LinkedIn Feed: Less JVM garbage collection and faster speed
1. ArrayList uses get(i) for simple random access, saving iterator overhead
2. Initializing the HashMap with an explicit size can avoid the overhead of resizing. The initial capacity uses the size of the input array divided by the load factor 0.7
3. Delayed expression evaluation
4. Pre-compiling the regular expression pattern will save every time CPU and memory consumed by compilation
5. As many caches as possible: But we use Guava cache
6. The Interns method of String is very useful, but It's also dangerous
http://www.4byte.cn/learning/84939/linkedin-feed-geng-shao-de-jvm-la-ji-shou-ji-geng-kuai-de-su-du.html
machine learning
http://blog.csdn.net/han_xiaoyang/article/details/49797143