Frequently Asked Questions

1. What does juliet stands for?

The name of the program, juliet, is not an accronym. It is a name one of us simply could not get out of their mind during the time it was being written :-).

2. juliet is taking too long to fit my dataset!

We suggest you check the number of dimensions/datapoints. Typically, for dimensions less than about 20, the algorithm shouldn't take long unless you are working of tens of thousands of datapoints, and fitting highly complex models (e.g., several non-celerite GPs). In this latter problem, we suggest you use one of the celerite GPs provided by juliet, as they can largely speed-up the computing time (from days to hours in cases of thousands of datapoints). For larger dimensions, consider using dynesty (see here) and/or multi-threading.

In general, MultiNest gets inefficient for dimensions larger than about 20, and then dynesty is the way to go. Also make sure to install the version of dynesty from the project's Github and not the pip install version, as the latter is not updated. Finally, take a look at the dynesty FAQs webpage, which sheds light on the possible configurations for the problem at hand. This issue might also be enlightening for understanding what methods, number of live points, etc. to use.

3. Can you implement X in juliet?

If you believe that implementing X to juliet will benefit the community, we welcome suggestions to implement it, which of course will be subject to time-constraints by us. If you are willing to work on this implementation yourself, however, we would be happy to have a small meeting with you so we can help you out with understanding the code structure of juliet. If you have a problem which needs a very specific input to juliet, or is a very specific problem, we would most likely decline the offer of implementing it to juliet --- so make your case! In any case, you are always welcome to work on adding this to juliet!