Law of large numbers

• Ensures that empirical averages computed from large samples are good estimates of the population mean Central limit theorem
• Enables the assumption of normally distributed data from a large sample even if the population is not normally distributed

Normal Distribution (Gaussian)

• A symmetrical bell shaped distribution where 68% of data fall within 1 standard deviation, 95% fall within 2 standard deviations and 99.7% fall within 3 standard deviations
• The PDF is given by $f(x|\mu, \sigma)=\frac{1}{\sqrt{2\pi\sigma^2}}e^{-\frac{(x-\mu)^2}{2\sigma^2}}$
  • $e$ is the natural exponent (Euler's number)
• A special case is the standard normal (also called z-distribution)
  • where the mean is 0 and standard deviation is 1, so is variance
  • denoted as $N(0, 1)$
• z-score is a measure that describes a value’s relative position to the mean of a group of values, or how many standard deviation away from the mean it is
  • $Z=\frac{x-\mu}{\sigma}$
• Key properties
  • a linear combination of normally distributed variables are also normally distributed
  • Central limit theorem
  • transformation - non-normal distributions can be transformed to follow normal using log transform, square root or box-cox transformation -> Feature Transformation Techniques
• Applications
  • Many ML and statistical inference techniques assume the data follows a normal distribution

Binomial Distributions

• A discrete probability distribution which models the number of successful outcome of independent Bernoulli trails, often used to model situations where there are exactly 2 outcomes
  • The distribution gives the probability of having $k$ successes out of $N$ trails given a success probability of $p$ and failure of $q=1-p$
  • calculated by $P(X=k)=\binom{N}{k}(p^k)(1-p)^{(n-k)}=\binom{N}{k}(p^k)(q^{(n-k)})$
    • $\binom{N}{k}$ is the binomial coefficient representing the number of ways of choosing $k$ out of $N$ trails calculated by $\frac{N!}{k!(n-k)!}$
    • factorials are the product of all positive integers equal and less than the number
    • factorial simplifications
      • 15!/3! = 5!
• The properties of the distribution are: mean ($\mu=Np$), variance $(\sigma^2=np(1-p))$, and standard deviation ($\sigma=\sqrt{np(1-p)}$)

Poisson Distribution

• A discrete probability distribution which models the number of times something will happen within a period
• Assumptions
  • The occurrence rate is a known constant ($\lambda$)
  • The events occur independently from each other
• Gives the probability of observing exactly $k$ events in interval
  • calculated by $P(X=k)=\frac{e^{-\lambda}\lambda^k}{k!}$
• Applications
  • Queueing theory to model the number of customers arriving
  • Epidemiology to model the number of rare diseases or occurrence in a population
  • Manufacturing to estimate the number of defects in a batch of products
• Relation to other distributions
  • The poisson distribution is a limiting case of the binomial distribution where $n$ is large and $p$ is small and $np=\lambda$ remains constant
  • The inter-event time follows an exponential distribution with parameter $\lambda$