dojo - Training classes for functions and algorithms¶

class dojo.Dojo(algorithms : sequence, environ : object[, runs=100])¶

A simple testing class for competing algorithms. Dojo is the base class for concrete implementations to measure the performance of algorithms that shall perform the same task.

algorithms¶: The list of algorithms to compare

environ¶: An arbeitrary object that simulates the execution environment for the algorithms. A copy of it will be passed as first argument to each algorithm.

Note

The copy will be created via copy.deepcopy() to ensure that (hopefully) no value of the original environment will be modified by an algorithm.

runs¶: The number of consecutive runs for each algorithm.

train(*args) → None¶

Executes the algorithms, comparing their performance.

This has to be implemented by inheriting classes.

class dojo.TimingDojo(algorithms : sequence, environ : object[, runs=100])¶

A Dojo implementation measuring the execution time of its algorithms.

train(*args) → object¶

Executes the algorithms and compares their run-time performance Dojo.runs is used to create a reliable average mean for the execution time and hence should not be chosen too small.

The method will return the best performing algorithm.

class dojo.FitnessDojo(algorithms : sequence, environ : object[, runs=100[, cmpfunc=min]])¶

A Dojo implementation measuring the fitness of its algorithms. The fitness is determined by the passed cmpfunc.

cmpfunc¶

Comparision function for the fitness measurement. It has to return a single object of a passed iterable and must accept a named argument key, since its execution looks like

dojo.cmpfunc(result_dict, key=result_dict.get)

train(*args) → object¶: Executes the algorithms and compares their return value, which must be a float. Dojo.runs is used to create a reliable average mean for the return value and hence should not be chosen too small.