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

 

Batch Details
Batch run name german_2
Description German 2 on Atlas
This report file final/german_2/german_2_report.html
data file /home/cssip/jsherrah/EPrep3/data/german_2.dat
Time of completion Tue Jun 2 11:57:00 1998
Duration of Batch 13 hours, 49 minutes, 6 seconds
Random Seed 2861081218 (from clock)
Average generations per run 21.30
Average failed feature creations per run 23592.80
Average fitness evalutions per run 174387.30
Test Set Improvement 1.60 %
 
 

Data Partition
 
Class Training Validation Test Total
0 350 175 175 700
1 150 75 75 300
Total 500 250 250 1000

Summary of Results

  Original Classification Errors (%)
Classifier Training Validation Test
Maximum Likelihood(ML) 16.80 30.00 28.80
Min. Distance to Means(MDTM) 40.40 33.20 41.20

  EPrep Best-of-run Classification Errors (%)
Run
Training
Validation
Test
McNemar confidence
# Features
# Inputs
# Nodes
Classifier
1 26.20 29.60 34.00 0.067 6 16 219 ML
2 23.40 22.80 28.00 0.618 14 15 32 ML
3 24.80 22.80 28.80 0.500 9 13 400 ML
4 19.80 24.00 27.20 0.729 16 17 149 ML
5 22.60 21.20 27.20 0.691 16 14 173 ML
6 22.80 23.60 30.00 0.359 16 19 356 ML
7 21.80 21.60 28.40 0.560 13 13 13 ML
8 22.20 24.80 30.40 0.248 14 14 14 ML
9 21.20 23.20 28.80 0.500 15 16 36 ML
10 20.60 25.20 26.80 0.770 17 17 94 ML
Ave. 22.54 (1.82 ) 23.88 (2.25 ) 28.96 (2.02 ) 0.50 (0.21 ) 13.60 (3.32 ) 15.40 (1.85 ) 148.60(133.24) ML

  Confusion matrix for Best Ever Individual from run 5
Class
Predicted
Total
1
2
Ground Truth
1
150 25 175
2
43 32 75
Total
193 57 250

Average Operator Probabilities
Operator Average Probability
Delete-Feature Mutation 0.095
Add-Feature Mutation 0.103
Hoist Mutation 0.098
Truncate Mutation 0.104
Swap Mutation 0.101
One-Symbol Mutation 0.097
All-Nodes Mutation 0.100
One-Node Mutation 0.101
Grow Mutation 0.099
High-Level Crossover 0.103
 

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

Related Data Files
Description Filename
german_2_gen.m Matlab plot generation function
german_2_bogf_ave.{eps,gif} Best-of-generation Fitness, averaged over 10 runs
german_2_bogv_ave.{eps,gif} Best-of-generation Validation Set Error, averaged over 10 runs
german_2_avef_ave.{eps,gif} Average fitness, averaged over 10 runs
german_2_stdf_ave.{eps,gif} Standard deviation of fitness, averaged over 10 runs
german_2_nftr_ave.{eps,gif} Average number of features per individual, averaged over 10 runs
german_2_nnode_ave.{eps,gif} Average number of nodes per individual, averaged over 10 runs
german_2_nint_ave.{eps,gif} Average number of introns per individual, averaged over 10 runs
german_2_ntrl_ave.{eps,gif} Average number of RAT trials per individual, averaged over 10 runs
german_2_optimp_ave.{eps,gif} Average improvement in fitness due to optimisation, averaged over 10 runs
german_2_opprob_ave.{eps,gif} Average probability of each genetic operator, averaged over 10 runs
german_2_run_x.dat Binary data file containing results of run x (read by Matlab functions)
german_2_bor_x.prep 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-run individual for run x
german_2_run_x.corr 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Feature correlation file for run x
german_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
german_2_bogf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Fitness for run x
german_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
german_2_avef_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average fitness for run x
german_2_stdf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Standard deviation of fitness for run x
german_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
german_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
german_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
german_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
german_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
german_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