Sd6.20 summary of intensive training

ok Today is not very sleepy Have a good start

I seem to have found something , I doubt whether my sleepiness is determined by the routine of today's question （ Whether the automatic body test is a routine If you don't follow the routine, help me release my sleepiness ）

blame

Time arrangement

8.10-10.00

Look at the question I feel that everything is quite done obviously T1 log about T2 It's a moment product （n^3 It feels like ) T3 It should be a block （1G）

And then he pushed T1 Thought of two points + Line tree approach Feel like running log^2 It was a little dangerous, so I thought for a while But I wrote it directly

Then a test 4s dangerous

10.00-10.20

Considering that there is no repair rmq direct st Just make the watch Less seconds 1log

So I changed wildly After that, I will speed up my reading In the end 0.6s 了

10.20-11.00

Thought about it for a while T2 T3,T2 Be able to do shape pressing ,T3 Thought about it for a while Feeling nq But I have no idea

11.00-12.05

Write the shape pressure After running the data, I found that it was impossible 20 And suddenly found k=5 I didn't open the array correctly in that file .

Suddenly a daze （ The calculator is wrong ） think k=5 That one should be written separately The more you write, the more strange it becomes （ The discovery complexity is the same ） Then I did it by hand I found it unnecessary to drive alone .

Wasted my precious time （ anger

12.05-12.50

Thought about it for a while T3 The reaction can be double pointer Get a bucket cnt Just maintain the double pointer （ Move the pointer when it is illegal The length does not increase ）（30

then nq You've gone too far 50）

Topic analysis

T1

Consider for $a[i]$ Find the latest successor $x-a[i]$

Then it divides the controllable range

And to avoid double counting the answers You need to pay attention to the two minutes

T2

Consider the pressure dp $f[s,x]$ The button set used for the expression is $s$ At the moment $x$ Number of alternatives be aware Because the big ones can only recover the small ones Finally, the big ones must not be recovered , So the answer must be limited , The state must also grow from small to large

then dp！

T3

Consider double pointers And then if $r-1$ after The answer has become illegal So let's $l++,r++$ ( Only when the interval becomes smaller can it become a better answer ）

---

I feel that today's performance is quite regular Play normal （ even to the extent that T3 It's a little beyond my imagination ）

But if it's better Maybe hope T3 It is more practical to do more （ The same is true of the list ）