20 Dec 2013
Sugar for the technical mind (Part II)
Second edition of “Sugar for the technical mind” (a.k.a., papers I’ve read recently).
As before, light reads are papers without much jargon and technical prerequisites. Heavier reads will usually require full concentration to understand and may contain a lot of equations and derivations.
Arc Diagrams: Visualizing Structure in Strings by Martin Wattenberg, 2002 (Medium) : Pretty static music visualization - nuff said.
How NOT to review a paper - The tools and techniques of the adversarial reviewer by Graham Cormode, 2008 (Light) : Purely for entertainment, how to review papers in all the wrong ways.
Generating Complex Procedural Terrains Using the GPU by Ryan Geiss, 2007 (Medium) : By now, most people should be used to seeing CGI landscape of valleys and mountains, which are basically a 2D grid with a heightmap applied. This technique, however, describes a way to generate terrain and other 3D features that cannot be created with just a mesh - and the results are very impressive.
Parallel Prefix Sum (Scan) with CUDA by Mark Harris, Shubhabrata Sengupta and John D. Owens, 2007 (Heavy) : There are a lot of cool algorithms for GPU, in particular on NVidia's GPU Gems. I'm picking this one specifically not because it produces pretty pictures, but because it presents a neat solution to a seemingly non-parallelizable problem. Further, this "scan" operation is a basis for many many GPU algorithms, such as array compaction, radix sort, etc.
Ray Tracing Deterministic 3-D Fractals by John C. Hart, Daniel J. Sandin, Louis H. Kauffman, 1989 (Medium) : This paper describes how, assuming you have an equation that gives you the closest point on a fractal for any point in 3D space, a way to raytrace 3D fractals very efficiently. The cool thing about this paper is that it was published much earlier than fractals such as the Mandelbulb and Mandelbox were discovered, and is now commonly used to render some of the most awesome fractal art ever created.
Real-Time Content-Aware Image Resizing by Huang Hua, Fu TianNan, Rosin Paul L. & Qi Chun 2010 (Medium) : I mentioned Shamir & Avidan's original paper a previous blog post. This optimized version (for speed) helped me win a Nokia hackathon</a> recently.
The camel has two humps by Saeed Dehnadi and Richard Bornat, 2006 (Light) : This one is not peer-reviewed and strongly controversial - it attempts to explain a phenomenon which I have observed myself: in introductory computer science classes, there often appears to be two groups of people. There are those who just never seem to "get it" and those who find it really easy and get bored to death. Their findings are to be taken with a grain of salt, but there are some interesting ideas in there, namely that the difference between the two groups lies in the whether a consistent model about how the computer works is ever formed.