• Paper reading slowly – learning triggers for heterogeneous treatment effects


    This paper is based on the recursive partitioning for heterogeneous casual effects, adding two new elements: Trigger: select personalized thresholds for different groups of treatments. E.g coupon strength, red envelope amount New node penalty: to enhance model generalization paper C. Tran and E. Zheleva, “Learning triggers for heterogeneous treatment effects,” in Proceedings of the AAAI […]

  • Flink knowledge map


    Noun concept Streams unbounded stream Infinite data flow bounded stream Limited data flow Time (time) Time is an important basis for us to judge whether the business status is lagging behind and whether the data processing is timely. There are three time types. Event time Event time, when the event occurred on its device. Data […]

  • HashMap source code — JDK1.8


    Important attributes The default capacity is 16 /** * The default initial capacity – MUST be a power of two. *N power of 2 is recommended */ static final int DEFAULT_INITIAL_CAPACITY = 1 << 4; // aka 16 Default load factor /** * The load factor used when none specified in constructor. */ static final […]

  • Query table space utilization


    Query tablespace utilization — consider self growth, where utilization considers tablespace utilization relative to threshold select a.tablespace_name, total || ‘M’, free+shortNum|| ‘M’ free, total-free-shortNum || ‘M’ as used,Round((free+shortNum)/total * 100, 3) as “FREE%”,Round((total – free-shortNum)/total * 100, 3) as “USED%” from (select tablespace_name,autoextensible,case when t.autoextensible=’YES’ then (maxbytes-bytes)/1024/1024 else 0 end shortNum,case when t.autoextensible=’YES’ then sum(maxbytes)/1024/1024 […]