I've got to hand it to Prof. Toma, my advisor on the summer research. Her massive alterations made the blurb much more readable to an interested lay person.
Monday, October 12, 2009
Monday, August 10, 2009
copyrights and incentivising information production
If you haven't seen it, I highly recommend this discussion between two of my favorite prognosticators, Charles Stross and Paul Krugman. There's a bunch of interesting play between the perspectives of the two intelligent guys, but the best bits are thrown in at the end once they really get going: copyrights, intellectual property, newspapers, financial ratings institutions, and how they all are interrelated, failing, and in need of a new approach.
Krugman, writer for the New York Times, is especially aware of that company's pending problem with paper subscribership, online consumption, and digital reproduction of their product. He points out the current dependance the paper has on the pleasant qualities a newspaper brings over a computer at breakfast, a "very thin lag in technology" that keeps the paper afloat because it is currently the best (most comfortable, 'normal', simple) news medium. One only needs to consider advances like the Amazon Kindle and RSS feeds and aggregators to remark upon how thin the lag is becoming.
Stross hits the big point, that the issue is one of information production on a much larger scale, where digital reproduction and the internet have replaced the librarian and made copyright irrelevant by eliminating the boundary that divided personal non-commercial reuse and for-profit redistribution. Stross himself releases many of his books, novellas, and short stories under a Creative Commons license, so it is especially interesting to hear his side of this issue.
There are two fascinating points, one by each theorist, that definitely deserve more consideration: bandwidth tax and non-compromised financial ratings. I suggest you read the original to learn how it all connects.
Krugman, writer for the New York Times, is especially aware of that company's pending problem with paper subscribership, online consumption, and digital reproduction of their product. He points out the current dependance the paper has on the pleasant qualities a newspaper brings over a computer at breakfast, a "very thin lag in technology" that keeps the paper afloat because it is currently the best (most comfortable, 'normal', simple) news medium. One only needs to consider advances like the Amazon Kindle and RSS feeds and aggregators to remark upon how thin the lag is becoming.
Stross hits the big point, that the issue is one of information production on a much larger scale, where digital reproduction and the internet have replaced the librarian and made copyright irrelevant by eliminating the boundary that divided personal non-commercial reuse and for-profit redistribution. Stross himself releases many of his books, novellas, and short stories under a Creative Commons license, so it is especially interesting to hear his side of this issue.
There are two fascinating points, one by each theorist, that definitely deserve more consideration: bandwidth tax and non-compromised financial ratings. I suggest you read the original to learn how it all connects.
The blurb
Improved Visibility Computations on Massive Grid Terrains
Jeremy Fishman and Laura Toma
Calculating visible points on grid or triangulated elevation maps is a computationally intensive task. A simplified nearest-neighbor interpolation for grid terrains equates to a Voronoi decomposition and permits a lower upper-bound on computation time. However, the well-known Van Kreveld algorithm under this simplification is still heavily I/O-bound on larger grids. We developed several much improved algorithms under the I/O model of Aggarwal and Vitter, including two cache-aware full visibility algorithms and one cache-oblivious approximation. Further work is focused on removing the Voronoi simplification in order to allow use of linear interpolations between grid points. A comparision of both interpolation models and the several I/O approaches can be performed using random viewpoints distributed among terrain features selected by flow modeling, and will allow us to quantify the trade-offs between computation time and accuracy on grids up to hundreds of gigabytes.
Jeremy Fishman and Laura Toma
Calculating visible points on grid or triangulated elevation maps is a computationally intensive task. A simplified nearest-neighbor interpolation for grid terrains equates to a Voronoi decomposition and permits a lower upper-bound on computation time. However, the well-known Van Kreveld algorithm under this simplification is still heavily I/O-bound on larger grids. We developed several much improved algorithms under the I/O model of Aggarwal and Vitter, including two cache-aware full visibility algorithms and one cache-oblivious approximation. Further work is focused on removing the Voronoi simplification in order to allow use of linear interpolations between grid points. A comparision of both interpolation models and the several I/O approaches can be performed using random viewpoints distributed among terrain features selected by flow modeling, and will allow us to quantify the trade-offs between computation time and accuracy on grids up to hundreds of gigabytes.
Saturday, August 1, 2009
Subscribe to:
Comments (Atom)
