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

 

Batch Details
Batch run name satimage_2
Description Satimage 2 on Cortex
This report file final/satimage_2/satimage_2_report.html
data file /home/cssip/jsherrah/EPrep3/data/satimage_2.dat
Time of completion Mon Jun 8 10:14:24 1998
Duration of Batch 95 hours, 51 minutes, 6 seconds
Random Seed 2720038922 (from clock)
Average generations per run 50.70
Average failed feature creations per run 6915.60
Average fitness evalutions per run 84420.90
Test Set Improvement 7.20 %
 
 

Data Partition
 
Class Training Validation Test Total
0 766 384 383 1533
1 351 175 177 703
2 679 340 339 1358
3 313 157 156 626
4 354 175 178 707
5 754 377 377 1508
Total 3217 1608 1610 6435

Summary of Results

  Original Classification Errors (%)
Classifier Training Validation Test
Parallelepiped(PPD) 36.28 35.39 37.95
Min. Distance to Means(MDTM) 22.44 20.65 23.66

  EPrep Best-of-run Classification Errors (%)
Run
Training
Validation
Test
McNemar confidence
# Features
# Inputs
# Nodes
Classifier
1 16.44 15.98 17.27 1.000 13 27 75 MDTM
2 21.45 19.96 23.04 0.500 30 30 30 MDTM
3 17.44 16.73 17.89 1.000 19 36 4581 MDTM
4 15.82 14.74 16.46 1.000 15 18 60 MDTM
5 17.90 17.79 21.55 0.278 15 36 365 PPD
6 21.48 20.90 21.68 0.579 4 4 7 MDTM
7 21.45 19.96 23.04 0.500 30 30 30 MDTM
8 16.66 15.49 16.71 1.000 13 14 38 MDTM
9 19.49 21.33 24.72 0.000 26 36 1896 PPD
10 16.47 15.67 17.52 1.000 13 19 64 MDTM
Ave. 18.46 (2.18 ) 17.85 (2.35 ) 19.99 (2.96 ) 0.69 (0.35 ) 17.80 (8.01 ) 25.00 (10.32 ) 714.60(1400.85) MDTM

  Confusion matrix for Best Ever Individual from run 4
Class
Predicted
Total
1
2
3
4
5
6
Ground Truth
1
363 0 10 4 6 0 383
2
1 147 0 0 29 0 177
3
2 0 298 35 1 3 339
4
0 0 30 102 2 22 156
5
3 0 2 8 144 21 178
6
0 0 6 70 10 291 377
Total
369 147 346 219 192 337 1610

Average Operator Probabilities
Operator Average Probability
Delete-Feature Mutation 0.101
Add-Feature Mutation 0.090
Hoist Mutation 0.090
Truncate Mutation 0.086
Swap Mutation 0.108
One-Symbol Mutation 0.096
All-Nodes Mutation 0.093
One-Node Mutation 0.112
Grow Mutation 0.107
High-Level Crossover 0.116
 

Number of Run Terminations attributed to each Criterion
Termination Criterion Number of Terminations
TP Criterion 0
GL Criterion 1
Max. Generations 9
Client Abort 0
Zero Validation Error 0
Total 10
 
 

Related Data Files
Description Filename
satimage_2_gen.m Matlab plot generation function
satimage_2_bogf_ave.{eps,gif} Best-of-generation Fitness, averaged over 10 runs
satimage_2_bogv_ave.{eps,gif} Best-of-generation Validation Set Error, averaged over 10 runs
satimage_2_avef_ave.{eps,gif} Average fitness, averaged over 10 runs
satimage_2_stdf_ave.{eps,gif} Standard deviation of fitness, averaged over 10 runs
satimage_2_nftr_ave.{eps,gif} Average number of features per individual, averaged over 10 runs
satimage_2_nnode_ave.{eps,gif} Average number of nodes per individual, averaged over 10 runs
satimage_2_nint_ave.{eps,gif} Average number of introns per individual, averaged over 10 runs
satimage_2_ntrl_ave.{eps,gif} Average number of RAT trials per individual, averaged over 10 runs
satimage_2_optimp_ave.{eps,gif} Average improvement in fitness due to optimisation, averaged over 10 runs
satimage_2_opprob_ave.{eps,gif} Average probability of each genetic operator, averaged over 10 runs
satimage_2_run_x.dat Binary data file containing results of run x (read by Matlab functions)
satimage_2_bor_x.prep 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-run individual for run x
satimage_2_run_x.corr 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Feature correlation file for run x
satimage_2_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
satimage_2_bogf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Fitness for run x
satimage_2_bogv_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Validation Set Error for run x
satimage_2_avef_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average fitness for run x
satimage_2_stdf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Standard deviation of fitness for run x
satimage_2_nftr_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of features per individual for run x
satimage_2_nnode_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of nodes per individual for run x
satimage_2_nint_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of introns per individual for run x
satimage_2_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
satimage_2_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
satimage_2_opprob_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average probability of each genetic operator for run x