neural networks live!
have yet to read the paper, but it looks exciting.summary
Science 28 July 2006
Reducing the Dimensionality of Data with Neural Networks
G. E. Hinton* and R. R. Salakhutdinov
art weekend
friday:mission17 gallery. friend jesse was a curator. indeed! nice pidgeons made of felt, with i think the stories of street people playing on squeeze-activated audio insides. others were cast in plastic and made into shakeable snow-pidgeons of dreams. or lost dreams. and then cancun, oh how we've missed out! 2 blocks away, so good.saturday:sears/jackson space...sunday: symphony in dolores park. tchaikovsky's piano concerto no. 1. louder and more powerful classical music is unlikely to be heard again in a very long while. then they eventually got on to the 1812 overture which, well, still reminds me of transporting conduction experiences with one hand out the driver's side window in the middle of the night back home.but the best part of the performance, though definitely aided by wine and beer, was the conductor - one alastair willis. it was like he was conducting with his whole body. in ways that, for example, you'd never see in masur. the swayin hips during the piano concerto were kindof seductive. part of me hopes this is a new trend - conducting through dance. but it has to be understated, or rather, it must make it seem like powerful energies are being held at bay, only letting through a glimpse of what could be.
verbal insight follows
scanning hundreds and hundreds of known words.
lolipop icicle is revealed in the etymology of popsicle.
disgusting suddenly makes sense with a gustation base.
and most importantly, chow is unveiled as an adjective to be used with mein.
awaiting other insights from an unhappily focused brain.
so
a favorite researcher is now posted at a school where i have already interviewed. my app obviously can't get worse, so . . . would this be what i want to do?
it's not necessarily affective, but it moves toward solving a hard problem, though not the hard problem (i.e. the feeling of what happen). i am constantly amazed at how the human brain develops and successfully navigates a valenced environment, and in this field i know i can generate tractable problems. and it satisfies a theory fetish, quite honestly.
things to do
i am occasionally strongly motivated and then struck with ennui, indecision, and down down.but when up for keeping going, this is where the going is going. and some of this is actually getting done:I. learn new words. it is honestly entertaining. but the difficulty is that in all my years, i have never learned how my brain best learns new stuff when it has to learn and not when it wants to (e.g. my berkeley record). i really don't know how to train myself. but i think it's a worthy challenge to right the injustice that was my near-perfect math score and near middling verbal, when all these years it's been the math that has been the nemesis (granted, gre math is for junior high students). any ideas for good word books?
II. personal statement. do it.
III. NSF fellowship application. bulletproof. make admission undeniable.
IV. reinvigorate ancient berkeley thesis work, submit it, and along the way reawaken my intellectual curiosity, back to where i feel my will to power.
V. apply. early.
X. decide what i care about, if anything.
solving the stem cell/unused embryo debate
