Main.BSF2006 History

Hide minor edits - Show changes to markup

October 16, 2006, at 12:57 AM by 194.90.37.5 -
Changed lines 9-10 from:
to:
October 16, 2006, at 12:55 AM by 194.90.37.5 -
Changed lines 12-15 from:
  • Other alternatives:....
to:
  • Random Walk with Choice in General Metric Space
  • Power of 2 Choice in (Randomize) Networks Algorithms
Changed lines 18-24 from:
  1. Improve random walk techniques for sensor networks. This includes the power of choice and reducing the resistance between nodes of interest.
    1. random walks on subgraph of random geometric graphs, like the delaunay triangulation and other planar subgraphs.
  2. trade-off between random walks and gossip in terms of energy and time. Combine both techniques and/or classify when to use each.
  3. Dynamic network (over time): many topics to cover. representation, mobility, links error, failure. What is a simple model that we can start with. There is a lot of space for theoretical work as well as simulation and I think the modeling itself is not a simple issue. Tzika has done some work in the area and there seems to be more work than I was aware of. Never the less I think there is much to do here. What are the question we want to ask on these models (after we'll have them)? connectivity, clustering, random walks, routing...
    1. determine the regions of dynamics for which it is better to use optimize deterministic algorithms and for which it is better to use randomize algorithms.
to:
  1. 16/10 Discussion: Concentrate on the Power of Choice:
    1. The hitting time on the line for RWC is O(n1.5), for the lollipop O(n1.5log(n)). Theoretical proofs.
    2. Simulation results for planar subgraphs of RGG: RNG, GG, random delaunay triangulation.
    3. Prove that the complete graph gives the least possible improvement.
    4. The affect of choice on mixing time and particular gossip algorithms for sensor networks.
    5. Theoretical results for the most visited node in RWC in some graphs.
  2. Previous Topics:
    1. Improve random walk techniques for sensor networks. This includes the power of choice and reducing the resistance between nodes of interest.
      1. random walks on subgraph of random geometric graphs, like the delaunay triangulation and other planar subgraphs.
    2. trade-off between random walks and gossip in terms of energy and time. Combine both techniques and/or classify when to use each.
    3. Dynamic network (over time): many topics to cover. representation, mobility, links error, failure. What is a simple model that we can start with. There is a lot of space for theoretical work as well as simulation and I think the modeling itself is not a simple issue. Tzika has done some work in the area and there seems to be more work than I was aware of. Never the less I think there is much to do here. What are the question we want to ask on these models (after we'll have them)? connectivity, clustering, random walks, routing...
      1. determine the regions of dynamics for which it is better to use optimize deterministic algorithms and for which it is better to use randomize algorithms.
Changed line 34 from:
  1. What should be the main outcome? theoretical results, or we will also want to have practical side (which will require equipment and simulation
to:
  1. What should be the main outcome? theoretical results, or we will also want to have practical side (which will require equipment and simulation.
October 01, 2006, at 02:33 PM by 194.90.167.95 -
Changed lines 21-23 from:
to:
  1. determine the regions of dynamics for which it is better to use optimize deterministic algorithms and for which it is better to use randomize algorithms.
October 01, 2006, at 02:27 PM by 194.90.167.95 -
Added line 18:
  1. random walks on subgraph of random geometric graphs, like the delaunay triangulation and other planar subgraphs.
Changed lines 20-21 from:
to:
  1. Dynamic network (over time): many topics to cover. representation, mobility, links error, failure. What is a simple model that we can start with. There is a lot of space for theoretical work as well as simulation and I think the modeling itself is not a simple issue. Tzika has done some work in the area and there seems to be more work than I was aware of. Never the less I think there is much to do here. What are the question we want to ask on these models (after we'll have them)? connectivity, clustering, random walks, routing...
Added line 28:
  1. Bhaskar and Tzika, sign up at the BSF site so I can add you to see the proposal on-line.
Changed lines 35-46 from:
to:
October 01, 2006, at 02:16 PM by 194.90.167.95 -
Added lines 5-6:
Changed lines 11-13 from:
to:
  • Current Title: Randomize Algorithms for Dynamic Sensor Networks
  • Other alternatives:....
Changed lines 17-18 from:
to:
  1. Improve random walk techniques for sensor networks. This includes the power of choice and reducing the resistance between nodes of interest.
  2. trade-off between random walks and gossip in terms of energy and time. Combine both techniques and/or classify when to use each.

Questions / Discussion

  1. What should be the main outcome? theoretical results, or we will also want to have practical side (which will require equipment and simulation
  2. What is the focus of the work? We need to be very clear on want we want to achieve in the proposal.
  3. For a startup request (2 years, $60K) what is the expected outcome, 3-4 papers of the investigators? we need to contact the group here at BGU that deals with the proposals to find that out.

To do List

  1. Divide the workload
  2. set a time line for the work (taking into account the paper work that is required at each university, signatures , etc)
October 01, 2006, at 01:56 PM by 194.90.167.95 -
Added line 29:
  • BSF login
October 01, 2006, at 01:54 PM by 194.90.167.95 -
Changed lines 16-18 from:

Links

to:

Links

October 01, 2006, at 01:53 PM by 194.90.167.95 -
Changed lines 6-18 from:
to:

Title

Possible Topics

References / Reading

Links

October 01, 2006, at 01:51 PM by 194.90.167.95 -
Changed lines 3-6 from:
to:
October 01, 2006, at 01:45 PM by 194.90.167.95 -
Changed line 6 from:
  • [Links]
to:
October 01, 2006, at 01:45 PM by 194.90.167.95 -
Changed lines 3-5 from:
  • [Title]
  • [Possible Topics]
  • [References / Reading]
to:
October 01, 2006, at 01:42 PM by 194.90.167.95 -
Added lines 1-6:

BSF 2006 - Grant Proposal

  • [Title]
  • [Possible Topics]
  • [References / Reading]
  • [Links]