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

 

Batch Details
Batch run name german_3
Description German 3 on Atlas
This report file german_3/german_3_report.html
data file /home/cssip/jsherrah/EPrep3/data/german_3.dat
Time of completion Sun Jun 21 18:22:58 1998
Duration of Batch 27 hours, 33 minutes, 45 seconds
Random Seed 2192058883 (from clock)
Average generations per run 29.60
Average failed feature creations per run 6868.00
Average fitness evalutions per run 93901.30
Test Set Improvement 0.80 %
 
 

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) 26.20 32.80 29.20
Min. Distance to Means(MDTM) 39.60 38.00 37.20

  EPrep Best-of-run Classification Errors (%)
Run
Training
Validation
Test
McNemar confidence
# Features
# Inputs
# Nodes
Classifier
1 29.00 27.60 27.60 0.652 1 11 283 MDTM
2 25.80 24.80 27.20 0.716 12 15 138 ML
3 30.00 26.40 30.00 0.417 12 16 984 MDTM
4 28.80 25.60 29.20 0.500 3 11 343 ML
5 27.40 24.80 26.40 0.761 11 16 716 MDTM
6 20.00 23.60 24.40 0.974 17 17 248 ML
7 29.80 26.00 26.40 0.831 8 8 19 ML
8 25.00 22.80 28.40 0.599 12 19 432 ML
9 27.20 24.80 28.40 0.606 11 11 19 ML
10 18.00 24.80 26.80 0.782 20 20 881 ML
Ave. 26.10 (3.89 ) 25.12 (1.30 ) 27.48 (1.53 ) 0.68 (0.16 ) 10.70 (5.40 ) 14.40 (3.75 ) 406.30(327.40) ML

  Confusion matrix for Best Ever Individual from run 8
Class
Predicted
Total
1
2
Ground Truth
1
143 32 175
2
39 36 75
Total
182 68 250

Average Operator Probabilities
Operator Average Probability
Delete-Feature Mutation 0.099
Add-Feature Mutation 0.104
Hoist Mutation 0.109
Truncate Mutation 0.103
Swap Mutation 0.098
One-Symbol Mutation 0.089
All-Nodes Mutation 0.090
One-Node Mutation 0.101
Grow Mutation 0.099
High-Level Crossover 0.109
 

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

Related Data Files
Description Filename
german_3_gen.m Matlab plot generation function
german_3_bogf_ave.{eps,gif} Best-of-generation Fitness, averaged over 10 runs
german_3_bogv_ave.{eps,gif} Best-of-generation Validation Set Error, averaged over 10 runs
german_3_avef_ave.{eps,gif} Average fitness, averaged over 10 runs
german_3_stdf_ave.{eps,gif} Standard deviation of fitness, averaged over 10 runs
german_3_nftr_ave.{eps,gif} Average number of features per individual, averaged over 10 runs
german_3_nnode_ave.{eps,gif} Average number of nodes per individual, averaged over 10 runs
german_3_nint_ave.{eps,gif} Average number of introns per individual, averaged over 10 runs
german_3_ntrl_ave.{eps,gif} Average number of RAT trials per individual, averaged over 10 runs
german_3_optimp_ave.{eps,gif} Average improvement in fitness due to optimisation, averaged over 10 runs
german_3_opprob_ave.{eps,gif} Average probability of each genetic operator, averaged over 10 runs
german_3_run_x.dat Binary data file containing results of run x (read by Matlab functions)
german_3_bor_x.prep 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-run individual for run x
german_3_run_x.corr 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Feature correlation file for run x
german_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
german_3_bogf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Fitness for run x
german_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
german_3_avef_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average fitness for run x
german_3_stdf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Standard deviation of fitness for run x
german_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
german_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
german_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
german_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
german_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
german_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