Package jaga.experiment

Interface Summary
ConfigurableExperiment Allows data to be fed in to this experiment after being constructed.
ConfigurableRandomInputExperiment Used for experiments where the randomness of the input sequence can be controlled for example so taht every generation the same input sequence is used for all individuals but this is modified for the next generation.
Experiment An Experiment is a purpose we wish the individuals to perform, such as being an AND gate, or surviving in a tank full of piranhas.
FitnessFunction A fitness function must provide a fitness value given the inputs, outputs and desired outputs of an individual evaluation.
TestPatternGenerator Used to plug-in to experiments so that their generateInput method can be configurable.

Class Summary
CompleteShuffledTPG Will generate a test pattern with all 2^I input vectors (where I is the nr.
ConfigurableRandomInputMultiOutputExperiment Multiple Output Experiments allow many experiments sharing inputs to be bundled into the same one.
CorrelationFitnessFunction Given desired output samples and actual output samples will calculate fitness as the statistical correlation between them.
ExperimentLib Provides common functionality used by many Experiment imlementations.
MultiOutputExperiment Multiple Output Experiments allow many experiments sharing inputs to be bundled into the same one.
MultipleSampleAvgFitnessFunction This Wrapper fitness function extracts multiple output samples from the actual output by averaging values at certain points.
NRepeatsTPG This Pattern Generator generates random input vectors at each step until all possible 2^I (where I is the nr.
PerRowFitnessFunction Creates a single output series whose value is 0 for rows where every output in Q is equal to those in desQ, and a 1 otherwise.
PessimisticFitnessFunction If a test pattern of length TP is applied containing only V unique test vectors, then this fitness function (FF) wrapper will provide a test pattern of length V to its internal FF with a 1 where an output was equal to the desired at every application of the current test vector and a 0 if at any application of that vector it was different.
SimpleCorrelationFitnessFunction A variant of the CorrelationFitnessFunction, it requires less computation.
SingleNoisySizeSampleAvgFitnessFunction Similar to the SingleSampleAvgFitnessFunction but with a varying size window.
SingleSampleAvgFitnessFunction A single sample at the end of each session while inputs are applied is used as the ouptput for that session, instead of using all inputSampleSeparation bits.
SumFitnessFunction This fitness function calculates fitness as (correct bits/total bits) averaged over all outputs.