1

The below is a picture which denotes the error of an ensemble classifier. Can someone help me understand the notation

enter image description here

What does it mean to have (25 and i) in brackets and what is ε^1 is it error of first classifier or the error rate raised to power i. Can someone explain this formulae.

Arun Jose
  • 125
  • 4
  • 2
    ${25 \choose i}$ is a [binomial coefficient](https://en.wikipedia.org/wiki/Binomial_coefficient). The whole expression is the calculation of bernoulli distribution – Nikos M. May 22 '22 at 09:57

1 Answers1

2

$\varepsilon^i$ is the error rate raised to the power i. So for each value i, the formula calculates the probability of i classifiers classifying a sample incorrectly, so for i=13 we have: $$e_{13\ wrong} = {25 \choose 13} \times \varepsilon^{13} \times {(1-\varepsilon)}^{12}$$ Assuming $\varepsilon = 35\%$, and calculating the binomial coefficient gives us: $$e_{13\ wrong} = 5,200,300 \times 0.35^{13} \times 0.65^{12} = 0.035$$ Repeat this for $i = 14, 15, ... , 25$, then sum all the results to get the final answer.

Lynn
  • 1,121
  • 1
  • 3
  • 18