The Evolutionary Pre-Processor (version 3.0)
Batch Run Report

 

Batch Details
Batch run name balance_3
Description Balance 3 on Nyquist
This report file final/balance_3/balance_3_report.html
data file /home/cssip/jsherrah/EPrep3/data/balance_3.dat
Time of completion Tue May 26 21:49:23 1998
Duration of Batch 0 hours, 21 minutes, 45 seconds
Random Seed 18446744071572555808 (from clock)
Average generations per run 9.50
Average failed feature creations per run 267.30
Average fitness evalutions per run 30236.30
Test Set Improvement 24.20 %
 
 

Data Partition
 
Class Training Validation Test Total
0 144 72 72 288
1 24 12 13 49
2 144 72 72 288
Total 312 156 157 625

Summary of Results

  Original Classification Errors (%)
Classifier Training Validation Test
Parallelepiped(PPD) 53.85 53.85 54.14
Min. Distance to Means(MDTM) 27.56 25.00 24.84

  EPrep Best-of-run Classification Errors (%)
Run
Training
Validation
Test
McNemar confidence
# Features
# Inputs
# Nodes
Classifier
1 15.38 15.38 22.93 0.382 1 4 31 PPD
2 0.00 0.64 0.00 1.000 1 4 7 PPD
3 13.14 13.46 48.41 0.000 3 4 85 PPD
4 0.00 0.00 1.27 1.000 1 4 10 PPD
5 15.38 16.03 15.92 0.875 2 4 66 MDTM
6 0.00 0.00 0.00 1.000 1 4 8 PPD
7 0.00 0.00 0.64 1.000 1 4 7 PPD
8 4.17 5.13 24.20 0.440 2 4 64 PPD
9 0.00 0.00 1.27 1.000 1 4 7 PPD
10 21.47 25.64 28.66 0.003 4 4 92 PPD
Ave. 6.96 (8.01 ) 7.63 (8.81 ) 14.33 (15.75 ) 0.67 (0.40 ) 1.70 (1.00 ) 4.00 (0.00 ) 37.70 (33.47 ) PPD

  Confusion matrix for Best Ever Individual from run 7
Class
Predicted
Total
1
2
3
Ground Truth
1
72 0 0 72
2
0 13 0 13
3
0 1 71 72
Total
72 14 71 157

Average Operator Probabilities
Operator Average Probability
Delete-Feature Mutation 0.100
Add-Feature Mutation 0.102
Hoist Mutation 0.099
Truncate Mutation 0.098
Swap Mutation 0.100
One-Symbol Mutation 0.101
All-Nodes Mutation 0.099
One-Node Mutation 0.100
Grow Mutation 0.101
High-Level Crossover 0.100
 

Number of Run Terminations attributed to each Criterion
Termination Criterion Number of Terminations
TP Criterion 0
GL Criterion 6
Max. Generations 0
Client Abort 0
Zero Validation Error 4
Total 10
 
 

Related Data Files
Description Filename
balance_3_gen.m Matlab plot generation function
balance_3_bogf_ave.{eps,gif} Best-of-generation Fitness, averaged over 10 runs
balance_3_bogv_ave.{eps,gif} Best-of-generation Validation Set Error, averaged over 10 runs
balance_3_avef_ave.{eps,gif} Average fitness, averaged over 10 runs
balance_3_stdf_ave.{eps,gif} Standard deviation of fitness, averaged over 10 runs
balance_3_nftr_ave.{eps,gif} Average number of features per individual, averaged over 10 runs
balance_3_nnode_ave.{eps,gif} Average number of nodes per individual, averaged over 10 runs
balance_3_nint_ave.{eps,gif} Average number of introns per individual, averaged over 10 runs
balance_3_ntrl_ave.{eps,gif} Average number of RAT trials per individual, averaged over 10 runs
balance_3_optimp_ave.{eps,gif} Average improvement in fitness due to optimisation, averaged over 10 runs
balance_3_opprob_ave.{eps,gif} Average probability of each genetic operator, averaged over 10 runs
balance_3_run_x.dat Binary data file containing results of run x (read by Matlab functions)
balance_3_bor_x.prep 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-run individual for run x
balance_3_run_x.corr 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Feature correlation file for run x
balance_3_tst_bor_x.pred 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Test set predictions for best-of-run individual from run x
balance_3_bogf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Fitness for run x
balance_3_bogv_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Validation Set Error for run x
balance_3_avef_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average fitness for run x
balance_3_stdf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Standard deviation of fitness for run x
balance_3_nftr_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of features per individual for run x
balance_3_nnode_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of nodes per individual for run x
balance_3_nint_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of introns per individual for run x
balance_3_ntrl_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of RAT trials per individual for run x
balance_3_optimp_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average improvement in fitness due to optimisation for run x
balance_3_opprob_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average probability of each genetic operator for run x