open source maps

stumbled across of open source map codes: roadnav and roadmap. both use census data for generating street maps, and both can use gps input to show your location. roadnav can compute paths for you, though neither knows about one-ways (it's not in the data). a fork of roadmap is working to possibly remedy that problem, however. might be interesting to put them on my laptop, for when i travel.

list.it information scrap organizer

i'm a big user of google notebooks and i'm always on the lookout for better ways to keep quick, short bit of info organized (and off of the paper that seem to surround me anyway). list.it sounds like an interesting concept, and the folks at csail always seem to know what they're doing.

investment entities

why do tax laws have to be so complicated? i found a site that gives some info on the various types of entities to use as vehicles for active trading, and the tax pros and cons for each. i'm afraid i'll need to deal with this at some point. i guess it might be a choice between hiring a lawyer before or an accountant after. or both.

some books on ica

an information-theoretic approach to neural computing, qa76.87.d47 1996 has intros for both information theory and anns, shows connection between pca and ica. has ref to mdl with supervised learning, but it's the old (not stochastic complexity) mdl. independent component analysis: theory and applications, tk5102.9.l44 1998 time-delayed decorrelation, nonlinear ica, some historical info in preface. separated discussions of methods for sub- and super-gaussian (kurtosis < or > gaussian kurtosis) good illustrations of how ica can solve some problems of pca. looks like some deep math, but intelligible advances in independent component analysis, qa76.87.a378 2000 temporal effects: multivariate time series, ref to financial time series; time-varying mixtures, particle filters nonlinear mapping, analyzing independence assumption, ica on noisy data the time-varying stuff really looks interesting independent component analysis: principles and practice, qa76.87.i516 2000 non-stationary sources, including particle filters nonlinear, though not many equations or examples a lot of material seems very similar to the previous book independent component analysis, qa278.h98 2001 lots of background, justification of ica approach preprocessing and time filtering, comparison of algorithms good overall tutorial financial applications independent component analysis: a tutorial introduction, qa76.87.s78 2004 shorter, more intuitive explanations probably better for getting the basic idea kind of dumbed-down in some ways (i know what lambda looks like!) skip the first 6 chapters if you have a decent math background a lot of the ica books seem connected not only with information theory, which i would expect, but also with other stuff on neural nets, which i didn't necessarily expect. this one might be worth checking out. refers to some generalizations of the ica/pca ideas, with good presentation. more stuff on the group's website.

ubuntu, sabayon, etc.

i've been thinking about trying ubuntu (or xubuntu) on my next machine. it seems to be the biggest fad to hit linux in a while, and i have to admit that my more geriatric hardware has a hard time keeping up with gentoo updates. and packages like checkinstall give me hope that i will still be able to pick up scruffy-looking hitchhiker software along the way. the best of both worlds, of course, would be a meta-distro built on top of gentoo. that way i could let somebody else build binaries for most of my packages and still grunge around when i want. i've read a little debate about how this could be accomplished, especially given the exponential combinations of useflag combinations. i lean toward the p2p approach, but i don't see anyone doing it yet. but apparently some people are working on distros based on gentoo, like the stupidly named sabayon project. maybe one of these will eventually be worth the trouble.

ica and svd

been poking around some more with svds, and came across this presentation that refers to independent component analysis. the wikipedia article is decent, with good refs. one way to view the ica is as a nonlinear generalization of pca; the components are not necessarily orthogonal but they minimize the mutual information and all the higher-order cross cumulants, rather than just the second like pca. the conditions of uncorrelated and independent are equivalent only for gaussion random vars, but ica has determinacy problems of more than one of the components are gaussian. this ebook has good info on it in chapter 6, if you can get access.

3d pdf objects

i tried for way too much time to compile the u3d reference library from sourceforge, with something tantalizingly close to success. the only way i could actually use it, though, was to use wine on the precompiled windows binary on an idtf file. the only way i have successfully created an idtf is by exporting an imported stl file from meshlab. i did manage to shove the resulting u3d into a pdf, and it looked pretty cool. but i need to find an easier way to make idtf files. EDIT: apparently vrml 97 can be converted to u3d with acrobat 3d. s2plot site gives some instructions, including some basic frame-based animation. also, some russian dude made a vtk->idtf exporter. not sure if he published anywhere. also also, mathgl supposedly has idtf export capability, so things like udav that use it should, too.

so many time wasters, so little time...

there are some neat os and free games out there. if only i were idle rich. planeshift looks very well done for a completely volunteer project. i think their idea for npc ai is really cool and opens up a lot of possibilities, basically allowing an interactive fiction game integrated into an mmorpg. i tried it once and couldn't see much in the way of npcs and other players. i hope they keep working on it. the spring project has an interesting engine for a variety of rts. tremulous is a fps with some rts elements. vega strike is a sci-fi themed mmorpg with some stunning graphics and very involved back story. just like eve online, if only players could disembark and walk around, this could be a killer game.

time-varying svd

recently i've been trying to see if anyone else has worked on time-varying singular value decompositions. there's surprisingly little out there on it, at least that i can find. i've tried thinking of the svd eigenmodes in terms of rotating coordinate systems. i remembered from dynamics classes once upon a time that some ways of representing rotation have a gimbal lock problem, but quaternions do not. and (i think) they also have linear propagation equations. the wikipedia article on it looks pretty good. i should probably read that and the 'see also' articles at the bottom.

random thought fragments on multiresolution analysis

someday i'd like to write a textbook on multiresolution techniques in surrogate modeling. in the intro i would present a unified approach to orthogonal decompositions like eigensystems, svd, and ffts with wavelets, linear model sensitivity analysis, and mdl. i would go on to answer questions like, 'how much information is gained in terms of the modes and singular values if i update an svd representation of a surrogate model?' and how do i do an svd surrogate model in the first place? will it capture features on different scales, or can i force it to? how do i use that predicted information gain to guide adaptive sampling?