mlp

mlp(
    @activation 2
    @batchsize 50
    @hiddenlayers 3 3
    @learnrate 0.01
    @maxiter 1000
    @momentum 0.9
    @tapin 0
    @tapout -1
    @outputactivation 0
    @validation 0.2
    @mode 0
    @useseed 0
) -> llll

Generates a Multilayer Perceptron Neural Network object for either regression or classification tasks.

---

Arguments

• @activation [int]: Activation function for hidden layers. (default: 2).
  • 0: Identity (-∞ to ∞—i.e., no activation).
  • 1: Sigmoid (0 to 1).
  • 2: ReLU. (0 to ∞).
  • 3: Tanh (-1 to 1).
• @batchsize [int]: Batch size. (default: 50).
• @hiddenlayers [list/int]: A list of integers, each specifying the desired size of every fully connected, hidden layer. (default: 3 3).
• @learnrate [float]: Learning rate. (default: 0.01).
• @maxiter [int]: Maximum number of iterations during training. (default: 1000).
• @momentum [float]: A scalar that applies a portion of previous adjustments to a current adjustment being made by the neural network during training. This has many benefits—an intuitive one is that even if some data point(s) suggest the neural network make some drastic, unique adjustments in a new direction, the history of adjustments from other points (it’s “momentum”) will prevent it from going too far off course of where most of the data is directing it. (default: 0.9).
• @tapin [int]: The index of the layer to use as input to the neural network for predict (0-based). The default of 0 is the first layer (the original input layer), 1 is the first hidden layer, etc. This can be used to access different parts of a trained neural network such as the encoder or decoder of an autoencoder. (default: 0).
• @tapout [int]: The index of the layer to use as output to the neural network for predict (0-based). The default of -1 is the last layer (the original output layer). This can be used to access different parts of a trained neural network such as the encoder or decoder of an autoencoder. (default: -1).
• @outputactivation [int]: Activation function for output layers. Ignored when @mode is 1. (default: 0).
  • 0: Identity (-∞ to ∞—i.e., no activation).
  • 1: Sigmoid (0 to 1).
  • 2: ReLU. (0 to ∞).
  • 3: Tanh (-1 to 1).
• @validation [float]: Factor of the dataset to use for validation during training. (default: 0.2).
• @mode [int]: MLP prediction mode, determining the type of set required during the fit process, as well as the type of set returned during prediction. (default: 0).
  • 0: Regression—meant to be fitted with a dataset as an output, via the fit function.
  • 1: Classification—meant to be fitted with a labelset as an output, via the fit function.
• @useseed [int]: Use random seed for parameter initialization. (default: 0).
  • 0: Off
  • 1: On

---

Output

Multilayer perceptron object [llll]

---

Usage

$indata = null;
$outdata = null;
## generate basic dataset based on "less-than" function
for $i in 1...100 do (
    $a = rand(0, 1);
    $b = rand(0, 1);
    $indata _= [$a $b]; ## input point
    $outdata _= [$a < $b] ## expected output (0 or 1)
);
$indataset = dataset($indata); ## input points
$outdataset = dataset($outdata); ## expected outputs
$model = mlp(@outputactivation 1); ## create mlp model
## fit (i.e., train) model to learn input/output mappping from dataset
for $i in 1...10 do ( ## repeat training to minimize loss
    print(fit($model, $indataset, $outdataset), 'Loss:') 
);
writeobject($model, "./mlp.json"); ## write as JSON for future use (optional)
$model = readobject("./mlp.json"); ## read pre-trained model from JSON (optional)
$xpoint = 0.25 0.75; ## sample point
$pred = predict($model, $xpoint); ## generate prediction
print($pred, "prediction:") ## should be (almost) 1.0, as 0.25 < 0.75 is true