Prove it - Ep2: Another random walk

Поделиться
HTML-код
  • Опубликовано: 13 янв 2025

Комментарии • 21

  • @vicente3j
    @vicente3j 6 месяцев назад +47

    optiver i watched the whole video can i get an internship

  • @billyb5225
    @billyb5225 5 месяцев назад +7

    Exercise 2: You can derive a recurrence relation for the sequence {a_k} on by assuming the formula E_0n = a_k + E_kn. In the fair coin example, a_k = k^2. In this case, the prob of going back is 2/3 and forward is 1/3.
    Applying the EV recurrence to house k you get E_kn = 2/3 * [E_k-1n + 1] + 1/3 * [E_k+1n + 1] and by E_kn = E_0n - a_k, then rearranging for E_0n you get
    E_0n = 3 - 2a_k-1 + 3a_k + E_k+1 = a_k+1 + E_k+1, netting a_k+1 = 3 - 2a_k - 1 + 3a_k or a_k = 3a_k-1 - 2a_k-2 + 3
    Solving this, you need to assume a_k is a linear combination of A*r^k where r is a root of the characteristic polynomial. Substituting this gets the characteristic polynomial r^2 - 3r+2 which has roots 1 and 2. But since the basis term 1^k i.e. constant is already present in the recurrence relation, we need to add a k term. So we assume a_k is in the form a_k = A*2^k + B + Ck.
    You can derive that a_1 = 1, a_2 = 6, a_3 = 19, and solving for A, B, C in 3 simultaneous eqns gets the final answer of a_k = 4*2^k - 3k - 4.
    In the end the expected number of blocks to go from house 0 to n is E_0n = 2^[n+2] - 3n - 4.

    • @linyidai9076
      @linyidai9076 4 месяца назад

      I put it in OEIS. thanks for letting me know "charasteristic polynominal"

    • @billyb5225
      @billyb5225 4 месяца назад

      @@linyidai9076 would you mind sending the link to it :)

    • @billyb5225
      @billyb5225 4 месяца назад

      @@linyidai9076 are you able to share the link to the contribution pls :)

  • @ccolombe
    @ccolombe Месяц назад

    These are great! More please :)

  • @vahidnaghshin6455
    @vahidnaghshin6455 5 месяцев назад +2

    We're still waiting for this month prove it. Any update?

  • @infinity624
    @infinity624 6 месяцев назад +10

    Guys this is a very easy problem no need to solve equations or do anything just give me like and pray I get into Optiver. Here is the one liner, instead of the reflection think your friends' house is on -4 and +4 u start from 0. u stop when u reach 4 or -4. u need expected stopping time of a symmetric random walk. That is 4*4= 16

    • @infinity624
      @infinity624 6 месяцев назад +3

      for those who aren't aware of this result E(S_n^2-n)=E(S_0^2-0) since that quantity is a martingale where S_n = X1+...+Xk. and E(S_n)=0 from there u get S_n^2 = (4^2+4^2)/2=N.

    • @madhuragrawal5685
      @madhuragrawal5685 5 месяцев назад

      Wow that's very clever! Very cool ​@@infinity624

    • @Antoine.Comeau.Racing
      @Antoine.Comeau.Racing 3 дня назад

      Quite right!

  • @sagnikbiswas3268
    @sagnikbiswas3268 6 месяцев назад +5

    for k, it’s (k-1)^2. So it’s 16. Very textbook problem

  • @janisaiad9505
    @janisaiad9505 6 месяцев назад +2

    everyone that manipulated random walks for once should get a good answer, wrong people will use markov chains and absorbing states and it's a big red flag to do so in an interview

    • @peizhengwang
      @peizhengwang 6 месяцев назад +1

      Hi. May I ask how to solve this question not using markov chain? The first thing came to my mind is using martingale. But I feel like this is not a martingale because it has an upward trend. Dont know how to solve it. 😢

    • @gqp3185
      @gqp3185 5 месяцев назад

      fuck me,you read me like a book,i just used the MC

  • @entropekomiko
    @entropekomiko 4 месяца назад

    I used gambler's ruin time + wald's equation😅

  • @jasonzhou3314
    @jasonzhou3314 5 месяцев назад

    I think there is a mistake the +1 should be on the other side. E(X_1) + 1 = 1/2E(X_0) + 1/2E(X_2) OPTIVER I THOUGHT YOU WERE a good firm!!