\MRNN

Most Retarded Neural Network ever. Yep, with single neuron.

Synopsis

class MRNN {

• // members
• protected float $weight = 0.01;
• protected integer $lastError = 1;
• protected float $smoothing = 0.0001;
• protected float $actualResult = 0.01;
• protected integer $correction = 0;
• protected integer $epoch = 0;
• protected array $trainStats = ;
• protected integer $statEvery = 5000;
• protected bool $debug = false;
• protected string $activationFunction = 'def';
• // methods
• public void __construct()
• public void setWeight()
• protected void setActivationFunc()
• public float processInputData()
• public float restoreInputData()
• protected float sigmoid()
• protected float unsigmoid()
• protected void train()
• protected bool learn()
• public bool learnDataSet()
• public array getTrainStats()
• public float getWeight()
• protected float getLastError()
• protected float getEpoch()
• public void setDebug()
• public string visualizeTrain()
}


Members

protected

• $activationFunction — string
  Contains network activation function type
• $actualResult — float
  Training routine result
• $correction — float
  Contains current weight correction
• $debug — bool
  Output of debug messages due train progress
• $epoch — int
  Contains current training iteration
• $lastError — float
  Last neuron training error
• $smoothing — float
  Smoothing factor
• $statEvery — int
  Contains train stats multiplier
• $trainStats — array
  Contains current training stats as epoch=>error
• $weight — float
  Initial weight

Methods

protected

• getEpoch() — Returns current training epoch
• getLastError() — Returns current train last error
• learn() — Train neural network on some single input value
• setActivationFunc() — Sets network instance activation function type
• sigmoid() — Just native sigmoid function
• train() — Do the neuron train routine
• unsigmoid() — Inverse of native sigmoid function

public

• __construct() — What did you expect?
• getTrainStats() — Retrurns current network instance training stats
• getWeight() — Returns current neuron weight
• learnDataSet() — Performs training of neural network with
• processInputData() — Returns data output processed by trained neuron (forward)
• restoreInputData() — Returns data input processed by trained neuron (backward)
• setDebug() — Sets debug state of learning progress
• setWeight() — Sets neuron instance weight
• visualizeTrain() — Performs network training progress