Remember the first time Baidu telephone interview

From two o'clock in the afternoon to three today, I had my first telephone interview in my life. Last 59 minutes, pure technical side.

Straight into the business, began the algorithm investigation. A total of two, about 40 minutes, the first in the hint reluctantly answered, the second is really not answered.

Then there are some issues related to the project, especially the information extraction module, and then asked me some special circumstances of some ideas and so on, alas ... No idea what they're asking.

Before a little preparation is not, still busy things on the project, originally some algorithms do not understand, this more tragic ...

Here are the details:

The first algorithm problem: There is an array, set it as N1, how to remove the N2 large number of before. And then with the qualification, suppose the numbers are between 0-10000.

began to see the topic, thought of heap sort. Because before you have seen a very large array, take the first few large numbers, and use the smallest heap. Take the first n small number of words, with the largest heap. The results asked me to implement the process, no impression--. Forget the biggest and smallest pile of things. I'm going to open this up and keep looking.

The interviewer says to talk about your ideas, no matter how they do it. I say you can think of a simple sort first, and then take the N2 before the OK. He asked the complexity of how much (he likes to ask complexity, if you say the idea can be achieved, ask you the complexity, the whole down, I have to calculate about 10 or so complexity), I said the simple bubble sort of words, should be n side. He said there is no easy way, I say the quick sort of words, should be nlogn. And then let me talk about the following quick sort, which seems barely clear--.

There was no further progress, the interviewer suggested that the condition was specific and the numbers were 0-10000. You see can count 10000 how many, 9999 how many and so on. And then...... Very glorious, I ran away again (--. I really do not know how to think, are prompted to this extent) I said can binary find, 5000-10000 of how many, then not enough, 2500-5000 how much. Ask the final complexity of how much, I calculate, looks like or Nlogn. - -。

Then after a thought at least to give the answer, is the end of the first question.

When I come back to discuss with my classmates, I generally think that we can sort by cardinal number ... The answer to the interviewer's tip is similar.

The second algorithm: There are two ordered array N1 and N2, how to find the number of digits.

With the first lesson, first from the simple come, say you can merge the sort first, and then take the median number. The interviewer asks, what is the complexity of the order of the two ordered arrays? I once again gorgeous 2 B. began to say min{n1,n2} it, and then after thinking, the affirmative answer is MAX{N1,N2}. Then the interviewer asks why, then various explanations. Finally, the interviewer said, whether it is max{n1,n2} or N1+N2, we continue to look down. I have a moment to smoke their own impulse, ni, is n1+n2 ah--.

To continue to look down means that there is no easy way. I have no idea what to do with my bile. (Why are you so stupid?) Then, the interviewer again prompts: If you give you a number of median size, how to insert it. An instant thought of the binary insertion, respectively, from the middle look up. Then ask the complexity, answer Yue Logn1+logn2. Then it goes on, which is basically a hint of the binary method. But for a long, long thought, there is no final solution. Alas, reluctantly into the next link--.

Project Experience Review:

First: Metasearch, let me speak briefly. Then asked me which part I was responsible for, and I said I did it all. Looks like the other people are not very interested, proceed to the following one.

The second is the information extraction aspect, detailed asked me the realization thought. Then raised a number of questions, and finally for some video sites, the text is particularly small, there is no train of thought? I am again the bile, said that for large sites can do based on rules of extraction. The interview brother asked, that some of the site of the minority, or a little thought--.

The next time is about the same, the following text classification, Hadoop, MPI did not ask the end, and then asked me if there is anything to ask him. I asked him to combine this interview performance, give a little advice God horse's. He replied that the basic algorithm and data structure to understand in-depth, can not stay on the surface, to think more, have their own ideas to do.

Well, my first time was so gorgeous that it ended in a miserable--.