This folder contains implementation of Edna data analysis in Python.
Normally you don't need to call this code directly, it is supposed to be called from edna-server, but sometimes you may want to test specifically analysis part of Edna.
We use poetry for dependency management, though you can install dependencies using your
favorite way (e. g. your system package manager) and not use poetry.
To get shell with all dependencies installed run poetry install and then poetry shell from this directory.
The list of features implemented in an analytic module:
- Compute parameters of fitted sigmoid.
- Detect potential outliers.
- Compute parameters of fitted sigmoid without outliers.
JSON containing list of experiments in the form:
[{"experiment": int,
"data": [[float, float], ...],
"find_outliers": boolean},
...]
It's expected to be passed to the standard input (stdin).
JSON containing list of calculated values per each experiment:
[{"experiment": int,
"params": [float, float, float, float],
"outliers": [int, ...],
"new_params": [float, float, float, float],
"status": str},
...]
experiment and status are always present in the output.
If "status" is not DONE, other fields are omitted. If find_outliers is False, outliers are not calculated.
If there are no outliers, new_params is omitted. Status may contain error descriptions.
ERROR: Empty data field.No data points have been passed.ERROR: Wrong experiment format.The experiment from .json provided is in the wrong form.ERROR: Too little data.The least amount of data points to fit sigmoid is 4.ERROR: Can not fit sigmoid.Impossible to fit sigmoid to this particular set of data points.