Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Fizz Buzz in Tensorflow

joelgrus.com

Post reply on HN
Print view View original

41–50 of 85 posts

Re: Fizz Buzz in Tensorflow

#41

Post by CodeMage »

couldwejust wrote: post #37
th0ma5 wrote: post #29

Earlier quoted context omitted.

FWIW, please be mindful of your biases. I recently had an interviewer with a major bank conglomerate who had some silly puzzle. I outlined a system for solving it in algorithmic notation, and the interviewer had no idea that such a response was appropriate and said sorry, that they were looking for someone with programming experience, not specific, just "many languages" ... Not saying you do this, but what if someone…

> what if someone provided Fizz Buzz in COQ or Elixir Wie waere es fuer dich, wenn ich in einer Sprache antworte, die du wahrscheinlich nicht verstehst?

Zavisi. Ako si mi rekao da mogu da koristim bilo koji jezik, onda je na tebi da pronađeš prevod ;)

Re: Fizz Buzz in Tensorflow

#43

Post by Bahamut »

gnaritas wrote: post #23
LaurenceW1 wrote: post #3

Earlier quoted context omitted.

I would have given him the job

If that were a real situation, I wouldn't have, he over-engineered the shit out of it and while cool, I prefer devs who can solve simple problems simply; they get more work done for the price.

Well, when you give a hilariously bad interview question...

Re: Fizz Buzz in Tensorflow

#44

Post by jedberg »

spc476 wrote: post #40
jedberg wrote: post #38

Earlier quoted context omitted.

I hate puzzles too, especially the gotcha ones with esoteric knowledge required. But for FizzBuzz I explain the problem, explain that there is no trick and the simplest answer that produces the correct output will be fine. To solve it you need to know about building a loop, what the mod operator does, and maybe keeping state depending on how you build it. I tell them to write it in the language they know best. None o…

You don't need mod, loops or recursion to solve it. For instance: local remove = table.remove local insert = table.insert local print = print local sequence = { false , false , 'Fizz' , false , 'Buzz', 'Fizz' , false , false , 'Fizz' , 'Buzz' , false , 'Fizz' , false , false , 'FizzBuzz' } local function o(v) local name = remove(sequence,1) insert(sequence,name) print(name or v) return v + 1 end local function t(v) r…

You're right, it doesn't (clever solution BTW).

Although we could debate that the way you call o and t is just you manually expanding a loop. In which case I would question why you didn't just use a loop. :)

I could also debate with you that you had to at least understand what mod is to build the sequence.

I would also give you bonus points for detecting that it is a repeating pattern.

But the point is, you solved it, you wrote code that could solve it. So now I know that you're at least capable of writing basic code, and we've got a couple of things to talk about!

Re: Fizz Buzz in Tensorflow

#45

Post by vonklaus »

gnaritas wrote: post #23
LaurenceW1 wrote: post #3

Earlier quoted context omitted.

I would have given him the job

If that were a real situation, I wouldn't have, he over-engineered the shit out of it and while cool, I prefer devs who can solve simple problems simply; they get more work done for the price.

Sometime it is better to hire for culture, someone who is intelligent but also, and this is critical, can take a joke.

Re: Fizz Buzz in Tensorflow

#46

Post by jjallen »

CodeMage wrote: post #41
couldwejust wrote: post #37

Earlier quoted context omitted.

> what if someone provided Fizz Buzz in COQ or Elixir Wie waere es fuer dich, wenn ich in einer Sprache antworte, die du wahrscheinlich nicht verstehst?

Zavisi. Ako si mi rekao da mogu da koristim bilo koji jezik, onda je na tebi da pronađeš prevod ;)

What are the odds my wife's language would show up in an HN comment?

Re: Fizz Buzz in Tensorflow

#48

Post by xyience »

jedberg wrote: post #44
spc476 wrote: post #40

Earlier quoted context omitted.

You don't need mod, loops or recursion to solve it. For instance: local remove = table.remove local insert = table.insert local print = print local sequence = { false , false , 'Fizz' , false , 'Buzz', 'Fizz' , false , false , 'Fizz' , 'Buzz' , false , 'Fizz' , false , false , 'FizzBuzz' } local function o(v) local name = remove(sequence,1) insert(sequence,name) print(name or v) return v + 1 end local function t(v) r…

You're right, it doesn't (clever solution BTW). Although we could debate that the way you call o and t is just you manually expanding a loop. In which case I would question why you didn't just use a loop. :) I could also debate with you that you had to at least understand what mod is to build the sequence. I would also give you bonus points for detecting that it is a repeating pattern. But the point is, you solved it…

My team isn't exactly doing "top-tier CS work", so if someone "showed off" how far above the average they are with a Coq solution or the quirky GP you replied to or the OP's neural net I'd be delighted and want to skip the rest of the interview, recommending an immediate hire (or at least a go-to-next-stage since I only give these questions over phone screens). The question is beneath them and they have demonstrated that, our team's work is also probably beneath them. We just use that question to filter out the totally unqualified. Their potential value is high because they are overqualified -- they didn't just pass it, they passed it with style. Most dev work at large companies is dull and doesn't require much knowledge that e.g. a CS degree supposedly gives, which is part of why it's amusing/frustrating we have such a bad whiteboard hazing culture insisting candidates can invert a tree or whatever.

I head the "what if they do it this other weird way?" question off by giving them a simple problem, having them solve it (or at least convince me their solution solves it), and then asking them to solve it a different way. My screens mostly follow the ideas from https://sites.google.com/site/steveyegge2/five-essential-pho... I allocate about 10 minutes max for "basic coding" where I just ask them (lately) to write a function that returns true if the given input int (which is assumed to be > 0) is even and false if odd. Then I ask them to do it in different ways. If they use bit-and, I figure they know enough about the "bits and bytes" section too for my team's purposes. Last time someone couldn't recall using bit-and but surprised me with a way I didn't consider, which was convert to string and check the last character. I'm sure there are other ways to do it I haven't considered either. Maybe someone will give me a solution with tensorflow someday, but what would be really impressive is if they bust that out within a few minutes. :) All I'm really after is "can you write code?", if the answer is no then very little time was wasted, if the answer was yes we can go into variations and other questions to rank the other yeses / find other red flags.

Re: Fizz Buzz in Tensorflow

#49

Post by aab0 »

brandmeyer wrote: post #8
mattmcknight wrote: post #4

Maybe the one-hot binary encoding wasn't the best feature set. Base 3 and Base 5 encodings might have worked better.

Stupid question: We are trying to train the network how to calculate modulus of division. Of a binary encoding. Why should we expect this to be learnable in only two layers?

Technically, it's not two layers. It's only two layers of parameters. A RNN runs the two layers repeatedly at each sequential input. So if my input/output is like '1000~>fizz', it can be seen as a feedforward NN with at least 8 layers. (This is how it's possible to train a RNN in the first place: it gets 'unrolled' over time and turned into a big NN.)
Post reply on HN