Bradley Arsenault

I would be in favour of just making the precision selectable, including both 64 bit doubles and 16 bit halfs. As someone relatively new to neural networks and Torch, I...

I support this issue. I just posted a duplicate which I am merging into this one. See the code below for an example of an algorithm that does this detection....

Yes I use ESLint in my other projects. Its a good tool. This project has different standards then what I use and I converted everything at the end. Clearly missed...

Hello, I have come back and made a few bug fixes after some testing. But I am having difficulty using the eslint configuration for this project. It seems to be...

I have signed the contributor agreement.

I think what I am intending would be your second suggestion. The ATPE algorithm can be implemented as an additional suggestor - the UI portion is just an extra which...

The biggest difference is that ATPE is not a pure bayesian optimizer. It has a pretrained component which injects some prior knowledge on machine learning models. Therefore, it obtains better...

@maxpumperla Sounds awesome. The ATPE model is the more interesting piece of work and broadly applicable. One note - I will have to introduce two additional dependencies to make the...

@maxpumperla Question: I've also made an implementation of Hyperband which can be wrapped around several different optimizers, including random, tpe and atpe. Hyperband is an optimizer which can take advantage...

@maxpumperla I have created an implementation of the ATPE algorithm into the core Hyperopt library, as pull request #466 What would be the next steps in getting this merged and...