Monday, October 18, 2010

Class notes: 10/18/10

Reminder: Final submission - Weds at noon!

What's a good intermediate number of lights?
Lauren: 5 is pretty good, it's an average case
Daniel: You can easily cover 50% of the map with 5 lights

We did some runs with G5.H map
G1: 270
Daniel: We find the spot with the most area and place the lights iteratively
Zach: We can construct linear and branching chains. We build the lights in a graph, then find the one that has the lowest average distance to all of the other lights, and put the collector there. We have a formula that calculates how long the pulses will be by looking at the "depth" from the
G5: 2013
Nitin: We don't do so great with the shadows
G2: 649
Jesse: G1 put the lights right on the perimeter of each other, whereas we make them somewhat closer
John: We create a 2D array and assign an "openness" value to each point

Nitin: We are all using a different approximation, and it seems like the people that take a finer granularity do better.
G3: 3972
Elba: Even if your chaining is slow, it's wayyyy better than random drift
G7: 3950
Dfed: We We have a "dumb" approach where we do a lot of random tries, then do whatever works best
Zach: Depending on how we place the lights, we can scale linearly
Andrew: The chain computation doesn't blow up, but the time to run the actual simulation takes time

G4: 850
G6: Didn't terminate
Yang Jian: We should have been chaining

Tournament configurations?
How many lights should we expect?
3, 10 came out of a hat. What is reasonable?
What about if you had a maze?

Upper bound: 50/100

How many times do we need to run to reduce variance?
Lauren - 5
We will rewrite the simulator so that it will run multiple runs with the same light configuration. I'll pick a reasonable number of runs (on the order of 10, probably)

How many lights?
3, 5, 8, 13, 100

What's an OK upper bound for a max rounds? 15,000

Blank
Boxes & Lines
Cage
Caged
G5 H
Steps
W

We will come up with:
1 maze
2 more hidden ones

No comments:

Post a Comment