Next:
Contents
Contents
METAL
The
METAL Machine Learning Experimentation Environment
V3.0 (
METAL-MLEE
)
Manual - Version 3.0
Johann Petrak
johann@ai.univie.ac.at
Austrian Research Institute for Artificial Intelligence
Contents
Introduction
Installation
System Requirements
Software Requirements
Obtaining and Installing the Program
Source code by other authors included in the package
What
METAL-MLEE
Does
Standard Database Format
Names File
Data File
Formal Format Description
The Programs
Main experimentation program:
run_exp
Synopsis
Important Options
Specifying a CPU time limit
Specifying learning algorithms
Passing parameters to the learning algorithms
Algorithm interface programs
Extracting information from the results:
parse_results
Synopsis and options
Normalize time measurements:
parse_times
Synopsis and options
Checking the database format:
check_database.pl
Synopsis and options
Checking experiment output:
check_results.pl
Synopsis and options
Other programs and helper files included in the distribution
Calculate quick measures from names files:
parse_names
Select a subset of features:
project
Sample interface scripts
The Clementine command line interface:
run_clem
Calculate landmarks:
landmark.pl
Adapting
METAL-MLEE
Adding Learning Algorithm Interface Scripts
Adding Preprocessing Algorithms
Running Experiments
Structure and Organization of Output Data
The log file
The
results
file
The
.stats
file
The
.dct
file
The targets files
The prediction files
Solutions to frequent problems
The experiment fails and I cannot figure out why?
A CPU timeout was specified, but the algorithms run much longer?
A CPU timeout was specified, but the algorithm seems to never stop?
It seems for some of the folds there are empty train/test files?
Glossary of Frequently Used and Exotic Terms
Bibliography
About this document ...
2002-10-17
