Algorithm Engineer was asked...October 26, 2022

Algorithm Developer was asked...December 14, 2021

↳

Intuitively the answer could be infinity if the first person we draw is the tallest man on earth. But the probability of this event will be small. So just model the distribution P(X) of heights and simulate draws from it. With a bit of math (integrating 1/f(x) gives log(x)) you get that the answer is indeed infinity. Less

↳

Some comment to my answer above; I foolishly forgot that as we are dealing with P(Z<0) and E(Z) = 0, this will be 0.5. Then the sum will be sum from k=1 to infinity of k*(0.5)^k = 2 as this is just equal to 1/2 * (1/(1-(1/2)^2)) using the power expansion of 1/(1-x^2), so the answer is 2, when using standard normals. Less

↳

Because the answer is not a probability, it's the expected number of draws. What's the expected number of times you have to roll a die before you roll a 0? Infinity. What's the expected number of times if you randomly pick the tallest person in the world? Infinity. Even a vanishingly small chance of picking the tallest person in the world still makes the EV infinity. Less

Algorithm Developer was asked...October 13, 2015

↳

You can use Bayes to get the real number, which is close to 0.5 (which you can see intuitively) . They wanted the exact number. Less

↳

@ Feb 4, You formula is correct, looks like typo. because 2^10=1024 so 1/(1+999/1024) ~= 1/2 (1*1/1000)/(1*1/1000+(1/2^10)*999/1000) Less

↳

Bayes. 10^-3/(10^-3 + 1/2^10 * (1-10^-3))

Algorithm Engineer was asked...January 6, 2021

↳

what was your aggregated total score? it shows on the codility report page like 30min after you finish Less

↳

37 aggregated total score

↳

yea codility is such bullcrap, i got all of them right on the surface, but i ended up with 78 agg score. they need to use hackerrank or something, codility just hides too much and we can't even see what we're doing wrong smh Less

Algorithm Developer was asked...November 11, 2015

↳

First algorithm: sort the sequence in O(n log n) time and select the kth element. Second algorithm: construct a max-heap out of the n elements, which can be done in O(n) time and then repeatedly remove the largest element until you reach the kth element. This step takes O(k log n) time and so overall the algorithm runs in O(n + k log n) time. Third algorithm: Use quick select: partition the sequence using a random element as the pivot and recurse on the partition that contains the kth element, which you can determine based on the sizes of the two partitions. This runs in expected time O(n) and runs with high probability in O(n) time. It is also possible to extend this algorithm to run deterministically in O(n) time (see Deterministic Selection algorithm). Less

↳

The third algo is O(n)

↳

we can maintain a min-heap of size k, each time if new element is larger than min element in the heap, remove the root and heapify, otherwise don't update Less

Algorithm Engineer was asked...May 6, 2015

↳

Two integrals along the joint distribution of the the two random variables.

↳

Computing the convolution of the pdf of both random variables.

↳

Hey! Thanks for sharing your experience. I have this interview at the location, can you please tell more about the onsite questions? It will be great if you could kindly tell about the questions focusing on coding(was it all in C?), DSP, image processing and ML. Less

Algorithm Developer was asked...January 16, 2022

↳

1. Not so trivial but idk P(5W 1R | 5W up) = P(5W up | 5W 1R) * P(5W 1R) / P(5W Up) = (1/6 * 6/27) / (1/6 * 6/27 + 1/27 * 1) = 1/2. 2. Helpful to remember that mean of die roll is 3.5 and variance is 35/12 ~= 3. Hence using CLT, the sum of 100 die rolls is approx. N(350, 300). Similarly, for a coin flip the expected number of heads is 1/2 and variance is 1/4, so the number of heads of 600 flips is approx. N(300, 150). So, X - Y ~ N(50, 450), and the probability this is greater than 0 is something like 99%. (sqrt(450) ~ 21, so X - Y would need to be over 2 standard deviations to the left). 3. As others have shown, answer is N(1/2)^(N - 1). Very common quant problem. A variation is: what is the probability 3 points on a circle form a triangle which contains the center? Notice that this is the complement of all 3 points lying on the same semi-circle (i.e. 1/4). Less

↳

1. Conditional probability. Trivial. 2. The mean and variance for X and Y are easy to calculate, approximately Gaussian with CLT. So mean and variance of X-Y is calculable. You can then get a quite good estimation of that probability with pen and paper without help from numerical simulations. 3. Pick any point, the probability of all the remaining n-1 points being on the semi-circle clockwise following this point is 1/2^(n-1). The total probability is the union of all n points being picked as the first point and they are actually mutually exclusive. So n*(1/2)^(n-1). Less

↳

1. 6/7 - There are 7 cubes which have 5 white sides - the center cube of each face and the center of the overall cube. Of those, the cubes on each face each have 1 red side. So, 6/7 2. This problem is on stack overflow and basically needs to be simulated, cannot solve intuitively. 3. n * (.5)^n-1, see stack overflow for explanation. Less

Multitouch Sensor Algorithm Engineer was asked...June 19, 2014

Algorithm Developer was asked...October 16, 2018

Algorithm Developer was asked...October 23, 2016

↳

^ how does this comment help anyone?

↳

Standard question. Just google it. I had seen this question while practicing some questions on the internet before the interview. Less

algorithm developermatlab programmervideo codec engineerwireless communication systems engineerquantitative analyst algorithmic tradingcompiler engineermachine learning engineerimage processing scientistelectrical engineer research and developmentsoftware engineer image processingpython django developerpcb layout engineersimulation software engineer