<< percentile CHE3007S plotall >>

CHE3007S >> CHE3007S > permutation_importance

permutation_importance

Function to determine permutation importance, and optionally plot results. based on R2 value.

Syntax

[pij, R2] = permutation_importance(x, y, W, repeats)
[pij, R2] = permutation_importance(x, y, W, repeats, sf)
[pij, R2] = permutation_importance(x, y, W, repeats, sf, af)
[pij, R2] = permutation_importance(x, y, W, repeats, sf, af, fig)
[pij, R2] = permutation_importance(x, y, W, repeats, sf, af, fig, labels)

Parameters

pij:

n x 1 vector describing overall variable importance

R2:

m x repeats matrix of resulting R2 values

x:

n x m matrix of input parameters, each column represents a different parameter

y:

n x p matrix of target values, each column represents a different variable

W:

hypermatrix of weights and biases resulting from trained neural network

repeats:

integer value determining how many permutations to execute

sf:

string variable identifing simulation function, default 'ann_FFBP_run'

af:

matrix of strings describing activation function for each layer incl output, default = ['ann_tansig_activ','ann_purelin_activ']

fig:

boolean variable indicating if plots should be made or suppressed, default %f

labels:

row vector of strings describing the input variables

Description

Function to investigate permutation variable importance of trained neural network


Report an issue
<< percentile CHE3007S plotall >>