From 732cba3f4335d3c28ce50e2abdb031b96adeb0fa Mon Sep 17 00:00:00 2001 From: Starofall Date: Tue, 17 Jan 2017 23:39:15 +0100 Subject: [PATCH] Enhanced documentation --- experiment-specification/definition.py | 21 ++++++++++++++++++--- 1 file changed, 18 insertions(+), 3 deletions(-) diff --git a/experiment-specification/definition.py b/experiment-specification/definition.py index 9a98062..4841a4c 100644 --- a/experiment-specification/definition.py +++ b/experiment-specification/definition.py @@ -5,10 +5,12 @@ system = { # Defines how to run experiments - # "sequential" -> Runs a list of experiments in a sequential way - # requires a "experiments_seq" array + # "sequential" -> Runs a list of experiments in a sequential way + # requires a "experiments_seq" array in the definition.py # "self_optimizer" -> Runs a self adaptation algorithm to optimize values - # requires a "self_optimizer" object in the definition + # requires a "self_optimizer" object in the definition.py + # "step" -> Goes through the range in steps (even on two dimensions) + # requires a "step_explorer" object in the definition.py "execution_strategy": "", # We can install a preprocessor like Spark to reduce data volume @@ -125,3 +127,16 @@ "sample_size": 1000, } ] + +# If we use ExecutionStrategy "sequential" -> +step_explorer = { + # If new changes are not instantly visible, we want to ignore some results after state changes + "ignore_first_n_results": 10, + # How many samples of data to receive for one run + "sample_size": 10, + # The variables to modify + "knobs": { + # defines a [from-to] interval and step + "variable_name": ([0.0, 0.4], 0.1), + } +}