2023-05-07 07:38:06 -07:00
|
|
|
# thursday
|
|
|
|
|
|
|
|
|
|
what comes after [humpday?](https://github.com/microprediction/humpday)
|
2023-05-04 08:24:04 -07:00
|
|
|
|
2023-05-04 15:52:08 -07:00
|
|
|
**note:** i have a personal bias against bayesian optimizers.
|
|
|
|
|
these results may not represent them fairly (or at all).
|
|
|
|
|
|
2023-05-07 07:38:06 -07:00
|
|
|
## usage
|
|
|
|
|
|
2023-05-04 08:24:04 -07:00
|
|
|
```
|
2023-05-07 05:35:37 -07:00
|
|
|
python3 -m thursday
|
2023-05-04 08:24:04 -07:00
|
|
|
```
|
2023-05-07 07:38:06 -07:00
|
|
|
|
2023-05-07 07:42:57 -07:00
|
|
|
## prerequisites
|
2023-05-07 07:38:06 -07:00
|
|
|
|
|
|
|
|
* Python 3.10 or greater
|
2023-05-07 07:52:39 -07:00
|
|
|
* numpy 1.23
|
2023-05-07 07:38:06 -07:00
|
|
|
* *TODO*
|
|
|
|
|
|
|
|
|
|
## structure
|
|
|
|
|
|
|
|
|
|
* `thursday/candidates` contains wrappers around optimizers to minimize objective functions on the unit cube
|
|
|
|
|
* `thursday/external` contains code written by other authors (note their licenses!)
|
|
|
|
|
* `thursday/internal` contains original implementations of general-purpose optimizers
|
|
|
|
|
* `thursday/solutions` is responsible for generating `thursday/go_benchmark_lists.py`
|
|
|
|
|
* `thursday/utilities` contains no surprises
|
|
|
|
|
* `thursday/` contains everything else (for now)
|
