Hacker News Reader

Comment by ojr | Hacker News Reader

I just did a phone screen with Meta, and the interviewer asked for Euclidean distance between two points; they definitely have some nerds in the building.

View on HN

That's like 8th grade math, what am misunderstanding about your comment?

E: wasn't the only one.

K closest points using Euclidean distance and a heap, is not 8th grade math, although any 8th grade math problem can be transformed into a difficult "adult" question. Sums are elementary, asking to find a window of prefix sums that add up to something is still addition, but a little more tricky

People saying it is a high school maths problem! I'd like to see you provide a general method for accurately measuring the distance between two arbitrary points in space...

Using a heap in 10 minutes, the Euclidean distance formula was given and had to be used in the answer; maybe they thought that was the question?

Compare to the speed of light, c at your two reference frames

They actually need to know so that they can train Llama.

I suppose the trick is to have an ipad running GPT-voice-mode off to the side, next to your monitor. Instruct it to answer every question it overhears. This way you'll ace all of the "humiliation ritual" questions.

there's a youtube channel made by a meta engineer, he said to memorize the top 75 LeetCode Meta questions and approaches. He doesn't say fluff like "recognize patterns. My interviewer was 3.88/4 GPA masters Comp Sci guy from Penn, I asked for feedback he said always be studying its useful if you want a career...

it wasn't just euclidean distance of course, it was this leetcode problem k closest points to origin https://leetcode.com/problems/k-closest-points-to-origin/des..., I thought if I needed a heap I would have to implement it myself didn't know I can use a library

I.e. the nearest neighbor problem. Presumably seeing if the candidate gave a naive solution and was able to optimize or find a more ideal solution

its not a nearest neighbor problem that is incorrect, they expect candidates to have the heap solution on the first go, you have 10-15 minutes to answer, no time to optimize, cheaters get blacklisted, welcome to the new reality

Finding the k points closest to the origin (or any other point) is obviously the k-nearest neighbors problem. What algorithm and data structure you use does not change that.

https://en.wikipedia.org/wiki/Nearest_neighbor_search

edit: If you want to use a heap, the general solution is to define an appropriate cost function; e.g., the p-norm distance to a reference point. Use a union type with the distance (for the heap's comparisons) and the point itself.

true, I am thinking, Node and neighbors, this is a heap problem, it actually does matter what algorithm you use, I learn that the hard way today, trying to implement quickselect vs using a heap library (I didn't know you could do that) is much easier, don't make the same mistake!

Wish it was more how you think than requiring boolean correct/incorrect answer on the whiteboard after 15min.

I don't get the joke.

no joke, stop pretending like you know the answer to every LeetCode question that utilizes Euclidean distance

That's basic high school math problem.

The foundation, like every LeetCode problem, is a basic high school math problem, when the foundation of the problem is trigonometry, way harder than stack, arrays, linked list, bfs, dfs...

Reply on news.ycombinator.com