Monday, March 9, 2009

that has never happened before i swear

blah blah blah i've written like tons of stuff in the last week to post here, and it's all been crap to my eyes. felt like busy work. thanx.

i'll write when i feel like it. suppose i can dump my stuff here 'til then.

the value of illusion
genre mixing (with regards to darla's world)
giant robots in general
i hate vampire kids

Wednesday, March 4, 2009

some latex constructs

here are some common constructs i use that i find myself frequently looking up. the first is the parallel symbol used in the Kullback-Leibler divergence. just using a pair of vertical bars squeezes the arguments together, but the \parallel puts the necessary space before and after the bars. below is the definition of skewed Kullback-Leibler divergence.


D_{SKL}(p,q) = D_{KL}\left(p \parallel \gamma p + (1-\gamma)q\right)


a similar problem occurs with conditional probabilities. \mid takes care of that. the next example uses Bayes' theorem, with a little \frac business for flavor.


P(A \mid B)=\frac{P(B \mid A)P(A)}{P(B)}


i use an old homework to show how to used the cases environment. each line can have only 1 &, which separating the result and the condition. not that the "amsmath" package needs to be loaded, or else you'll get funky "Misplaced alignment tab character &" errors. i use the \text command to get normal script in the math environment. (in the homework, i was referring to Prototype-driven grammar induction, by Haghighi and Klein)


P(p_{ij} \mid X) =
\begin{cases}
0.6 & \text{if $p_{ij}=X$}\\
\text{uniform} & \text{otherwise}
\end{cases}


that's it for now. if any latex stuff comes up, and i think it's interesting (or that i might need it for future reference), i'll be posting.