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

 

Batch Details
Batch run name spirals_3
Description Spirals 3 on Kaiser
This report file spirals_3/spirals_3_report.html
data file /home/cssip/jsherrah/EPrep3/data/spirals_3.dat
Time of completion Sat Jun 20 22:03:39 1998
Duration of Batch 6 hours, 13 minutes, 14 seconds
Random Seed 2726987811 (from clock)
Average generations per run 51.30
Average failed feature creations per run 526.10
Average fitness evalutions per run 450262.40
Test Set Improvement 19.21 %
 
 

Data Partition
 
Class Training Validation Test Total
0 100 45 240 385
1 94 52 239 385
Total 194 97 479 770

Summary of Results

  Original Classification Errors (%)
Classifier Training Validation Test
Maximum Likelihood(ML) 44.33 49.48 51.15
Parallelepiped(PPD) 44.33 46.39 51.57
Min. Distance to Means(MDTM) 44.85 45.36 51.36

  EPrep Best-of-run Classification Errors (%)
Run
Training
Validation
Test
McNemar confidence
# Features
# Inputs
# Nodes
Classifier
1 18.56 19.59 32.15 1.000 2 2 101 ML
2 21.13 29.90 27.14 1.000 2 2 48 ML
3 17.53 27.84 24.01 1.000 2 2 21 ML
4 20.62 30.93 30.27 1.000 2 2 70 ML
5 23.71 32.99 28.81 1.000 2 2 30 ML
6 20.62 25.77 20.46 1.000 2 2 25 ML
7 31.44 34.02 36.33 1.000 2 2 33 ML
8 21.65 31.96 37.16 1.000 2 2 10 ML
9 35.05 35.05 30.48 1.000 1 1 14 ML
10 15.98 28.87 28.39 1.000 2 2 24 ML
Ave. 22.63 (5.76 ) 29.69 (4.32 ) 29.52 (4.83 ) 1.00 (0.00 ) 1.90 (0.30 ) 1.90 (0.30 ) 37.60 (26.82 ) ML

  Confusion matrix for Best Ever Individual from run 1
Class
Predicted
Total
1
2
Ground Truth
1
158 82 240
2
72 167 239
Total
230 249 479

Average Operator Probabilities
Operator Average Probability
Delete-Feature Mutation 0.072
Add-Feature Mutation 0.092
Hoist Mutation 0.078
Truncate Mutation 0.107
Swap Mutation 0.127
One-Symbol Mutation 0.092
All-Nodes Mutation 0.080
One-Node Mutation 0.097
Grow Mutation 0.103
High-Level Crossover 0.152
 

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

Related Data Files
Description Filename
spirals_3_gen.m Matlab plot generation function
spirals_3_bogf_ave.{eps,gif} Best-of-generation Fitness, averaged over 10 runs
spirals_3_bogv_ave.{eps,gif} Best-of-generation Validation Set Error, averaged over 10 runs
spirals_3_avef_ave.{eps,gif} Average fitness, averaged over 10 runs
spirals_3_stdf_ave.{eps,gif} Standard deviation of fitness, averaged over 10 runs
spirals_3_nftr_ave.{eps,gif} Average number of features per individual, averaged over 10 runs
spirals_3_nnode_ave.{eps,gif} Average number of nodes per individual, averaged over 10 runs
spirals_3_nint_ave.{eps,gif} Average number of introns per individual, averaged over 10 runs
spirals_3_ntrl_ave.{eps,gif} Average number of RAT trials per individual, averaged over 10 runs
spirals_3_optimp_ave.{eps,gif} Average improvement in fitness due to optimisation, averaged over 10 runs
spirals_3_opprob_ave.{eps,gif} Average probability of each genetic operator, averaged over 10 runs
spirals_3_run_x.dat Binary data file containing results of run x (read by Matlab functions)
spirals_3_bor_x.prep 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-run individual for run x
spirals_3_run_x.corr 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Feature correlation file for run x
spirals_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
spirals_3_bogf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Fitness for run x
spirals_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
spirals_3_avef_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average fitness for run x
spirals_3_stdf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Standard deviation of fitness for run x
spirals_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
spirals_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
spirals_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
spirals_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
spirals_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
spirals_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