux,
but
the
the
same
thing
I'm.
A
B
A
Be
applied
to
all
of
these
other
temporal
data
streams
that
are
arising.
Riverview
is
one
that
I
wrote
soprano
open
data
API.
It
is
really
powerful,
open-ended
guys
free
for
anyone
to
use.
You
can
consume
hundreds
of
gigabytes
of
data
through
the
API.
We
can
also
try
and
find
interface
like
this
on
top
of
that,
not
to
mention
a
link
which
could
also
be
a
potential
data
source
or
data
processor
that
fit
integrate
with
the
system
like
this.
So
the
by
the
sky
idea-
and
this
is
apparently
to
meet
pi-
is.
A
But
I'm
going
to
show
you
this
endeavor,
so
this
is
the
this.
Is
the
structure
of
the
devil?
There's
a
lot
of
moving
parts?
It's
one
of
the
reasons
I
needed
to
simplify
this
and
come
up
with
in
plus
HTM
is
because
there's
a
lot
of
moving
parts.
First
of
all,
smart
things
is
an
IOT
platform.
That's
a
commercial
entity
platform.
My
samsung
I
know
the
guys
matter
the
confidence
for
one
of
the
reasons:
I
music,
but
it's
a
very
happy
boy,
LT
popcorn,
so
you
can
get.
A
Your
sensors
for
the
easily,
so
basically
you
buy
hub,
become
too
much
sensors
installed
around
the
office
or
the
house
or
whatever
you
monitor
is
going
to
the
bathroom
all
the
time.
What
motion
sensors
going
off
and
when
switches
are
being
thrown
our
consumption
of
the
coffee
machine
I've
got
all
these
things
and
you
can
go
get
that
dia,
a
very
smart
app,
which
I
did
that.
Basically
just
sits
in
that
tub
or
wherever
and
every
time
in
one
of
those
sensors
and
that's
anything.
It
just
posts.
A
Both
erase
them
raw
to
invest
ed
in
three
times
three
days
he
has.
He
had
me
how
to
graph
it.
So
I
can
see
over
time
how
much
power
consumption
my
television
is
using
and
once
I
do
get
the
HTM
system
in
place
right
additions
for
not
only
the
information.
I
can
grab
that
the
top
of
the
sensory
input.
A
A
Get
a
new
pieces
from
smart
days:
you
push
it
too
long.
That's
exists
on
a
cloud
I'm,
not
sure
it
badly.
The
model
results
and
ready
to
it
was
TV.
What
I
got
out
with
that
was.
It
then
became
really
a
bulky
me
to
experiment
with,
because,
first
of
all
that
I
get
breaking
because
it
was
you
run
for
five
bottles,
nipples
Pascal
Wednesday
to
it.
That
was
really
just
a
toy
implementation
that
founded
in
McKinsey
and
I
wrote
and
it
it
falls
over.
A
A
A
A
Guy's,
ever
okay,
we
had
a
mini
fridge
in
our
office.
This
is
its
power
consumption
over
time.
As
you
can
see,
it
has
pretty
standard
pattern.
Every
four
hours,
or
so
its
spikes
runs
in
terms
of
okay,
ignore
this
flat
and
that
was
I.
Don't
know
something
I
Denari
can
remove
her
another
one
of
my
expressions.
A
B
A
Five
minutes
really
see
the
shape
of
the
day
to
change
as
I.
Do
that
and
I
found
that
you
know
five
minutes
to
15?
Isn't
that
big
a
difference
right?
So
what's
the
point
going
any
higher
resolutions
on
taking
this
a
15
minute
meeting
aggregation
period?
Ok
and
now
I'm
going
to
actually
run
HTM
using
inputs,
ischium.
A
A
Writing
group
and
basically
is
I,
have
a
start,
processing
app
start
class,
stop
processing
app
that
defines
my
endpoints
an
aggregation
period,
which
is
15
minutes
here
kind
of
path
to
some
crackers
that
they
are.
You
knew
before
you
know
you
got
a
hopper
and
that
justifies
what
popular
type
of
model,
anyone,
how
all
different
settings
work
you
and
then
I
just
run
while
I
send
to
them
and
in
the
run
model,
is
basically
going
to
go
to
influx
GD.
Do
that
mean
aggravation
over
that
time
period
and
point
out
of
time
passed?
A
B
A
Be
able
to
your
update
this
rush
back
to
one
month,
a
good
15
minutes
three
so
as
I
hit
update,
that's
the
east
and
ateh
coming
in
over
top
of
the
center
day,
and
so
I
can
turn
out
like
they
almost
or
light
behind
it.
You
about
the
lightly.
This
little
bit
difference
its
story
because
you're,
better
indication,
whether
it's
really
not
as
hyper,
sometimes
anonymous
for
go
up
and
down
like
crazy.
So
the
anomaly
like
the
good
and
I
have
not
really
tuned
in
this
very
well
I
just
got
this
morning.
So.
A
Wish
I
would
be
runnin
runnin
to
model
for
a
reason,
but
it
does
at
least
give
you
an
indication
that
you
have
clip.
Obviously
this
is
anomalous.
There's
a
big
flat
line
of
data
here,
remember
what
happened
and
still
running
here
and
in
the
fighting
is
the
usage
shape
of
the
data
once
it
gets
over
your
change
guitar
to
figure
out
why?
But
it
definitely
is
not
comfortable
that
change,
especially
when
you
know
flatline
plateaus,
that
a
different
value
that's
usually
see
before.
A
A
A
A
A
Measurement,
which
is
power,
model
or
power
difference
which
the
timestamps
are
completely
synced
together.
So
if
I
want
to
sit
those
together,
it's
easy
to
do
and
efficacy
in
charting
is
two
different
complexity
games
it
together,
but
they
both
have
different
basic
settings.
You
know
maybe
eight
different
intervals
that
have
to
be
was
submitted.
The
Robbie
have
much
more,
but
every
inference
of
operas
old,
passed
away
on
some.
B
Can
I
make
the
point
that
the
Flint
stuff
blank
slate
has
a
philosophy?
We're
straining
is
supposed
to
subsume
bash
us,
so
the
past
and
sort
of
goes
over,
but
the
past
people
have
had
sort
of
x,
48
approach,
they
had
their
code
written
for
batch
and
their
low
Britain
first
streaming,
and
they
call
us
a
land
on
architecture.
This.
B
B
A
Buses,
as
I
get
retention
policies
that
allow
you
to
take
completely
raw
data
and
take
for
like
an
hour
and
then
runs
continue
square
is
too
abstract
being
greedy
aggregation
periods.
Anyone
for
all
the
data
is
longer
attached
box,
so
you
always
have
like
this.
Cash
is
one.
That's
all
data
be
around
a
40-story
forever
yo
Gotti's
extracting
things.
You
want
a
normal.