amle module¶
Automated Machine Learning Environment (AMLE)
This code is a simple shim to help automate running machine learning (ML) tests to reduce effort and make it easier to innovate.
- Requires various packages including YAML:
- sudo apt-get install python-pip git git-flow python-pytest python-yaml
- Requires PIP packages coloredlogs, voluptuous and numpy. Install with:
- pip install coloredlogs voluptuous numpy
Principles (aspirational):
- Generic. Just a shim, does not contain ML code, and tries to not be opinionated about how ML works or data types
- Reproducibility. Run the same test with same inputs and get the same output(s) - or at least statistically similar.
- Reduce experimentation work effort. Support comparative testing across different parameters and/or ML algorithms, retains historical parameters and results
- Add value to experimentation. Support evolutionary genetic approach to configuring algorithm parameters
- Visibility. Make it easy to understand how experiments are running / ran
-
class
amle.
AMLE
(CLI_arguments)¶ Bases:
baseclass.BaseClass
This class provides core methods for an Automated Machine Learning Environment (AMLE)
-
load_aggregator
(agg_name)¶ Passed file location for an aggregator module and return it as an object
-
load_algorithm
(alg_name)¶ Passed file location for an algorithm module and return it as an object
-
run
()¶ Run AMLE
-
run_aggregator
(agg_name, agg_parameters)¶ Run an aggregator, as per spec from policy
-
run_experiment
(experiment_name)¶ Run an experiment, as per spec from policy
-
-
amle.
print_help
()¶ Print out the help instructions