diff options
Diffstat (limited to 'houdre')
-rw-r--r-- | houdre/01_foundations | 58 |
1 files changed, 58 insertions, 0 deletions
diff --git a/houdre/01_foundations b/houdre/01_foundations new file mode 100644 index 0000000..aedcdd2 --- /dev/null +++ b/houdre/01_foundations @@ -0,0 +1,58 @@ +Probability is the study of randomness. + +Random Experiment: + Ex: Flip a coin, roll a die, results of a Covid test, ... + - It is random because the outcome is uncertain + - Can list all possible outcomes + +Defun 1: The sample space associated with an experiment, Omega, is the +set of possible outcomes of an experiment. + +Ex: + Flip a coin: Omega = {heads, tails} | {heads, tails, sides} + Roll a seven-sided die = {1..7} + The experiment, rigorously, is the # of dots on the top face. + +Cardinality of set is number of elements of set. +Cardinality of empty set is 0. (why is this a convention and not a +self-evident fact?) + +Card S = #S = cardinality of S + + Flip a coin until first tail, and stop. Count # of flips. + Omega = {1,2,3,...} = Natural numbers. (Sometimes also {0,1,2,...}) +Card N = +\infty + +Suppose you flip a coin and want to assess the odds of getting tails on +a flip. For any given coin, we don't know. How can we figure out? +One possibility: Keep flipping coin. # of tails / # of flips is +intuitively close to the probability of getting tails. +T(n) = # of tails / # of flips. +Intuitively, +0 <= Omega(n) = T(n)/n <= 1. +For n large, Omega(n) ~ probability of tails on a single flip. + +The law of large numbers tells us that T(n)/n will converge as +n -> +\infty, to p---the probabiity of getting tails on a single flip. + +"Wisdom of Crowds" somehow obeys the law of large numbers. + +This is the "frequentist" approach for probability theory + +"Subjective approach" to probability theory + - Better at answering question like "what is the odds that Dow Jones + grows by 113 points by EOD?" + +d) Pick a point at random in the unit square and observe its coordinates + What is the probability space? + Omega = [0,1] x [0,1] = [0,1]^2 = {(x,y) : 0<=x<=1 : 0 <=y<=1} + - Could also have excluded the boundary, making it (0,1)^2 + #Omega = +\infty + +Assuming a uniform distribution: +The odds of choosing a point in the bottom left corner triangle is 1/8. +The odds of choosing a circle with a radius of 1/4 is pi/16. +Generally, area/total = odds. +Point area = 0, so odds of choosing it is 0. + +How to formalize the uniform distribution? |