questions/notes from other repo

[ronboger]

• last time i wrote "notation: k cannot be number of modes and number of parameters"
• read rice chapter 1
• get better with ipython
• it's just sufficient, not necessary because definition of concavity in discrete domain is limiting. consider f(x) = {5, 4, 2, 4, 5}, which doesn't satisfy the concavity definition but does satisfy having a local minimum having an absolute extrema
• how to properly define dimensionality of x, theta. need further guidance
• BIC for single Gaussian is $$nln(RSS/n) + kln(n)$$
• if set of points that have likelihood of inf have measure 0, look more deeply on measure theory
• should i prove why it's not necessary?
• perhaps it is also better to consider proving the direct case instead of switching signs and thus concavity later on. the sufficient condition definition holds, make notation on this easily understood