Queries on Spatial

kneewastefulAI and Robotics

Oct 29, 2013 (3 years and 5 months ago)

67 views

Reviewers:

Lydia Monikonda

Usha Kumar


Review of Group 8 Final Presentation

Title:

Categorical Range Queries on Spatial Networks
,

by

G8
Rahul Saladi, Xiaofei Zhao, Akash Agrawal and Anuj Karpatne

http://www
-
users.cs.umn.edu/ rahuls/cs8715/

Queries on Spatial


Problem Motivation:

What kind of research (e.g. Basic or Applied research) did the talk

relate to? If it was basic research what fundamental questions or open problems did the

speaker relate to? If it was applied research, what applica
tion use cases / examples/ questions

did the speaker relate to? Do they have societal significance/are they made up? If they are

made up what are some realistic application questions that can be addressed? Briefly explain

within 75 words.


The presentation

foc
used on the applied research related to spatial range queries i.e, finding places of
interest wit
hin a particular distance range on spatial networks. The main focus is to consider the
network distance between the places, depending upon the connectivi
ty between them. These kind of
queries are extensively used in mobile applications to find places of interest along the way.

The main
social
significance

is in mobile applications which need to show where the nearby restaurant is on one’s
tra
v
el route.


Problem Statement: What research problem / software prototype did the talk attempt to

solve / build? Are the listed constraints reasonable? Why or Why not? Was the input / output

well defined or easy to understand via a simple example? (75 word limit)


The

presenters are attempting to solve the
problem

of “Given a spatial network,

the current location of
a person and the distance within which he/she need to identify the places of interest, how
quickly/efficiently one could solve this problem”. The exis
ting algorithms available for this purpose are
not very good at handling worst case
scenarios

and also in handling sparsely populated data. The
talks
clearly define

input and
output

with easy to follow examples. The constraint is not explained in the fin
al
presentation
, perhaps the time
taken to

execute the query could be considered as a constraint here, as
it needs to be executed in a real time timely manner.


Challenges: Are the challenges articulated clearly (and possibly illustrated via an example)?

If so, what are they and indentify their category (e.g. computational, statistical, others)?

Briefly describe some improvements that can help articulating challenges easier? (75 Word

limit)


The challenge is in coming up with a new algorithm to solve the p
roblem of range queries, by addressing
some of the issues like scanning entire spatial networks in worst case
scenarios
, and also making sure
that this algorithm performs better than the existing schemes. If the presenters had given an example
of how the
existing algorithms implement this and why they show poor performance in comparison to
the proposed approach; that would have tremendously improved the understanding of the
presentation.


Proposed Approach: Did the talk explain the key elements of the prop
osed approach clearly

via the use of suitable examples? If so what were the key elements? Did the proposed approach

honor all the constraints listed in the problem statement while achieving the goals listed in

the problem or were there some (simplifying) a
ssumptions? If so, what were they? Briefly

explain (and possibly include suggestions to improve the proposed approach to solve the

stated problem). (150 Word Limit)


The
depiction

of the nodes in
facility neighborhood graph in the
slide helped to explain
some of the

key elements.

As the constraint is not well stated, it is hard to judge if it honored all the constraints.

The presenters could

have improved the presentation of their proposed approach
by providing

the
constraints as well as an evaluation m
ethodology that supports their claims.


Novel/Better: Did the talk give a reasonable (e.g. thorough / acceptable) classification of

related work and identify their limitations via examples? Was the novelty of the proposed

approach

clear from the classification scheme used in the talk? If not, how can it be improved?

Did the talk emphasize on how the proposed approach was better with respect to related

work via additional examples? If not, how can it be improved? (100 Word Limit)


N
ot much information is presented on related work in the final presentation. It seems from the
presentation that the team has done some research on the existing algorithms like
Djikstra's or A*,
which are predominantly used for solving this kind of proble
m and in this new approach they
restrict the search space by identifying neighborhood facility graph to restrict the search. The
claim is that since they are not scanning the entire network, it will speed up the execution.

The presenters could have stated

a few papers on the range query algorithm implemented using the

Djikstra's or A*, to substantiate their claims.

Validation: Did the talk provide hints about the validation methodology or provide examples

to

validate their contribution claims? (50 Word Limit)

Validation methodology is not clearly discussed during presentation, however, the presenter mentioned
that they planned to take it up us a furture work to validate theis proposed approach on real world
n
etworks and also compare the performance of the algorithm with respect to the existing algorithms.


Presentation Critique: Rate the talk on a scale of 0(poor) to 10 (excellent) and provide a

brief justification (50 Words) while suggesting areas for improve
ment on the following:





Was the talk accessible t
o an "intelligent lay person"? 9

The talk had clear citation of examples and diagrams which are quite easy to understand and follow.





Did the talk emphasize a central message that conveys the overall v
alue of the work

being executed? 9




yes, it clearly states the new algorithm they have proposed.


Did the talk attempt to relate to the audience and showed effort in conveying key ideas

clearly? 10




yes, hand plenty of slides and diagrams explain thei
r work in an easy to understand manner.


Was the speaker's response to questions satisfactory? 10




yes. He handles the question very well.


How did the talk do on covering the 6 elements? Kindly rate each element separately

and include a brief justificat
ion for each.

8.

The talk did not have much on related work and evaluation methodology, though the talk provided
some hints in this direction.


Problem definition: 9



Could have provided more insight on this. The talk just

Mentioned its use in mobile
app.

o


Problem importance: 10


mobile app as well as in day to day activities.

o


Problem hardness: 8
, should have explained well as to why the traditional algorithms do not perform
well.


o


Description of Methods: 10 the methods were explained
clearly
.


o


Novelty of methods: 9
, the novelty is mentioned clearly, however, it did not previous papers in the
presentation, though the presenter mentioned about a particular authors work during the talk. Could
have provided this in the slides for the audien
ce to go back and read.

o


How are the methods better? 10 provided examples on how their methods

are better


o

10.

They provided an optimal approach by using nearest neighbor graph.