"the web, the chain, the tree"

Last week I went to a talk by Steven Pinker on elements of style (http://web.mit.edu/nse/events/communicating-science-and-technology.html). Of the whole talk, what I found most memorable was his description of the data structures for the representation of information: as he puts it - 'the web, the chain, the tree'. Our ideas are stored as an interconnected web, and we are often faced with the dilemma of writing them down in a linear structure (one sentence at a time, transitioning between at most two ideas: the previous, and the current sentence). To do this, we arrange our web into a hierarchy of ideas, which then dictates the overall order in which we can express it. In other words, we go from web to hierarchy, and then parse the hierarchy to arrive at a linear structure which we can put down on paper. This is a great way to put it! I've always been faced with the dilemma of wanting to say too many things at once, transitioning from any one idea to ten others, but have always been constrained by the necessary linear ordering of paragraphs. By the time you're done explaining the connection between the first and second ideas, you've lost a possible transition from the first to the third ideas (often requiring a return to the first idea and a new transition). I don't think this linear structure serves as a great representation of the ideas I want to express - I'd much rather hand in my essays in web form...

The Memex

Interesting article: http://en.wikipedia.org/wiki/Memex
Idea put forth by Vannevar Bush in 1945. Precursor to hyperlinking, but detailing a much grander idea of personal paths through collections of information. A user would indicate a linear sequence ("personal trail") for working through information, motivated by personal associations. Supplementing this with personal notes, the result could be stored and shared with others.  

educational websites

Favourite educational websites (in no particular order):

(1) http://videolectures.net/
(2) https://www.coursera.org/
(3) http://www.khanacademy.org/
(4) http://www.ted.com/

LMNN

Distance Metric Learning for Large Margin Nearest Neighbor Classification (Journal of Machine Learning Research, 2009) -- Weinberger, K. and Saul, L.


A very nice paper with a clear overview of distance metric learning, and describing a method which is "the logical counterpart to SVMs in which kNN classification replaces linear classification"

TTC stress experiment

An experiment I would love to run: 

(1) Measure the amount of stress hormone released on average by a single person when experiencing a subway delay for some small fixed amount of time;

(2) Integrate over the length of the delay;

(3) In total, end up approximating the total amount of stress hormones released over the course of a subway delay, by all the angry commuters;

(4) Include statistics about the frequency of occurrence of subway delays.

(5) Calculate total stress hormone secretions (sum over all commuters) caused by subway problems over the course of a year;

(6) Find correlations with health problems and life expectancy;

(7) Present documents to TTC to demonstrate negative impacts of TTC on the lives of Toronto citizens.

drift

Why, when walking with another person, does one continuously drift in the direction of one's companion (to the point of continuous embarrassing collisions)?

why I need blogger

Well, I just realized how much I've been missing (cliche, I know). From now on, blogger will house the websites I like, nice quotations I stumble upon, and random ideas that pop into my head. It won't be my diary, but it WILL be my storehouse of things I want to remember, incorporate into conversations, and otherwise make use of at some point in the future. I now declare that this blog will be an extension of my brain.

mind/brain quote

"Mind as a secretion of the brain." (Ambrose Bierce)

wikihow

I have yet to explore the full usefulness of this website: www.wikihow.com/, but so far, I'm very happy with the step-by-step (with picture) recipes that it contains! In fact: with many things you may want to learn how to do, the approach of 'K simple steps' and guided pictures, is the fool-proof way to do it :)