Class Dimlp

Inheritance Relationships

Base Type

Derived Type

Class Documentation

class Dimlp : public BpNN

The Dimlp class represents a Discretized Interpretable Multi-Layer Perceptron.

This class provides functionality for training and evaluating a Dimlp neural network, which is a type of neural network that includes discretized levels for interpretability.

Subclassed by BagDimlp

Public Functions

inline float Error(DataSet &data, DataSet &target, float *accuracy)

Computes the error and accuracy of the Dimlp network on a given dataset.

Parameters:

  • data – The dataset to evaluate.

  • target – The target values corresponding to the dataset.

  • accuracy – Pointer to store the computed accuracy.

Returns:

The computed error.

inline void Train(DataSet &train, DataSet &trainTar, DataSet &test, DataSet &testTar, DataSet &valid, DataSet &validTar, const std::string &accuracyFile, bool fromBT = false)

Trains the Dimlp network using the provided datasets.

Parameters:

  • train – The training dataset.

  • trainTar – The target values for the training dataset.

  • test – The testing dataset.

  • testTar – The target values for the testing dataset.

  • valid – The validation dataset.

  • validTar – The target values for the validation dataset.

  • accuracyFile – The file to save accuracy metrics.

  • fromBT – Flag indicating if training is done with bagging.

~Dimlp() override = default

Virtual destructor for Dimlp.

Dimlp(float eta, float mu, float flat, float errParam, float accuracyParam, float deltaErrParam, int discrLevels, int showErrParam, int nbEpochsParam, int nbLayers, const std::vector<int> &nbNeurons, const std::string &weightFile, int seed = 0)

Constructs a Dimlp network with specified parameters.

Parameters:

  • eta – Learning rate.

  • mu – Momentum.

  • flat – Flatness factor.

  • errParam – Error threshold.

  • accuracyParam – Accuracy threshold.

  • deltaErrParam – Delta error threshold.

  • discrLevels – Number of discretized levels.

  • showErrParam – Show error parameter.

  • nbEpochsParam – Number of epochs.

  • nbLayers – Number of layers in the network.

  • nbNeurons – Number of neurons in each layer.

  • weightFile – File to save weights.

  • seed – Seed for random number generation (default: 0).

Dimlp(const std::string &readFile, int nbLayers, const std::vector<int> &nbNeurons, int discrLevels, int netId = 1)

Constructs a Dimlp network by reading weights from a file.

Parameters:

  • readFile – File to read weights from.

  • nbLayers – Number of layers in the network.

  • nbNeurons – Number of neurons in each layer.

  • discrLevels – Number of discretized levels.

  • netId – Network ID (default: 1).

Dimlp(const std::string &readFile, float eta, float mu, float flat, float errParam, float accuracyParam, float deltaErrParam, int discrLevels, int showErrParam, int nbEpochsParam, int nbLayers, const std::vector<int> &nbNeurons, const std::string &weightFile, int seed = 0)

Constructs a Dimlp network by reading weights from a file and sets training parameters.

Parameters:

  • readFile – File to read weights from.

  • eta – Learning rate.

  • mu – Momentum.

  • flat – Flatness factor.

  • errParam – Error threshold.

  • accuracyParam – Accuracy threshold.

  • deltaErrParam – Delta error threshold.

  • discrLevels – Number of discretized levels.

  • showErrParam – Show error parameter.

  • nbEpochsParam – Number of epochs.

  • nbLayers – Number of layers in the network.

  • nbNeurons – Number of neurons in each layer.

  • weightFile – File to save weights.

  • seed – Seed for random number generation (default: 0).