-
很多recursive容易的算法也要考虑iterative方法。因为掌握iterative代表你对问 题结构理解上升了一个高度。e.g. reverse linked list， tree traversal
-
i) top k (kth) elements: heap O(n+klogn), quickselect O(n) average O(n^2) worst, median of medians O(n) worst. cons and pros. Extension: what if all data can not fit into memory. heap size of k O(nlogk) for single machine, many machines see 3. ii) get median in data stream: max heap + min heap
-
kth element in many m machines: binary search, pick a pivot and see how many less and larger among all machines, then discard halves accordingly ( distributed quickselect)
if sorted in single machine: find smallest (k/m)th element among all machines and discard the less partition.
-
stack support O(1) getMin queue support O(1) getMin e.g. k sliding window, most frequently clicked url in past 12 months.
-
reservoir sampling for infinite stream, generate random(1-7) with random( 1-5), card shuffle, string permute in place
-
data structure with O(1) insert, delete, getRandom, get: hashmap + array
LRU data structure: hashmap + doublelikedlist.
binary search tree with rank() (position of inserted or queried data) (add treesize attribute for each node)
-
bit operation and bitset. e.g. find missing number in large data, reverse binary number,
-
java multi-threading, blocking queue, nonblocking queue, H20, philosophy dining, deadlock checking. 现在是个公司都问concurrency，一定要好好准备。
-
OOP: elevator design, parking lot design distributed: large log file design, social network design, distributed cache design ....