Jacob Schreiber

Right. I think the idea is that neural networks are not necessarily the right tool when you already have informative features, like tabular data. Rather, neural networks are good at...

Seems similar to #10 and #11, though a little more specific.

I think that it's important to try out a few good baselines, such as a well tuned linear model and perhaps gradient boosting, but I really don't think it's necessary...

I think that train/validate/test is the more commonly associated name, but I agree that train/tune/test would make it more intuitive what each set is for. I haven't seen many people...

Perhaps the rule here should be that, when discussing these splits, one should make sure to define how you are using each one clearly enough that people from another field...

I like this rule, with its qualifiers. The "be surprised" aspect suggests that you should rigorously validate it before claiming success not that you shouldn't ever believe it. I've found...

My comment wasn't to say that you should be surprised if deep networks outperform humans on tasks that humans do well at, but you should be surprised when deep networks...

Howdy. That should be enough. Read this tutorial for more: https://github.com/jmschrei/pomegranate/blob/master/tutorials/C_Feature_Tutorial_5_Custom_Distributions.ipynb

I'm reluctant to merge a new distribution without adding to the suite of unit tests. Would you mind adding in unit tests similar to those of other distributions?

That does sound like an issue. Unfortunately, I'm no longer actively adding new features to pomegranate, so I don't think I'll be able to fix this. I'd be happy to...