Transcriptions
Note: this content has been automatically generated.
00:00:03
do you know that by two thousand and fifty seventy
00:00:06
percent of the global population believe in c. d.'s
00:00:11
for such overpopulated c. d.'s to be sustainable it is critical
00:00:15
that city officials make informed decisions about the city planning
00:00:20
however cities are complex environments that are shaped by multiple factors
00:00:26
any location in the city may have all kinds of information associated with it
00:00:32
from the axial currencies to noise complaints and tax
00:00:36
activity and this information can be pretty detailed
00:00:39
for example taxi records can have several attributes such as they pick
00:00:44
up and drop off time fair deep and number of passengers
00:00:49
all this complex information is usually publicly available in
00:00:53
the form of numerous urban data set
00:00:56
to penalise these data sets and the correlation some on them
00:01:00
city planners there are two these were utterly takes systems
00:01:03
using the systems instead of looking at numbers they look at callers
00:01:08
on the mob as you can see on my slide
00:01:10
while they can compare multiple data sets using just the single visualisation
00:01:15
to generate these visualisation as usual analytic systems rely
00:01:20
on what we call spatial application realise that
00:01:23
partition the data over to different regions and then compute summarise information for each region
00:01:31
this data partitioning is achieved by performing point in polygon test
00:01:36
to determine the data points contained within each polygon or save region
00:01:42
based tests are computationally heavy and considering the hundreds of millions or
00:01:48
even billions of points that the data set might contain
00:01:51
these poses a big challenge for these what analytic systems making interactive response is hard to achieve
00:01:58
in fact no existing technique can answer spatial application queries in real time
00:02:04
our research battles this challenge by evaluating spatial navigation queries using
00:02:10
renderings which are operations highly optimised for graphics hardware
00:02:16
just to give you some intuition our approach drones the data points on agreed it current fuss and keeps
00:02:22
track of intersections between the points and the canvas but
00:02:25
maintaining compassion aggregates in the canvas greed cells
00:02:29
it then throws the polygonal regions on the same canvas and calculates the aggregate result
00:02:35
from the bachelor art gets off the grid cells that intersect with each polygon
00:02:40
this approach leverage is that massive parallelism provided by more than graphics hardware
00:02:46
and thereby evaluates spatial education queries on the fly in real time
00:02:52
that way we enable city planners to interactively navigate urban