NNdatasets: All Datasets in One List

In pkR-pkR/NNbenchmark2019: Datasets and Functions to Benchmark Neural Network Packages

Description

NNdatasets is a list with the 12 datasets presented in this package and the recommanded number of hidden neurons for each neural network model.

• mDette: 5 neurons.

• mFriedman: 5 neurons.

• mIshigami: 10 neurons.

• mRef153: 3 neurons.

• uDmod1: 6 neurons.

• uDmod2: 5 neurons.

• uDreyfus1: 3 neurons.

• uDreyfus2: 3 neurons.

• uGauss1: 5 neurons.

• uGauss2: 4 neurons.

• uGauss3: 4 neurons.

• uNeuroOne: 2 neurons.

Each item of the list is itself a list with 5 components:

• ds: character. The name of the dataset.

• neur: integer. The recommanded number of hidden neurons in the NN model and in fmlaNN.

• nparNN: integer. The number of parameters in fmlaNN.

• fmlaNN: the formula of the corresponding neural network, with tanh() as the activation function in the hidden layer.

• Z: matrix or data.frame. The dataset itself.

Using attach() and detach() gives a direct access to these items.

Examples

ht(NNdatasets, n = 2, l = 6)

NNdataSummary(NNdatasets)

pkR-pkR/NNbenchmark2019 documentation built on June 22, 2020, 12:10 a.m.