张乐博士的工具包mexent 笔记

张乐博士最大熵工具包：下载地址 点击打开链接

我自己数据放进去，迭代20轮，程序处理完后输出：
Total 1063226 training events and 0 heldout events added in 2.63 s
Reducing events (cutoff is 1)...
Reduced to 935244 training events

最后程序报错：

std::runtime_error caught:lbfgs routine stops with an error

 IFLAG= -1
  LINE SEARCH FAILED. SEE DOCUMENTATION OF ROUTINE MCSRCH
ERROR RETURN OF LINE SEARCH: INFO=  6
  POSSIBLE CAUSES: FUNCTION OR GRADIENT ARE INCORRECT
  OR INCORRECT TOLERANCES

关于std::runtime_error caught:lbfgs routine stops with an error工具发布首页的解释是：

• This is a rare event and most users do not encounter this problem. However, when this does happen it is usually an indication of "over-training", which can be caused by either using a very small dataset, or trying to perform too many training iterations on a large dataset. The ad-hoc solution is to use a larger dataset, or simply use fewer training iterations. Future releases will probably turn this bug into a useful feature:-)

  
  

我的数据集合还是太小了？Really？