Hiroaki Mikami

Thank you for reporting the issue. I think [the StackOverflow post](https://stackoverflow.com/questions/44217316/how-do-i-use-stdoptional-in-c) you mentioned is related. I can't reproduce the build error because I don't have OS X machine. But I...

I'm sorry about my mistake in `enumerator-test.cc` (I forgot to add `#include `). The latest master branch fixed this issue. And I tested that it can be compiled with g++-6.3.0...

Now I cannot check the gtest version because the AWS instance was already terminated. But I fixed these errors by adding `static_cast(...)` to `EXPECT_TRUE` and `EXPECT_FALSE` macros. Would you use...

I converted the AST to a sequence of rules at https://github.com/HiroakiMikami/mlprogram/blob/master/mlprogram/actions/action_sequence.py#L310 by using DFS.

### Notes: The DNN model used in the "small" setting may too big, and the model can synthesize the program without reinforcement learning.

Sorry for the late reply. I do not value paper reproduction (such as TreeGen) now for 2 reasons. The first reason is that using a large model with NLP technique...

> Do you mean that all models perform relatively similar across all data sets? That's almost right. I think all models show a similar performance if the used computational resource...