• Activity
  • Votes
  • Comments
  • New
  • All activity
    • Showing only topics in ~science with the tag "mathematics". Back to normal view / Search all groups

    27. A common misconception is that the risk of overfitting increases with the number of parameters in the model. In reality, a single parameter suffices to fit most datasets

      Tweet

      @lopezdeprado: A common misconception is that the risk of overfitting increases with the number of parameters in the model. In reality, a single parameter suffices to fit most datasets: https://t.co/4eOGBIyZl9 Implementation available at: https://t.co/xKikc2m0Yf

      5 votes

    49. How do you compute the probability of covering an entire population given you take an arbitrary number of random samples?

      Ask
      I suck at probability, so I thought I would ask here. To clarify, given a population of size P, a sample size of K, and an arbitrary number of trials N, how do I compute the probability of having...

      I suck at probability, so I thought I would ask here.

      To clarify, given a population of size P, a sample size of K, and an arbitrary number of trials N, how do I compute the probability of having included each member of the population at least once in the experiment?

      This problem is difficult to wrap my head around. It seems like it uses a combination of combinatorics and dependent events, which really throws me off.

      Edit: This problem isn't the coupon collector's problem (please see some of my responses below). Think of the coupon collector's problem as being a special case of this problem where K = 1. My question is meant to cover an arbitrary K >= 1.

      9 votes